Joint estimation of insurance loss development factors using Bayesian hidden Markov models

This repository holds code and the workflow to build the paper:

Goold, C. 2025. (preprint). Joint estimation of insurance loss development factors using Bayesian hidden Markov models

A preprint is available on Arxiv here.

The code and data are available in the code/ and code/data/ directories, respectively.

To run the code, you will first need to install the requirements, ideally within a virtual environment. Using Python 3.10:

python3.10 -m venv .env
source .env/bin/activate
python3.10 -m pip install -r requirements.txt

You will also need a working version of CmdStan, which can be installed after the above requiremnets using (on Linux/MacOSX):

mkdir -p code/.cmdstan
install_cmdstan -d code/.cmdstan -v 2.36.0

See CmdStanPy's installation page for more information.

Each of the Python files in the code/ directory can be run as standalone modules, using:

python3.10 [file].py

replacing [file] with a particular filename. Results will automatically be saved out in the code/results/ directory.