This repository accompanies the paper Markowitz Portfolio Construction at Seventy. It contains a reference implementation of the Markowitz portfolio optimization problem and the data used in the paper. Please note that the tickers of the stocks have been obfuscated to comply with the data provider's terms of use.
Please run all experiments using
make experiments
This first installs uv on the fly and then creates and replicates the environment defined in the script section of the 'experiments.py' file as described in the documentation of uv run
If you have uv installed you can bypass make and run directly
uv run experiments.py
The script section is as of today
# /// script
# requires-python = ">=3.12"
# dependencies = [
# "mosek==11.0.27",
# "loguru==0.7.3",
# "numpy==2.3.2",
# "pandas[output-formatting]==2.3.1",
# "matplotlib==3.10.5",
# "cvxpy-base==1.7.1",
# "clarabel==0.11.1"
# ]
# ///
A large fraction of our experiments have been performed using MOSEK as the underlying solver. We assume a valid license for MOSEK is installed. If not, you may want to apply for a Trial License
If you want to reference our work in your research, please consider using the following BibTeX for the citation:
@article{boyd2024markowitz,
title={Markowitz Portfolio Construction at Seventy},
author={S. Boyd and K. Johansson and R. Kahn and P. Schiele and T. Schmelzer},
journal={Journal of Portfolio Management},
volume={50},
number={8},
pages={117--160},
year={2024}
}
or for the arXiv version:
@misc{boyd2024markowitz,
title={Markowitz Portfolio Construction at Seventy},
author={Stephen Boyd and Kasper Johansson and Ronald Kahn and Philipp Schiele and Thomas Schmelzer},
year={2024},
doi = {10.48550/arXiv.2401.05080},
url = {https://arxiv.org/abs/2401.05080}
}