A toolbox for macroeconometric modeling
Version 0.0.1
- High-level language for model description (parser based on Lark)
- backward looking modeling with AR / ECM processes
- Dulmage - Mendelsohn block decomposition of the model
- Symbolic computation of the jacobian
- Several choices of numerical solvers (based on Scipy, or high-order Newton methods)
- Time-series management based on Pandas
A macro model is defined by a set of static and dynamic equations, which determines the evolution of economic variables (such as GDP, interest rate, etc). The toolbox is able to simulate a trajectory (yearly or quarterly) of a model, based on a sample of time series (a training set). With this training set, the coefficients of the dynamic equations can be estimated, and the residuals of the model computed.
- Clone the repository
git clone https://github.com/InseeFrLab/Macronometrics.git
- Install the package
cd Macronometrics
python setup.py install
- Clone the repository containing an illustrative model
cd ..
git clone https://github.com/InseeFrLab/Macronometrics-Notebook.git
- Run the Jupyter notebook
Colibri.ipynb
- Numba just-in-time compilation of the solving functions
- Estimation of the coefficients of the model (OLS)
The code for Dulmage - Mendelsohn block decomposition is implemented with courtesy of Bank of Japan research team :
Hirakata, N., K. Kazutoshi, A. Kanafuji, Y. Kido, Y. Kishaba, T. Murakoshi, and T. Shinohara (2019) "The Quarterly Japanese Economic Model (Q-JEM): 2019 version" Bank of Japan Working Paper Series, No. 19-E-7.
Some features of the toolbox are inspired from the Grocer package for Scilab, and implemented with courtesy of Eric Dubois, lead developer of Grocer : http://grocer.toolbox.free.fr/
Institut National de la Statistique et des Etudes Economiques
Direction des Etudes et Synthèses Economiques
Département des Etudes Economiques
Division des Etudes Macroéconomiques
Alexandre Bourgeois - Benjamin Favetto (@BFavetto) - Adrien Lagouge - Matthieu Lequien (@MLequien) - Olivier Simon