Contagious disruptions and complexity traps in economic development

This repository contains the code used to produce the figures and results in the paper

Charles D. Brummitt , Kenan Huremovic, Paolo Pin, Matthew H. Bonds, and Fernando Vega-Redondo. "Contagious disruptions and complexity traps in economic development". Working paper, 2017.

Below is an explanation of the contents of this repository and of the depencies of the code.

Scripts and notebooks

The folder scripts contains code and notebooks for replicating the results of the paper. Instructions for how to replicate each figure are given in the notebooks below. Wherever a long calculation is done, the results of the long calculation are provided in the folder scripts/simulated_data, and each notebook has code for both re-running the long calculation and for loading the results of the long calculation.

Mathematica notebooks and Wolfram Language Code

• figures_1_6.nb combines several datasets, runs some simulations of the model [for figure 6(b)], and creates figures 1 and 6;
• figures_3_4.nb creates the phase portraits in figures 3 and 4;
• figure5.nb creates the phase diagram in figure 5;
• figure2.key is a Keynote file containing figure 2;
• set_strategies_that_could_be_best_response.wl is a script in the Wolfram Language that computes the set of strategies that could be a best response and the chance of successfully producing. The notebook [Explore the set of strategies that could be a best response.nb](Explore the set of strategies that could be a best response.nb) contains a few examples of the set of strategies that could be a best response.

Dependencies: The Mathematica notebooks were made in version 11.0. They will work in version 10.4 as well. These notebooks can be viewed for free using the Wolfram CDF player, available on Windows, Mac, and Linux, and they can be run for free in the Wolfram Cloud.

Jupyter notebook and Python code

• The Jupyter notebook Figure_SI_1.ipynb contains the Python code used to create Figure SI-1. This notebook uses the following code written in Python for simulating the model:
  • ABM.py defines an economy and agents;
  • EconomySimulator.py (which is a wrapper for ABM.Economy that simulates the economy and collects information about it).

Dependencies: The Python code was run with Python 3.5.2, Matplotlib 2.0.0, Pandas 0.19.2, Seaborn 0.7.1, NumPy 1.11.3, progressbar 3.12.0, and joblib 0.10.3. The first five of these come with the installation of Anaconda; the latter two can be installed using pip by running pip install progressbar and pip install joblib from a terminal.

Figures

The folder figures contains the seven figures as PDF files. It also contains the figure compare_std_dev_F.pdf made at the end of the Jupyter notebook Figure_SI_1.ipynb, which illustrates the difference in standard deviations of 200 simulated time-series like those shown in Figure SI-1(b).

Empirical data

Empirical datasets are located in the folder empirical_data. This data is described, cleaned, and used in the notebook figures_1_6.nb, located in the scripts folder.

Simulated data

Data generated from simulating the model was exported into the folder scripts/simulated_data by the notebooks

• figures_3_4.nb
• figures_5.nb
• figures_1_6.nb
• Figure_SI_1.ipynb

All these notebooks are located in the scripts folder.