Inheritance simulation

An agent-based model simulating the UK population, focusing on the effects of inherited wealth.

Requirements

• Python 3.6
• Packages listed in requirements.txt

Running the simulation

Before your first run, initialise the data:

make data

or

python data.py

To run the simulation:

make run

or

python adapter.py model.yaml

Outputs

Running the simulation produces an output.h5 HDF5 file, containing the following tables:

• data: top-level data at each tick (year) of the simulation
• by_age: data subdivided by the age of the agents, at each tick
• history: full history of each agent through the simulation
• skill_dist: number of agents at each skill level at each tick
• tree: full list of family links between pairs of agents
• trusts: data on each inheritance trust at each tick

A function is provided for plotting the life history of an individual agent:

import pandas as pd
import sinks

data = pd.HDFStore('output.h5')

sinks.plot_history(data, 123)

Input parameters

Parameters of the model are set in model.yaml. Under control_parameters you can set the number of ticks and number of agents. The available model_parameters are:

• inheritance: type of inheritance, "direct" or "trust"
• trust_first_generation: first generation to benefit from trusts (1 = children)
• trust_last_generation: last generation to benefit from trusts (1 = children)
• assortative_mating: strength of assortative mating (0 = none, 1 = very strong)
• savings_rate: fraction of income that gets saved
• interest_rate: interest rate for wealth and trusts

Files

• README.md: this file
• requirements.txt: python packages required
• raw_data/*/*.xls: raw data from the ONS and census
• data.py: script for processing raw data into numpy arrays
• model.yaml: model specification
• inheritance.py: core model logic
• adapter.py: wrapper for inheritance.py that loads in the data and parameters
• scenarios.py: initialises agent population
• sinks.py: processes output data

Resources

PyData London 2017

• Slides: https://www.slideshare.net/Sandtable/forecasting-social-inequality-using-agentbased-modelling
• Video: https://www.youtube.com/watch?v=RglNX4c_dfc