Testing the gradient predictions of the monocentric city model in Addis Ababa

Note

This repo contains replication code and data for the paper Beze (2024).

The real estate dataset is available on Zenodo. For more details, see the data section below.

Requirements

Tip

The order in which the scripts should be run is provided in script/main.sh.

Expand

• R 4.3.3

The necessary R packages are listed in the renv.lock file. You can install them by running the following command in the R console:

# renv::init() # to initialize renv on the project if you don't clone the repo
renv::restore()

• Python 3.12

The necessary Python packages are listed in the requirements.txt file. You can install them with uv:

uv pip install -r requirements.txt

Data

The data used in the analysis constitutes two main parts: real estate data and building footprint data.

Housing data

Important

Data availability

The dataset has been published on Zenodo and can be accessed here.

Variable description

var	description	group	remark
id	ID of the property (prepended with the provider name)		The ID uniquely identifies properties; in the raw data, it may not have been, even within a provider.
listing_type	Listing type (for rent or sale etc.)	listing and property types	Parsed if not provided
property_type	Property type (house, apartment, etc.)	listing and property types	Parsed if not provided
price	Price of the property in local currency (Ethiopian Birr (ETB))	price	Other currency units are converted to ETB
price_type	The type of price (fixed, negotiable, etc.)	price	Parsed if not provided
price_adj	Price of the property adjusted for inflation	price
price_sqm	Price of the property per square meter	price
price_adj_sqm	Price of the property per square meter adjusted for inflation	price
size_sqm	Floor area of the property in square meters	size	Imputed if not provided
size_sqm_is_imputed	Yes if the floor area of the property was imputed	size
plot_size	Lot size of the property in square meters	size
address	Address of the property (untouched as provided)	address
address_main	Address of the property (manually corrected or cleaned)	address	The address of the property has been manually corrected or cleaned. Addresses for properties have been manually extracted from the description of the property.
address_alt	Address of the property (extracted with Gemini Pro)	address	Equals to address_main if extraction failed or null
unique_address_grp	Address group counter	address	This variable identifies properties with the same addresses.
place_name	The name of the geocoded place, from the geocoding api,address	address
place_id	The id of the geocoded place	address
subcity	The subcity name	address
lng	The longitude of the property location	address
lat	The latitude of the property location	address
is_lng_lat_sampled	Yes if lng,lat is sampled	address	When the address is broad like “Bole” or even “Addis Ababa” a random (lng,lat) can be sampled from the subcity or Addis polygons.”
date_published	The date the property was published on the website	time
time	The month (formatted year-month-01) the property was published on the website	time
year	The year the property was published on the website	time
quarter	The quarter the property was published on the website	time
title	The title of the property ad	description
description	The description of the property ad	description
num_bedrooms	The number of bedrooms in the property	features
num_bathrooms	The number of bathrooms in the property	features
num_images	The number of images in the property ad	features
features	A list of additional features of the property	features	A list of additional features, unstructured.
condition	The condition of the property	features
furnishing	The furnishing level of the property	features	E.g. fully furnished, semi-furnished, etc.
pets	Yes if pets are allowed in the property	features	Applicable to rentals. Parsed if not provided
floor	The floor location of the property	features	Applicable to apartments. It may refer to the number of floors in some cases.
garden	Yes if the property has a garden	features	Parsed if not provided
parking	Yes if the property has parking	features	Parsed if not provided
kitchen	Yes if the property has a kitchen	features	Parsed if not provided
elevator	Yes if the property has an elevator	features	Parsed if not provided
balcony	Yes if the property has a balcony	features	Parsed if not provided
water	Yes if the property has water	features	Parsed if not provided
power	Yes if the property has electricity	features	Parsed if not provided
seller_address	The address of the seller mentioned in the ad		Phone number, email or social media information about the seller/agent.
dist_meskel_square	The distance from the property location to the CBD (Meskel Square) in km	Distance to the CBD
dist_arat_kilo	The distance from the property location to the CBD (Arat Kilo) in km	Distance to the CBD
dist_piassa	The distance from the property location to the CBD (Piassa) in km	Distance to the CBD
exchange_rate	Monthly Birr to USD exchange rates		Source: National Bank of Ethiopia
misclassified_or_outliers_flag	Yes if the property’s listing or type are thought to be misclassified or outlier.

If you want to reproduce the data using the scripts, you can follow the steps in script/main.sh.

If you run the scripts successfully, you will have: The primary dataset for the analysis is constructed from data/housing/processed/listings_cleaned.csv, a cleaned version of the scraped data from all providers. The raw data is available in data/housing/raw for the providers included in the analysis. Missing attributes in the dataset are imputed using Gemini Pro, and the imputed data can be found in data/housing/processed/structured/tidy. Finally, property addresses are geocoded using Google Places API and OSM nominatim. The georeferenced data is available in data/housing/processed/tidy/listings_cleaned_tidy__geocoded.csv.

Important

During web scraping, I tried to respect the robots.txt file of the website. See the contents in data/housing/robots_txt.

A list of real estate providers in Addis

name	num_ads
Loozap Ethiopia	75358
Cari Africa Homes	42612
AfroTie	30000
JIji	12272
Qefira	8121
Ethiopia Property Centre	3649
Engocha	2059
Real Ethio	1585
Airbnb Addis Ababa	1000
EthiopianHome	990
Ethiopian Properties	880
Sarrbet	741
Ethiopia Realty	717
Ermithe Ethiopia	645
LiveEthio	625
ZeGebeya.com	560
Zerzir	539
Real Addis	513
Beten	495
Kemezor	434
HahuZon	400
Ethiobetoch	315
Verenda	285
Mondinion	268
Yegna Home	247
Expat	233
Keys to Addis	219
Ebuy	216
Addis Agents	195
Rent in Addis Agent	175
Betoch	126
Sheger Home	120
Ethio Broker	105
Betbegara	83
Addis Property Listings	76
Shega Home	60
Realtor Ethiopia	33
Addis Gojo	32
Notes: The number of ads is as of April 2024. Qefira shut down in June 2023.

Building footprint datasets

The building variables are extracted from two sources:

• The German Aerospace Center (DLR): the World Settlement Footprint (WSF) 3D and WSF 2019v1 datasets.
• Open buildings from Google.

Citation

Please cite the paper or dataset for any use of the code or data in this repository.

@article{Beze_2024,
  title = {Testing the Gradient Predictions of the Monocentric City Model in Addis Ababa},
  ISSN = {1556-5068},
  url = {https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4803607},
  DOI = {10.2139/ssrn.4803607},
  journal = {SSRN Electronic Journal},
  publisher = {Elsevier BV},
  author = {Beze,  Eyayaw},
  year = {2024}
}

@misc{Beze_2024_dataset,
  title = {Georeferenced real estate data for Addis Ababa},
  author = {Beze,  Eyayaw},
  year = {2024},
  doi = {10.5281/ZENODO.11205969},
  url = {https://zenodo.org/doi/10.5281/zenodo.11205969},
  publisher = {Zenodo},
  copyright = {Creative Commons Attribution 4.0 International}
}