This dataset is a harmonisation of the Domestic Electrical Load Survey (DELS) 1994-2014 dataset. The DELS 1994-2014 questionnaires were changed in 2000. Subsequently survey questions vary between 1994-1999 and 2000-2014. This makes data processing complex, as survey responses first need to be associated with their year of collection and corresponding questionnaire before they can be correctly interpreted. This Key Variables dataset is a user-friendly version of the original dataset. It contains household responses to the most important survey questions, as well as geographic and linking information that allows for the households to be matched to their respective electricity metering data. This dataset and similar custom datasets can be produced from the DELS 1994-2014 dataset with the python package delprocess. The data processing section includes a description of how this dataset was created. The development of the tools to create this dataset was funded by the South African National Energy Development Initiative (SANEDI).
Kind of Data
Unit of Analysis
v1: licensed data, available under conditions
The scope of the South African Domestic Electrical Load Survey includes:
HOUSEHOLD: Household characteristics, occupant demographics of gender and age, employment, income, number of household occupants with income
DWELLING: Dwelling characteristics, wall material, roof material, floor area, water source
APPLIANCES: stove, hotplate, kettle, heater, iron, geyser, washing machine, TV, fridge, freezer, microwave
BEHAVIOUR: frequency of appliance usage
ELECTRICITY: time since electrification, circuit breaker
The study had national coverage.
Settlement/suburb and municipality
The dataset covers South African households in the DELS 1994-2014 dataset. These are electrified households that received electricity either directly from Eskom or from their local municipality.
Producers and sponsors
University of Cape Town
South African National Energy Development Initiative
[Expert knowledge for] understanding database design and conveying details around data collection
The dataset includes all households for which survey responses have been captured in the DELS1994-2014 dataset.
Data is provided as is and no weightings have been applied.
Dates of Data Collection
Data Collection Mode
This dataset has been constructed from the DELS 1994-2014 dataset using the data processing functions in the delprocess python package (www.github.com/wiebket/delprocess: release v1.0). The delprocess python package takes the complexities of the original DELS 1994-2014 dataset into account and makes use of 'spec files' to specify the processing steps that must be performed. To retrieve data for all survey years, two separate spec files are required to process survey response from 1994-1999 and 2000-2014. The spec files used to produce this dataset are included in the program files and can be used as templates for new custom datasets. Full instructions on how to use them to process the data are in the README file contained in the delprocess package.
SPEC FILES specify the following processing steps:
1. List of search terms for which survey questions will be searched, and variables returned
2. Transformations (addition, subtraction, multiplication) of variables retrieved from search output
3. Bin intervals for variables (requires numeric data)
4. Lables for bins (requires binned data)
5. Details of bin segments
6. Replacement (encoding) of coded variable values
7. Higher level geography detail
In particular, the DELSKV 1994-2014 dataset has been produced by specifying the following processing steps:
* monthly_income from 1994 - 1999 is the variable returned by the 'income' search term
* monthly_income from 2000 - 2014 is calculated as the sum of the variables returned by the 'earn per month', 'money from small business' and 'external' search terms
* Appliance numbers from 1994 - 1999 is the count of appliances (no data was collected on broken appliances)
* Appliance numbers from 2000-2014 is the count of appliances [minus] the count of broken appliances (except for TV which included no information on broken appliances)
* A new total_adults variable was created by summing the number of all occupants (male and female) over 16 years old
* A new total_children variable was created by summing the number of all occupants (male and female) under 16 years old
* A new total_pensioners variable was created by summing the number of pensioners (male and female) over 16 years old
* A new total_unemployed variable was created by summing the number of unemployed occupants (male and female) over 16 years old
* A new total_part_time variable was created by summing the number of part time employed occupants (male and female) over 16 years old
* roof_material and wall_material values for 1994-1999 were augmented by 1
* water_access was transformed for 1994-1999 to be 4 [minus] the 'watersource' value
* Appliance usage values have been replaced with:
* water_access values have been replaced with:
3=tap in yard
4=tap inside house
* roof_material and wall_material values have been replaced with:
Appliance usage information was only collected after 2000.
No binning was done to segment survey responses for this dataset.
* The 2000-2014 survey questions contain no variable for 'number of females: 50+', which goes against the pattern of other occupant age categories.
* Spacing in the original questions is irregular and can cause challenges when specifying transformations (eg. 'number of males: 16-24' and 'number of males: 25 - 34', 'part time' and 'parttime').
* Spelling mistakes in the original questions can cause challenges when specifying transformations (eg. 'head emploed part time').
Missing values have not been replaced and are represented as blanks except for imputed columns (total_adults, total_children, ...) and appliances after 2000, where missing values have been replaced with a 0.
DataFirst Support Site
University of Cape Town
Licensed use files, available for non-commercial use only
Toussaint, Wiebke. Domestic Electrical Load Survey - Key Variables 1994-2014 [dataset]. version 1. Johannesburg: SANEDI [funders]. Cape Town: UCT Energy Research Centre [producers], 2014. Cape Town: DataFirst [distributor], 2019. DOI: https://doi.org/10.25828/mf8s-hh79