Ethnic social origin is sensitive personal data under Kenya's Data Protection Act, 2019. DLA Piper's summary of section 2 says that the definition covers data revealing a person's race, health status, ethnic social origin, conscience, belief, genetic and biometric data, property details, marital status, family details, sex and sexual orientation. Bowmans says this category gets a higher level of protection. Anyone collecting data on communities should start here. This page isn't legal advice.
Key points:
- Ethnicity gets the higher protection.
- Guessing someone's ethnicity from a name creates the same risk.
- You need a clear lawful ground.
- Minimise what you collect, and secure it well.
Why does the law treat ethnicity differently?
Because misuse can cause serious harm, especially in Kenya. Article 31 of the Constitution protects the right to privacy, including the right not to have information relating to family or private affairs unnecessarily required or revealed.
Kenya's history shows what lists of communities can do.
What counts as sensitive personal data?
The definition in section 2 is above. Bowmans notes that it was widened to include property details, marital status and family details, including names of a person's children, parents and spouse.
So a survey that asks about family and land may well be collecting sensitive data, even if it never asks about ethnicity.
When can you process it?
Part V of the Act, from section 44, sets out the grounds for processing sensitive personal data. Read the Act itself, and take advice. A public body with a legal duty, such as NCIC running an ethnic audit, stands on different ground from an NGO or a company.
Is inferring ethnicity allowed?
Treat it as processing sensitive data. If a system guesses a person's community from a name, language or home area, it's producing data that reveals ethnic social origin. TI's pack treats inferred ethnicity as sensitive personal data, whatever the intent, and it doesn't allow it.
What should a survey do?
Ask yourself if you need the question at all. Often age, gender and area answer the research question. If you do need ethnicity, explain why, make it optional, and get clear consent.
Keep it apart from names and contacts. Report only in large groups. Think about what would happen if the file leaked.
What about consent?
Consent has to be informed and freely given, and people must be able to withdraw it. In a tense place, or with a government office present, a person may not feel free to refuse. Train interviewers to explain this.
What about NCIC audits?
NCIC collects staff ethnicity data to check the one-third rule, under its statutory role. Ethnic audits are a good example of a justified, focused purpose. Results are reported by institution and community, which is a decision for a commission with that legal duty. Others shouldn't assume that they can do the same.
What about employers?
Employers may need to report on staff under the National Cohesion and Integration Act. Keep such data apart from general HR files. Limit access to the people who need it. Use it for the purpose stated.
What about county governments?
Counties need to check compliance with the rule on hiring. Do it through a defined process, with clear ownership. Don't circulate spreadsheets. See ethnic balance in county jobs.
What about complaints?
A complainant may mention ethnicity in their own words. That's their choice, and it's part of their account. Keep it there, and don't turn it into a sortable field unless you have a clear reason and basis.
What about election campaigns?
Parties and candidates are especially tempted to use ethnic data for targeting. That raises serious legal and cohesion concerns. The electoral library's page on campaign conduct and data law covers this.
What about peace committees and Nyumba Kumi?
Household lists that note a family's community are dangerous. Don't keep them. Nyumba Kumi roles sets out safer practice.
What security should you have?
Encrypt data. Use passwords and separate logins. Keep access logs. Train staff. Avoid personal phones and shared drives. Plan for a breach, and know who you'd tell.
How long should you keep data?
Only as long as you need it. Set a date, and delete or anonymise after that. Old election data is a risk that doesn't earn its keep.
Should you do an impact assessment?
For work with sensitive data at scale, a data protection impact assessment is a sensible step. It makes you think through who could be harmed, and how to prevent it. Ask the Office of the Data Protection Commissioner for guidance.
Who is responsible?
Work out who's the controller and who's the processor. An institution that decides why and how data is used is responsible for it. A software provider that follows its instructions has its own duties. Put this in writing.
What rights do people have?
They can ask what you hold about them, correct it, object and ask for deletion, within the limits of the law. Tell them how. Treat requests seriously.
What's the practical test?
Ask this before you add any field: would I be comfortable if this appeared on the front page, with my organisation's name on it? If not, don't collect it.
What should a consent line say?
Keep it short and plain. Say who you are, why you're asking, who'll see the answer and how long you'll keep it. Say that the person can skip the question, and that skipping it costs them nothing.
A sample could read: "This question is optional. We ask it only to check that our service reaches everyone. Your answer is kept apart from your name, and only two staff can see it. We'll delete it after 12 months." If that wouldn't be true of your work, don't ask the question.
Read next
- Conflict early warning system for Kenya: the TI pack page for this work
- Cohesion monitoring on TI: all the pages in this section
- Monitoring cohesion without profiling
- Ethnic audits by NCIC
- The cohesion barometer
- Takwimu Intelligence (TI) home page
- National cohesion in Kenya: the guide to this library
- The TI platform and how we check our numbers
Where TI fits
Takwimu Intelligence (TI) holds consent and withdrawal on each respondent record, keeps identity apart from analysis, and treats the institution as controller of its own data.
To see it on your own work, book a walk-through.