For the 83% of Kenyans without a formal payslip, the SHA cannot simply look at a tax return. Instead, it uses a statistical “proxy”—a set of observable characteristics that correlate with wealth.
When you register on the Afya Yangu portal, the algorithm processes 43 specific data points to “guess” your household income.
The Proxy Variables include:
Housing Quality: Does your house have a thatched roof, iron sheets, or tiles? Are the walls mud, brick, or stone?
Utility Access: Do you have a token meter for electricity? Is your water from a borehole, a piped connection, or a river?
Household Composition: How many children or dependents are under your care? (Ironically, in some models, more children can lead to a higher “estimated” wealth based on consumption patterns).
Asset Ownership: Do you own a motorbike, a cow, or a television?
The “Error by Design” Controversy
The Africa Uncensored investigation highlighted a critical flaw: correlation is not causation.
“The system sees a stone house and assumes a high income, but it cannot see the debt taken to build it, or the fact that the breadwinner lost their job last month.” — John-Allan Namu, Investigative Journalist.
The Algorithm’s Two Big Mistakes:
Exclusion Errors: Truly poor families are flagged as “middle class” because they live in an inherited brick house, forcing them to pay premiums they cannot afford.
Inclusion Errors: Wealthy individuals with “minimalist” lifestyles or hidden assets may be assessed at the base rate of Ksh 300, depriving the fund of necessary revenue.
SHA’s Defense: The “Inexact Science”
Responding to these technical critiques, SHA officials argue that while PMT isn’t perfect, it is significantly better than the old system.
The Privacy Concern: Where is Your Data?
The MTI tool doesn’t just rely on what you tell it. Under the Digital Health Act, the SHA has the power to “ping” other government databases to verify your status.
KRA Data: To check for registered businesses or tax returns.
Lands Registry: To see if you own property.
NTSA: To check for vehicle ownership.
This “data mesh” is what allows the algorithm to create a financial profile of you in seconds. However, civil society groups like Article 19 have raised red flags over how this data is stored and whether Kenyans gave informed consent for their “proxies” to be used this way.
Final Thought: Can an Algorithm Have Empathy?
The Social Health Authority insists that the MTI will “learn and improve.” But for a family in rural Kenya being asked to pay Ksh 1,500 instead of Ksh 300 because they own a few goats, the “learning curve” feels like a threat to their survival.
