PolitIQ Kenya · Nairobi · 2026-08-19 · Information cutoff 2026-08-19
Scope. This note is about how protest is counted in Kenya and what the count supports. It uses published third-party data and our own aggregate collection. It makes no assessment of any named officeholder, and it identifies no participant.
Protest counts alone are very likely a poor measure of political pressure (moderate confidence). A count treats a peaceful county assembly and a dispersal with live rounds as one event each. The distribution of severity carries the signal; the total mostly carries the news cycle.
The state response is likely the more informative variable (moderate confidence). Amnesty International Kenya, Odipo Dev and VOCAL Africa, in the Kenya Freedom Index published in May 2026, document 1,002 protests over 2020–2025 alongside 138 deaths, 1,227 injuries and 2,128 arrests, with over 30% of protests dispersed using tear gas. The response figures separate periods that the protest count does not.
Concentration in Nairobi is real but overstated by collection (low confidence). The same index places 29% of protests in Nairobi against 5% in Nakuru and 4% in Mombasa. Media presence is itself concentrated in Nairobi, so some of that gap is coverage rather than incidence. We cannot currently separate the two, and we say so rather than report the ratio as settled.
Over the last 30 days our event detection recorded 43 county-attributed events across 15 counties. Their composition matters more than the total:
| Category | Severity | Events |
|---|---|---|
| Political | Low | 23 |
| Public order | High | 21 |
| Socio-economic | Low | 15 |
| Socio-economic | Medium | 3 |
| Political | Medium | 1 |
| Socio-economic | High | 1 |
Nearly every political event sits at low severity — rallies, tours, assemblies — while public-order events cluster at high. Counting events without weighting them would put the two on the same footing and describe a country that does not exist.
The reading we are arguing against is that a rising protest count signals rising instability. It is not unreasonable: mobilisation does precede escalation. But the same rise is produced by a widening of what gets reported — more outlets, more phone cameras, more civic documentation — with no change on the ground at all. Over 2020–2025 both were happening in Kenya at once. We do not think a count can separate them, which is why we weight severity and read the state response rather than the tally.
Third-party figures are from the Kenya Freedom Index (Amnesty International Kenya, Odipo Dev, VOCAL Africa; published 21 May 2026, covering 2020–2025). Our own figures are aggregates from our event detection over the trailing 30 days, derived from published news; they are not a census of protest, and small counties with thin media coverage are underrepresented. Our detection currently weights harm to persons but does not yet weight arrests separately, which means the state-response variable we argue for is one we do not yet fully measure ourselves. That work is in progress.
The standards this is written to are in our analytic standards; the pressure measure referred to above is described in how we measure political pressure.