Skip to content

guide

A cohesion barometer for Kenya: tracking the 2027 Roadmap

A barometer asks the same cohesion questions every quarter, so a commission or county can see whether its strategies are working. How to build one from NCIC's Roadmap.

Takwimu editorial team, Editorial team6 min read

A cohesion barometer is a survey asked the same way, in the same places, every few months. One reading tells you where things stand, and a series tells you whether a strategy is working. TI's barometer takes the five roadblocks in NCIC's Roadmap to a violence-free election and asks about each one by county. Four of its measures start from NCIC's own published consultation figures.

Key points:

  • It asks about trust, tolerance, youth, digital harm and violence.
  • Four indicators carry NCIC baselines.
  • Results are reported by area, never by community.
  • There's no combined score.

Why use a barometer?

A consultation run once produces a number. It can't tell you if that number is rising. The same instrument run each quarter produces a series, and a series is the only thing that can show if a strategy worked.

It also gives early warning. A sharp fall in trust in one county is a signal you can act on before an incident occurs.

What does it measure?

NCIC's Roadmap names five roadblocks, and each has a matching strategy: a trust deficit, political intolerance, youth vulnerability, negative use of digital technology, and an entrenched culture of violence. TI's barometer has 20 items built around those.

So a commission can report in the language of its own document.

What baselines exist?

NCIC's Roadmap consultation gives four useful starting figures. Three in four participants (75 percent) saw hate speech as a pressing threat, and 73 percent said the same of political intolerance. Some 69.3 percent believed gangs are hired by politicians. And 63.4 percent expected violence from unusual security deployments. Use them with care.

These came from a stakeholder consultation, and they aren't a national sample. A barometer can start from them, but it should state the difference clearly.

What questions work well?

Ask about what people have seen and felt. Do you trust the IEBC to run a fair election? Would you feel safe voting? Have you seen threats against any group in your area? Do you believe gangs are being hired here? Would you accept a neighbour from any community?

Keep the wording fixed from round to round. Even a small change can make a trend meaningless.

How do you sample?

Draw a sample that represents the area you want to report on. County results need a large enough sample in each county. Say what your margin of error is. In TI, county is the level that can be weighted today, since it has the reliable population figures.

Keep the method the same across rounds, and record any change. A panel, which returns to the same people, has different strengths from a fresh sample each time.

How often should it run?

Quarterly works in the run-up to an election. Monthly may help in the last months. Longer gaps suit quieter years. Match it to your budget and to the decisions it needs to inform.

How should sensitive questions be asked?

Don't ask for anyone's ethnicity unless it's truly needed. Don't ask who someone will vote for in a tense area. Ask about what's around them, as well as about themselves. Train interviewers. Let people skip any question.

Interviewers should reflect the area. Check conflict sensitivity before you start.

Why is there no overall score?

A combined index would be the first thing that any critic attacked, and the hardest to defend. Counts and shares are quotable. Report each indicator with its sample and date. If someone wants a composite, they should publish how it's built, as the SCORE index does.

How should results be reported?

Report by county, or by larger region if the numbers are small. Show the change since last time, with margins of error. Say what's a real change, and what might be noise.

Never publish results by community. Avoid ranking counties in a way that could shame them.

How can you read a change?

Look for movement outside the margin of error. Look at several questions together. A fall in trust along with a rise in fear means more than either alone. Then ask local contacts why.

What should happen after a reading?

Each round should end in decisions. Which county needs attention? Which strategy isn't working? Who will act? A barometer that doesn't change any decision is an expensive report.

Who should see it?

Share what's safe. A commission may publish national and county trends, while detailed tables stay with those who act on them. Partners can see what's relevant to them by role.

What does it cost?

Phone surveys cost less than face-to-face interviews, and they're faster, but they miss those without phones. Face-to-face suits hard-to-reach areas. Many teams mix the two. Check that you're comparing like with like across rounds.

A barometer tells you what people feel. Complaints tell you what's been reported. Media monitoring tells you what's being said. When all three point the same way, you've found something real.

Who should own the data?

The institution should. Survey data about public trust is sensitive, and it's a public asset. Make sure that any contract says who owns it, and where it's kept.

What does one round look like?

Fix the questions and the sample. Train interviewers, and test the tool in a local language. Collect the data over one or two weeks, so that a single news event doesn't skew it. Check the quality. Weight the results where the population figures allow it.

Then write a short note: what's changed, where, by how much, and what you think it means. Hold a meeting within a week, because results go stale quickly in an election year.

What should you do if results alarm you?

Check them first. Look at the sample in that county, and ask local contacts what's happening. If the change is real, tell those who can act, and hold back from publishing until you've thought about the effect of doing so.

Where TI fits

Takwimu Intelligence (TI) carries this barometer ready to run, with the Roadmap's roadblocks, NCIC's four baselines and county results.

To see it on your own work, book a walk-through.

Takwimu Intelligence (TI) is a conflict early warning system for Kenya.

Sources

  1. 1

    SCORE methodology, Bosnia and Herzegovina · Social Cohesion and Reconciliation Index, with UNDPOpened 5 November 2026.

More from the library