Comparing One Ward Across Two Elections Without Fooling Yourself

The same ward, two elections apart, is frequently not the same ward.

Comparing it to itself is the most natural thing in the world and the easiest place to go wrong, because everything about the comparison looks stable and almost nothing about it is.

Check the boundary before you check anything else

Kenya has 47 counties, 290 constituencies and 1,450 wards, and those units are subject to periodic boundary review by the IEBC.

If a ward's boundary was redrawn between the two elections you are comparing, the two result sets describe different populations wearing the same name. The votes did not move. The lines did. Related: why the boundaries you campaign in may be older than you think.

This is the first check because it is the only one that invalidates the comparison entirely. The others merely distort it.

The register changed even if the boundary did not

A register is not a fixed quantity between elections.

New voters enrol. Existing voters transfer to other stations. Some die. A ward can gain thousands of registered voters without gaining a single new resident, simply because a drive worked. Related: why registration and eligibility are separate numbers.

That means percentages and raw counts can point in opposite directions, and both can be honestly reported.

Compare shares, and then compare counts, and say which you mean

Do both, in that order, and never quietly switch between them.

Suppose a ward casts 6,000 votes in one election and 7,200 in the next. Suppose a party's candidate wins 3,000 votes in the first and 3,300 in the second. Raw votes are up by 300. Share has fallen from 50 per cent to about 46 per cent.

Somebody claiming growth is telling the truth. Somebody claiming decline is telling the truth. The honest sentence contains both halves: more people voted for this candidate, and a smaller proportion of the ward did.

Same office, same event, same source

Three things have to match, and they are easy to get wrong in a hurry.

Same office: an MCA result compared with an MCA result. Comparing a ward's MCA vote with the presidential vote cast in the same ward measures split-ticket behaviour, which is interesting and is not a swing.

Same event: a general election compared with a general election. A by-election is a different electorate on a different day with different stakes. Related: what a by-election can and cannot show.

Same source: station-level forms compared with station-level forms, or published official totals compared with published official totals. Mixing a provisional figure remembered from a broadcast with a final published figure produces a discrepancy that is entirely an artefact of your own filing.

Sources drift in a quieter way as well. A number copied from a summary two years ago may have been provisional at the moment it was copied, and nothing in the note you made will tell you that afterwards. Write the source and the date beside every figure at the moment you record it.

The candidate changed, and that is data too

A ward's numbers move for reasons that have nothing to do with mood.

A popular incumbent standing down, a former candidate switching parties, a strong local name entering or leaving, three candidates from one side splitting a vote that one candidate consolidated last time. Each of those moves a share substantially, and none of them is the electorate changing its mind about anything national.

Write the candidate list for both elections side by side before you interpret the numbers. Half the swings people announce dissolve at that step.

Party labels move too. A ward's vote can look as though it transferred between parties when the same individual simply stood under a different name, which says a great deal about that person and almost nothing about the ward.

The five-line note that makes the comparison honest

Before 10 August 2027, build one sheet per ward you care about and put five lines at the top of it.

Boundary: changed or unchanged between the two elections. Register: the size at each. Office: which of the six. Event type: general or by-election. Candidates: who stood each time.

Then, and only then, write down the numbers. A comparison with that header is something you can defend to a sceptical journalist or a hostile opponent. A comparison without it is a story you told yourself, and the ward will correct you on election night.