How to read election results without confusing votes and coverage

Before comparing election numbers, establish what was counted, where it was counted and how complete the information is. A percentage with no denominator or a total with no reporting level can mislead even when the arithmetic is correct. The worked examples below are invented for explanation.

Attach a label to every number

Start with the election, office, constituency, reporting level and timestamp. Then distinguish registered voters, accredited voters, votes cast and valid votes. A party's score belongs to the election and geographical unit shown on its source record.

INEC's 2026 remarks on BVAS safeguards emphasise checks involving accreditation totals and entered scores. That is a useful reminder to read the supporting figures, not only the largest party number. An apparent inconsistency warrants checking the underlying record and any correction.

Know which denominator the percentage uses

Suppose an illustrative unit has 1,000 registered voters, 400 accredited voters, 390 votes cast and 380 valid votes. A candidate with 190 valid votes has 50% of the valid vote in that example. That is 19% of registered voters, a different calculation answering a different question.

Using votes cast divided by registered voters gives 39% turnout in this example. Using accredited voters divided by registered voters gives an accreditation rate of 40%. If a report uses “turnout” differently, ask for its formula before comparing it with another report.

• Candidate vote share: candidate votes divided by the stated vote total.
• Reporting coverage: units received divided by units expected at that level.
• Turnout or participation: check the numerator as well as the denominator.

Reporting coverage is not vote coverage

If results are available from 60 of 100 expected units, that is 60% of units. It does not establish that 60% of expected votes have been counted. Units can differ in registration size and participation.

The order in which places report can also affect the visible lead. A sample of available results is not automatically representative of the remaining areas. State what is missing and avoid projecting the remainder without a defensible method. A portal's upload count and a tally of readable, verified records may also differ.

Prevent double counting and silent revisions

Choose a consistent reporting level. Do not add polling-unit figures to a ward subtotal that already contains them. Give every row a unique location identifier and retain the source image or document reference.

When a figure changes, keep the earlier value, corrected value, reason and time. Describe a corrected transcription differently from a newly received result. Compare the change with the underlying official record and any published correction, rather than relying on an unattributed spreadsheet passed between chat groups.

Separate arithmetic from the outcome

You can check a sum without declaring a winner. Winning requirements and the completion of the official process are separate questions. Report a visible lead as a lead, explain its coverage, and identify an official declaration when one is made. This makes an update useful even while important results remain unavailable.