Party Agent Reports vs Official Results: Where the Numbers Didn't Match in 2022

Every polling station in Kenya has at least two sets of eyes on the count: the IEBC's presiding officer, who fills out Form 34A, and the party agents, who independently record the results. In theory, these two records should match exactly. In practice, they often don't.

During the 2022 general election, discrepancies between party agent reports and official Form 34A results were flagged at an estimated 2,300 polling stations — roughly 5% of Kenya's 46,229 stations. The vast majority of these discrepancies were minor arithmetic errors or transcription mistakes. But the pattern reveals important weaknesses in Kenya's electoral architecture.

How Discrepancies Happen

A discrepancy between agent reports and official results can arise from several sources:

1. Arithmetic errors on Form 34A: The presiding officer hand-writes results. If the total doesn't add up correctly — votes for candidates plus rejected ballots doesn't equal total votes cast — a discrepancy exists. In 2022, arithmetic errors on Form 34A were found in approximately 9,800 forms (21% of all forms), according to analysis by the Kenya Elections Observatory.
2. Transcription errors: A presiding officer may correctly count the votes but write the wrong number on the form — swapping two digits, for example (writing 231 instead of 321)
3. Agent recording errors: The agent may misread the form, mishear the announcement, or make their own arithmetic mistakes when phoning in results
4. Legitimate recounts: If a recount occurs at the polling station, the final Form 34A may differ from what the agent recorded during the initial count
5. Deliberate tampering: In rare cases, a presiding officer may alter the form after agents have recorded the initial results

The Scale of the Problem

Of the estimated 2,300 stations with agent-reported discrepancies:

• ~1,400 stations (61%): Discrepancies of fewer than 10 votes — almost certainly arithmetic or transcription errors
• ~580 stations (25%): Discrepancies of 10-50 votes — could be errors or could indicate more systematic issues
• ~230 stations (10%): Discrepancies of 50-200 votes — concerning and warranting investigation
• ~90 stations (4%): Discrepancies exceeding 200 votes — significant mismatches requiring explanation

The 90 stations with discrepancies exceeding 200 votes are the most concerning. These represent cases where the agent's tally differed substantially from the official result. A 200-vote discrepancy at a single station, in a constituency race decided by 500 votes, can change an outcome.

Geographic Patterns

Discrepancies were not evenly distributed across the country. Certain regions showed higher rates:

Higher discrepancy rates (above 7% of stations):

• Parts of Nairobi (particularly Embakasi constituencies) — high-density stations with rushed counting processes
• Mombasa (especially Changamwe and Likoni) — areas where the gubernatorial race was particularly contested
• Bungoma and Trans Nzoia — multiple competitive races created complex tallying environments
• Narok — scattered stations in remote pastoral areas with limited oversight

Lower discrepancy rates (below 3% of stations):

• Central Kenya (Kiambu, Murang'a, Nyeri) — well-organized party agent structures and less competitive presidential race at the local level
• Nyanza (Kisumu, Siaya, Homa Bay) — strong single-party dominance meant fewer contested tallies
• Uasin Gishu and Nandi — strong UDA organization ensured tight agent management

The pattern suggests that discrepancies are more common where races are competitive and counting processes are under pressure, rather than in areas of single-party dominance where the outcome is not in dispute.

The Form 34A Problem

At the heart of many discrepancies is the design of Form 34A itself. The form requires presiding officers to:

1. Write each candidate's name and votes in words and figures
2. Total the valid votes
3. Record rejected ballots
4. Record disputed ballots
5. Calculate the total ballots cast
6. Record the total number of registered voters at the station
7. Have agents sign the form

This is done by hand, often under poor lighting conditions (many stations have no electricity), after a long day that began at 5 AM. The presiding officer may have been working for 18 hours when they fill out this form. Under these conditions, arithmetic errors are not surprising — they're predictable.

The fact that 21% of Form 34As contained arithmetic errors doesn't necessarily indicate fraud. It indicates a system that relies on human calculation under exhausting conditions. This is why electronic verification — cross-checking the hand-written form against an electronically tabulated result — is so important.

What Agents Can and Cannot Do

Under Kenyan election law, party agents have significant rights at the polling station:

• They can observe the entire voting and counting process
• They receive a signed copy of Form 34A
• They can challenge the count and request a recount (which the presiding officer may grant or deny)
• They can file complaints at the constituency tallying center

What they cannot do:

• They cannot handle ballots
• They cannot stop the counting process
• They cannot refuse to sign Form 34A (though they can note objections)
• They cannot unilaterally demand a recount — the presiding officer has discretion

The effectiveness of party agents depends almost entirely on training and technology. A well-trained agent with a smartphone app that instantly transmits results to a central tally is a powerful check on the system. An untrained agent with no reporting mechanism is little more than a witness.

Lessons for 2027

The 2022 discrepancy data points to clear improvements needed:

1. Digital Form 34A: Moving to an electronic form that auto-calculates totals would eliminate arithmetic errors overnight. The technology exists — it's a question of implementation.
2. Agent technology: Equipping agents with apps that photograph Form 34A and auto-extract results using OCR would create an instant, verifiable parallel record.
3. Real-time comparison: Systems that automatically compare agent-reported results with IEBC-transmitted results — and flag discrepancies immediately — would force resolution while evidence is still fresh.
4. Better training: Both presiding officers and party agents need more rigorous training on form completion and result verification.

The 2,300 stations with discrepancies represent 5% of all stations. That percentage needs to be driven toward zero. Every discrepancy is a potential dispute, and every dispute erodes public confidence in the electoral process.

Your agents are your eyes and ears. Give them the right tools. Votrack's agent reporting platform ensures every Form 34A is photographed, results are transmitted instantly, and discrepancies are flagged in real time. Request a demo and never wonder about your numbers again.