← All posts

What international election observers actually watch for

International election monitors spend weeks inside a country's polling stations, tallying rooms, and result-transmission lines — and what they find reveals exactly which parts of any election system are impossible to verify from the outside.

It is the evening of November 28, 2011, and somewhere in Kinshasa, a results sheet is being handed to a clerk at the Commission Électorale Nationale Indépendante. The Democratic Republic of the Congo has just held a presidential election across a country the size of Western Europe, with some polling stations accessible only by river or footpath. Millions of voters stood in lines that sometimes lasted hours. And now the sheets are arriving — thousands of them, from hundreds of jurisdictions, compiled locally, then regionally, then nationally.

The European Union Election Observation Mission is watching. And what it eventually writes down is damning.

Not because it found someone stuffing a ballot box. Because it couldn't find anything at all.

The Observers' Job Is Not to Say Whether They Trust the Result

This is the first thing most people get wrong about international election monitoring. Observers don't verify the count. They observe the conditions under which the count happens. They walk through polling stations, sit in tallying rooms, photograph result sheets, and track whether the rules that are supposed to make an election checkable are actually being followed.

When those conditions are absent, the result cannot be trusted — not because fraud is proven, but because it cannot be disproven.

The EU Election Observation Mission's final report on the DRC's 2011 elections is a document worth reading slowly. It describes a results compilation and publication process that "lacked transparency and credibility." It notes problems with the voter register. It records that the electoral commission, the CENI, operated without adequate visibility into how local and national tallies were assembled. And it notes something that cuts especially deep: the absence of a functioning Constitutional Court meant that disputes over results had nowhere to go. Observers were denied access to the relevant decisions. The adjudication process itself — the institution meant to be the last resort — was opaque.

The full EU EOM final report is available here.

The mission's conclusion wasn't "we found fraud." It was: "we cannot confirm this result is a valid reflection of what voters decided." That distinction matters enormously. It means the burden isn't on critics to prove something went wrong. The burden is on the system to produce evidence that it didn't.

The Six Checkpoints Monitors Actually Assess

When an international mission deploys — whether the EU EOM, the OSCE/ODIHR, or another recognized body — its observers are looking for something specific at every stage of the electoral cycle. Not impressions. Not vibes. Concrete, documentable conditions.

Here is what those checkpoints actually are.

1. Voter registration: is the list accurate and accessible? A register with phantoms (dead voters, duplicate entries, non-existent addresses) is an attack surface. Observers check whether registration was open, whether the list was published for public inspection, and whether challenges to the list could be heard and resolved through a documented process. In the DRC in 2011, the mission flagged problems with the voter register as a foundational issue — before a single ballot was cast.

2. Ballot security and chain of custody: where is the paper, and who touched it? From the printing facility to the polling station to the sealed box to the count room, observers track whether ballots and result sheets are handled by authorized people, in documented sequences, with tamper-evident packaging. A missing seal, an unsigned chain-of-custody form, or a ballot traveling with someone whose role doesn't authorize them to carry it — these are the precursors of unverifiable results. Austria's Constitutional Court annulled a presidential election in 2016 precisely because absentee-ballot envelopes had been opened by unauthorized people before the legally mandated time — no fraud proven, election void regardless, because the chain of custody was broken in ways that exceeded the winning margin.

3. Polling-day procedures: is the vote genuinely secret? This is where the OSCE/ODIHR findings get visceral. In Georgia's October 2024 parliamentary elections, observers recorded potential compromises to ballot secrecy in over 30 percent of their observations — some because of how ballots were inserted into boxes, some because of polling-station layouts, some because of how voters marked their ballots in view of others.

In Albania's April 2021 parliamentary elections, OSCE/ODIHR documented widespread allegations of vote-buying, leading to a number of investigations, and raised concerns about pressure on voters and the misuse of state resources.

These are not edge-case anomalies. They are documented, monitored, recurring failures in countries inside and adjacent to Europe.

The underlying mechanism matters: secrecy of the ballot is not a courtesy — it is the structural defence against bribery and coercion. A vote-buyer can only enforce the purchase if they can verify how the voter actually voted. A genuinely secret, unprovable ballot makes the transaction unenforceable. When monitors observe secrecy being compromised, they are watching the moment when vote-buying becomes economically viable.

4. Results compilation: can the numbers be traced from the polling station to the national tally? This is the step that failed most visibly in the DRC. Results sheets are supposed to travel upward through a verifiable chain: precinct level, provincial level, national level, each step signed, photographed, or transmitted in a way that allows any discrepancy to be caught and explained. When that chain is opaque — when compilation happens in rooms observers can't enter, or when results are announced before the underlying forms have been received and checked — there is no way to confirm the national total reflects what the individual sheets said.

Kenya's Supreme Court annulled its 2017 presidential election on exactly this basis. Not all polling-station result forms had been electronically and simultaneously transmitted to the national tallying centre as required by law. Result forms lacked consistent security features. The chairperson declared a winner before receiving all the underlying forms. The court held that the count itself was legally and democratically unsound — not because anyone proved the numbers were changed, but because the system didn't produce the evidence that would let anyone check.

5. Dispute resolution: is there an independent adjudicator with real power? An election with no credible dispute-resolution mechanism is an election where any problem, once it occurs, is permanent. The DRC had no functioning Constitutional Court in 2011. Observers were denied access to whatever decisions were made. There was no path by which a losing candidate or a citizen could point to evidence and have it examined under published rules by an independent body. This is not a procedural nicety. It is the last line of defense in a system that relies on the willingness of officials to follow the rules voluntarily.

6. Transparency of the entire process to observers and the public This is where the six checkpoints collapse into a single question: can anyone outside the system verify that it worked? Can a political party, a citizen, a journalist, or an international monitor sit down with the official numbers and trace them backward to signed, photographed, published precinct-level sheets? Or does verification require trusting the people running the count to have run it correctly?

The answer, in the DRC in 2011, was the latter. The answer, in Georgia in 2024, was partial at best in dozens of observed stations. The answer, in Albania in 2021, was compromised by a market for votes that observers could not quantify but also could not ignore.

What a Verifiable-by-Design System Would Make Trivially Checkable

Now here is the thing worth sitting with.

Every one of these six checkpoints — voter-list integrity, chain of custody, ballot secrecy, results compilation, dispute resolution, public transparency — describes information that exists at the moment an election happens. The number of ballots in a box is a fact. The name on a chain-of-custody form is a fact. The image of a result sheet is a fact. The transmission timestamp of a precinct's totals is a fact.

The problem is not that this information doesn't exist. The problem is that in most electoral systems, it exists inside a process that only the people running the process can see.

A system designed for observability from the ground up would change that architecture entirely. Each precinct's result, signed cryptographically by the officials who produced it and by the machines that counted it, published the moment it is finalized. Machine-readable, downloadable, reconcilable by anyone with a spreadsheet. Precinct-level data, not just county-level aggregates. Transmission logs with tamper-evident timestamps. Chain-of-custody records that don't require an observer's physical presence to verify.

This isn't science fiction. The EAC already urges officials to publish results in downloadable formats like .csv and .xml. States like Idaho already publish precinct-level results with raw data downloads. Colorado has run statewide risk-limiting audits since 2017 — statistical checks that provide defined confidence the reported winner actually won, sized to the margin of the race rather than a fixed percentage of ballots.

But the gap between "some jurisdictions do some of this" and "any independent observer worldwide can verify any step of any count" is still enormous.

In Kenya, the system was supposed to transmit form images electronically from every polling station in real time. The Supreme Court annulled the result because it hadn't actually done that consistently. The promise of observability and the architecture of observability are not the same thing.

The Difference Between "Observers Were Present" and "The Count Was Verifiable"

Here is the uncomfortable truth about international election observation: it is, by design, a sample.

An OSCE/ODIHR mission deploys hundreds of observers across thousands of polling stations. They visit a fraction of locations. They stay for hours, not all day. They observe counting procedures where they can, but they cannot be in every room where results sheets are compiled. When the EU EOM in the DRC reported that the results process "lacked transparency," it was reporting what it could see — which was, itself, insufficient evidence on which to pronounce the result valid.

That is not a criticism of the observers. It is a description of what observation can and cannot do.

Observer presence is a deterrent, not a guarantee. A polling station behaves differently when an international observer is sitting in the corner. But observers leave. The results sheets travel onward. The compilation happens in offices that observers may or may not enter. The national tally is announced by officials who may or may not have checked the underlying forms.

What would actually close this gap? A system where the underlying forms are themselves the public record — published, photographed, cryptographically signed, and downloadable — the moment each level of compilation is complete. A system where the observer's job is not to be physically present to witness the chain of custody, but to verify against a public, tamper-evident record that is accessible to anyone, anywhere, after the fact.

The German Federal Constitutional Court put the principle in plain language in 2009: an election is only legitimate when ordinary citizens, not just experts, can independently verify each essential step from ballot to result. That ruling concerned voting machines with no paper record — but the principle generalizes entirely. It applies to a results-compilation system that observers cannot enter. It applies to a transmission chain that leaves no tamper-evident log. It applies to any step in the process that asks you to trust the official account because the underlying evidence is not yours to check.

What "Credible" Actually Has to Mean

The EU EOM concluded that the DRC's 2011 results were "not credible." The OSCE/ODIHR found vote secrecy compromised in over 30 percent of observed polling stations in Georgia in 2024. Widespread vote-buying was documented in Albania in 2021. These findings exist in published, publicly available reports, sourced from trained observers operating under established methodologies.

They are claims. They are well-sourced claims. But they are also, structurally, the same kind of claim as an official government announcement that the election was free and fair. The official says: trust us. The monitor says: we observed these specific conditions. Neither gives you a system you can independently check.

The answer to both is the same: publish the evidence, not the conclusion.

Publish every precinct's result the moment it is final, in a format anyone can download and reconcile. Publish the transmission logs. Publish the chain-of-custody records. Publish the audit trail that lets a journalist in Brussels or a citizen in Kinshasa or a researcher in Tbilisi trace a national total back to a signed sheet from a single polling station.

When that infrastructure exists, international observers stop being the only people who can assess what happened. The public becomes the observer. Every political party becomes an auditor. Every journalist with a spreadsheet can cross-check the announced totals against the published precinct data.

That is what makes vote-buying harder — not because observers are watching, but because the transaction leaves a verifiable trace in a public record. That is what makes results-compilation fraud harder — not because officials are honest, but because any discrepancy between precinct totals and national totals is arithmetically visible. That is what makes the DRC scenario impossible — not because of better-intentioned officials, but because the record that would expose the manipulation is, by design, already public before anyone announces anything.

What Is Still Not Checkable — and What Would Fix It

Even with everything the EU EOM and OSCE/ODIHR found and published about the DRC in 2011, Albania in 2021, and Georgia in 2024, there are things that remain structurally unverifiable.

Whether a results sheet that was publicly photographed reflects what the actual ballot count produced — or whether the count itself was conducted honestly — cannot be confirmed without an independent audit of physical ballots against those sheets. Whether vote-buying in Albania changed the outcome, rather than merely occurring, cannot be determined from monitor observations alone. Whether the compromised ballot secrecy in Georgia affected individual races, or how many, is not calculable from aggregate observation rates.

These are not failures of the monitors. They are failures of the underlying systems to produce independently verifiable evidence in real time.

What would make them checkable? Precinct-level results published instantly and machine-readably, so anyone can aggregate and cross-check. Cryptographic signing of result transmissions, so alterations in transit are detectable. Physical paper ballots with voter-verifiable markings, audited against machine totals by statistical methods powerful enough to catch errors at the margin — like the risk-limiting audits Colorado has run since 2017. And a dispute-resolution process that operates on published records, not on the court's access to evidence it has to ask officials to produce.

An election no independent observer can fully verify in real time is an election that depends on trust. Trust is not verifiability. Trust is what you're forced to substitute when verifiability is absent.

The monitors have been telling us what the checkpoints are for decades. The technology to make every one of those checkpoints publicly observable already exists. The gap is not technical. It is architectural — a choice, in how systems are built, about who gets to see the evidence that democracy produced the right answer.

See how common these gaps are across the world →

Read the 2-minute version of why observability is the core problem →

Sources

See all postsExplore the gaps