CRTA :Unified Voter Register audit – status and outlook before the election campaign

Conducting a comprehensive, system-level audit capable of establishing the overall accuracy, completeness, and integrity of the UVR is not feasible under current conditions. A meaningful set of partial audits and diagnostic tests remains possible, but their findings cannot be generalized into a reliable assessment of the register as a whole.

Despite some recent, but long overdue, partial progress, it is increasingly doubtful that even a minimally comprehensive initial assessment of the voter registration system can be completed by the 28 October 2026 reporting deadline. What can realistically be delivered by then should therefore be understood as a set of partial diagnostic findings, not as a completed audit capable of establishing the accuracy, completeness and integrity of the voter register.

The Commission had approximately seven months for substantive audit work, after the first two months were required for its expedited establishment and the adoption of the rules, work plan, and methodology. Of those seven months, approximately four were then consumed in securing access to basic voter-level UVR data and its history of changes, with additional time required to establish the technical capacity needed to process them.

Today, with the reporting deadline fast approaching, we can identify anomalies more easily, but we still cannot systematically determine what produced them. Without that understanding, we cannot reliably recommend how they should be addressed or identify the institutions and officials responsible for creating or failing to correct them.

The outcome of the audit is constrained by the following problems:

the Commission still lacks effective and reliable access to the source registers; without it, there is no systematic cross-checking of the accuracy of UVR data – the cooperation with the IT office as an imposed intermediary in this process is very slow, has not yielded any comprehensive results and does not allow the Commission to independently verify the completeness and integrity of the data provided;
access to voter-level UVR data and its history of changes was delayed by approximately four out of the seven months available for substantive audit work and was secured only after difficult and protracted negotiations;
a noticeable difference in the level of engagement and initiative among Commission members. In practice, most substantive audit activities and proposals to advance the process have so far been initiated by the members nominated by civil society and opposition groups, while participation from members nominated by the ruling parties has generally been limited and reactive;
significant time has been lost overcoming inadequate technical conditions, including computers and a secure analytical environment; and
In the first months of its work, Commission members were informally told by relevant officials from the Ministry and the Government that the Commission would not be permitted to conduct an IT audit of the UVR software and its underlying information system, despite the Commission’s clear legal mandate to do so. This effectively excluded from scrutiny an entire area in which vulnerabilities or interference could potentially compromise the integrity of the UVR.

The practical result of these constraints is that the Commission has ultimately been granted minimal access to avoid being formally excluded from the system, but insufficient access, time, and technical capacity to carry out even the core elements of the audit within the statutory deadline for its first report.

Under these conditions, the process cannot credibly fulfill the purpose of ODIHR’s longstanding recommendation for a meaningful audit of the voter and civil registers. It is also unlikely to produce, validate and implement recommendations capable of addressing identified weaknesses and restoring public confidence in the voter register.

What can realistically be delivered by October 28, 2026:

Statistical overview of the UVR and its main voter-data variables;
Assessment of voter-data quality and internal consistency (missing voter data, errors, inconsistencies);
Identification of apparent UVR software shortcomings that can contribute to registration errors;
Analysis of likely duplicate voter records;
Audit of the voter registers in the 10 local self-government units that held elections in March 2026;
Identification of locations and patterns indicating potentially illegitimate residence registration and organized voter migration;
Assessment of weaknesses in the legal and procedural framework for voter registration contributing to errors and illegitimate residence registration;
Analysis of discrepancies between addresses in the UVR and the official Address Register.

What cannot credibly be completed by October 28, 2026:

A comprehensive assessment of UVR accuracy through systematic cross-checking with the underlying civil-status, citizenship and residence registers. The Commission can perform limited individual checks against these registers, but the available service does not support systematic or large-scale verification.
Full source-level verification of individual voter records (historical data, applications, supporting documents, administrative checks, approvals and decisions).
Independent verification of voters’ status against other relevant state databases, including health insurance, pension and tax records. These sources remain unavailable to the Commission.
Attribution of potentially unlawful mass registrations to responsible officials or institutions. The Commission cannot systematically establish who initiated, processed, verified or approved individual citizenship or residence cases, whether within the MoI, MPALSG, MFA or another institution.
An IT and information-security audit of the UVR and the entry-points of the systems that feed it. Without such an audit, the Commission cannot determine whether access controls, audit logs, privilege management, database changes and software deployments provide adequate safeguards against unauthorized or untraceable modification of voter data.
Comprehensive and statistically robust list-to-people and people-to-list field tests, as envisaged in ODIHR voter-registration audit methodology, to assess whether registered voters actually exist and reside where recorded, and whether eligible citizens are missing from the register.
A systematic assessment of the professional and institutional capacity of voter-registration authorities, including recruitment, training, staffing, resources, internal controls and administrative inspection.

What these limitations mean in practice

The following examples illustrate the difference between identifying a suspicious pattern through reviewing UVR data and establishing what actually produced it.

Example 1 · Citizenship, residence and subsequent voter registration

A person who has lived in Republika Srpska acquires Serbian citizenship, shortly afterward registers residence in a Serbian municipality, and subsequently appears in the UVR.

Each individual step can be entirely lawful. UVR data alone cannot establish the legal basis on which citizenship was acquired, how the application was processed, what evidence supported the residence registration, whether the person actually established residence at the declared address, or whether the same sequence occurred in a coordinated manner for a larger group of people.

The available source-register service may confirm that the person appears in the citizenship register and show the person’s current registered residence. It still does not provide the historical and procedural information needed to reconstruct how and when these changes occurred.

The evidentiary threshold changes if records show, for example, that dozens of people were deliberately registered at the same residential address despite not actually residing there, particularly where the registrations were coordinated for electoral purposes. Such evidence would indicate manipulation rather than a simple statistical anomaly.

To establish that distinction, the Commission would need to connect UVR history with citizenship case files, residence-registration records, the evidence submitted in those procedures and, where necessary, field verification. Without those sources, the Commission may identify a pattern requiring investigation, but cannot establish how or why it was produced.

Example 2 · Responsibility for suspicious residence registrations

Suppose the Commission identifies a cluster of residence registrations that appears highly unusual in timing, concentration or structure.

The UVR may show the resulting registrations, and the source-register service may confirm the current residence of individual persons. Neither provides the complete administrative chain behind those registrations.

The Commission cannot systematically reconstruct who received the applications, what evidence was submitted, who verified that evidence, what checks were performed, who approved the registrations, or whether the same officials repeatedly processed related cases.

Without this chain of evidence, the Commission cannot reliably distinguish between administrative error, weak controls, individual misconduct and coordinated abuse, nor can it identify the officials or institutions responsible.

Example 3 · An implausibly old voter

A 108-year-old person remains registered as an active voter in the UVR.

The Commission can identify the record as highly unusual. It cannot determine from the UVR alone whether the person is alive, whether a death occurred abroad and was never transmitted to the Serbian civil registry, whether the civil registry itself contains an outdated record, or whether the apparent age results from an incorrect date of birth.

The same limitation applies to other categories of voters whose continued presence in the register may appear implausible. Without systematic large-scale cross-checking against the underlying civil registers and other relevant official records, the Commission can slowly identify cases requiring verification, but cannot establish their actual status.

Example 4 · An address that does not correspond to the Address Register

Suppose voters are registered at an address in the UVR that cannot be matched to the current official Address Register.

This may indicate an invalid registration, but several administrative explanations are also possible: a street may have been renamed, an address code may have changed, an older identifier may still be used in another register, or the linkage between the two systems may not have been updated.

Historical address identifiers and the underlying registration records are therefore necessary to determine whether the discrepancy reflects a legitimate administrative legacy or an invalid residence registration.

Example 5 · A change to voter data shortly before an election

A voter’s registered address changes shortly before an election. UVR history may show that the change occurred and record its date.

That information alone does not reveal the complete technical origin of the change. Without end-to-end system logs and source-system audit trails, the Commission may be unable to determine whether the change resulted from a lawful update in a source register, a manual action by an authorized official, an automated software process, or an unauthorized intervention.

This limitation applies to the integrity of the UVR as a whole. Independent verification requires the ability to reconstruct how every material change entered the system and who or what initiated it.

The central limitation

The Commission can increasingly identify suspicious records, inconsistencies and patterns in the UVR, but its ability to verify them against source records remains severely constrained.

This must be seen in light of the Commission’s experience in obtaining access to the UVR itself. Securing even basic voter-register data required sustained engagement and repeated escalation with the Speaker of the National Assembly, the Ministry of Public Administration and Local Self-Government, the Ministry of the Interior, the Secretary-General of the Government and the Office for IT and eGovernment. Access was ultimately provided four months after the deadline set in the Commission’s adopted work plan, and without the technical capacity needed to process the data, which required several additional weeks to establish.

Access to source registers remains similarly limited. The verification service allows only individual checks based on a known JMBG, rather than systematic searches or analysis of the underlying registers. The information returned is restricted, particularly for citizenship data, historical and procedural records are largely unavailable, and each individual check requires a time-consuming access procedure.

Access to source registers

Access to source registers: available, but severely constrained for audit purposes

The Commission has been provided with a service through which information on an individual voter can be checked against selected source registers, including civil registers, the citizenship register and the residence register. This provides a useful mechanism for verifying individual cases. Its design, however, places major limits on its use for a systematic audit of the UVR.

There are three main constraints.

1. Every search requires a known JMBG

The service can retrieve source-register information only for a specific person whose JMBG is already known. The Commission cannot search the underlying registers by name, address, date of birth, municipality or other variables, nor can it query or analyse the registers as datasets.

This creates a particular problem precisely for records that may contain identity-data errors. If a person has an incorrect JMBG in the UVR, or no JMBG recorded at all, the Commission cannot use that record to search the source registers and establish whether the correct person exists there.

The service therefore works well for the question: “What do the source registers contain for this particular JMBG?”

It cannot answer questions such as:

“Who appears in the source register but is missing from the UVR?”
“Which people registered at this address obtained residence during a particular period?”
“Which records contain inconsistent or missing JMBGs?”
“Which citizenship or residence records share a common administrative pattern (dual citizenship, date of registration…)?”

These are essential audit questions because they require searching and comparing populations of records rather than checking one already identified person at a time.

2. The information returned is substantially limited

Even when a person can be retrieved through JMBG, the Commission does not receive all information contained in the relevant source register.

For residence, for example, the service provides information on the person’s current registered residence, but not the history of previous residences. Such historical information exists within the administrative system and is available to authorized officials, but is not exposed to the Commission through the verification service. This prevents the Commission from systematically reconstructing residence changes over time, including changes immediately preceding elections.

The limitations are even greater for citizenship. The service essentially allows the Commission to establish whether a particular JMBG is or is not present in the citizenship register. It does not provide other information that may be essential for the audit, such as:

when Serbian citizenship was acquired;
on what legal basis it was acquired;
whether the person holds another citizenship;
when the person was entered into the citizenship register;
which authority processed or approved the case;
whether related applications were processed together or within the same period;
or what documentation and decisions supported the entry.

Consequently, the Commission can verify the existence of a citizenship record for a known individual, but cannot use the service to analyse the administrative process through which citizenship was acquired.

3. Verification is technically too slow for large-scale auditing

Every individual lookup requires a separate access procedure before the information can be retrieved. This process takes at least approximately 15 seconds per query, before the returned information itself is reviewed or recorded.

Because queries must be performed individually, this makes verification of large numbers of voter records extremely slow. The service may therefore be workable for examining a limited number of individual cases, but it cannot realistically support systematic verification of large populations of voters within the Commission’s statutory timeframe.

The practical consequence is that the Commission has case-by-case verification access, but lacks the searchable, sufficiently detailed, historically complete and scalable access required for a comprehensive audit.

Bottom line

By 28 October 2026, the Commission may be able to partially identify certain anomalies, structural weaknesses, and high-risk patterns in the UVR, but it will remain unable to determine what caused them, how they should be corrected, and who is responsible for doing so. Limited access to underlying source records, administrative decisions, and technical logs will prevent it from reliably distinguishing between administrative error, outdated data, weak controls, individual misconduct, and coordinated manipulation, or from independently verifying the integrity of the UVR’s audit trail.

The first report should therefore be understood as a diagnostic assessment and a set of partial audits, rather than a comprehensive verification of the accuracy, completeness, and integrity of the UVR as a whole.

A credible audit requires searchable access to source-register data, relevant historical and administrative records, and a full IT and information-security audit. Without a rapid change in cooperation, the Commission will remain unable to trace suspicious voter-register findings to their source, establish responsibility, or verify the overall integrity of the UVR, leaving both the core purpose of ODIHR’s recommendation unmet and the longstanding public distrust in the voter register unresolved.