Voter register

The voter register on the basis of which voting will take place on 25 October was closed on 9 October without an audit having been carried out. The Commission, the first body with that task, was not given the conditions to carry one out in eight months. What members of the Commission were nonetheless able to establish through ten official notes is enough to conclude that the register contains tens of thousands of records that need to be checked, and that the system by which it is kept is such that it is hard to say what is an error and what is misuse.

The audit has not been carried out

The Act on the procedure for the audit, which the Commission adopted unanimously on 3 April, provides for two steps without which there is no audit: comparing the register with the source records in both directions, in order to establish who is in the register but should not be and who is not but should be, and checking the extracts on the basis of which voting takes place. Commission for the audit of the voter register, Act on the procedure for the audit of the Unified Voter Register (Belgrade, 3 April 2026). The Commission could carry out neither of them on its own. In March, the Office for IT and eGovernment referred it to the ministries; the Ministry of Public Administration and Local Self-Government (MDULS) took the position that there was neither a legal basis nor the technical means for broader access to the register application or for access to the civil registry, and the Ministry of the Interior (MUP) took the position that only individual look-ups were possible. Commission for the audit of the voter register, Official note on the process of securing access to the data necessary for carrying out the audit of the UVR (Belgrade, 5 October 2026). Unless otherwise stated, the chronology in this section is taken from that note. The Commission was given access to the source records on 5 May, as an individual query by personal identification number through the eZUP system, at four to five minutes per voter; a request to extend the search was rejected. It received an export of the data from the register itself only on 6 July, and the technical means to carry out analyses of voter data from the Unified Voter Register (UVR) independently only in August.

The cross-checking of the register against the source records remained in the hands of the Office: a nationwide test was proposed on 9 September, the day the elections were called, and the platform was presented on 9 October, the day the register was closed. The data for the second direction of the check, on persons who meet the conditions for entry but are not in the register, were not provided to the Commission even for the pilot municipality of Sevojno. Ibid. The Commission requested the second direction of the difference for Sevojno (123 persons, according to a letter from the Office for IT and eGovernment) on 2 September; by the beginning of October the data had not been provided. Members of the Commission appointed on the proposal of the opposition have publicly assessed that the Commission's work was obstructed; MDULS states that the Commission was given access to all relevant data. “BIRN: Voter register still not audited, ODIHR recommendations ignored for 15 years”, Nova.rs, 19 October 2026, accessed 20 October 2026, https://link.crta.rs/17y; Law on Amendments to the Law on the Unified Voter Register, “Official Gazette of the RS”, No. 96/2025, Article 8. CRTA's assessment is that the conditions for an audit were absent not because of technical obstacles, but because the bodies holding the data did not decide to remove those obstacles. The first report to the National Assembly, which the Commission must submit by 28 October, will state that the audit has not been carried out.

Ten notes on the state of the register

Members of the Commission appointed on the proposal of CRTA prepared ten official notes on findings from the data available to the Commission and submitted them to the Commission during October. The Commission is made up of ten members: five appointed on the proposal of the parliamentary groups of the ruling majority, three on the proposal of opposition parliamentary groups and two on the proposal of associations (CeSID and CRTA). Members and deputy members appointed on the proposal of other nominators also took part in preparing some of the notes. The notes are preliminary findings by members, on which the Commission has not yet taken a position, but for the first time they describe the register from the inside, on the basis of data rather than assumptions. Commission for the audit of the voter register, official notes by members: Official note on the process of securing access to the data necessary for carrying out the audit of the UVR; Description of the data in the Unified Voter Register; Multiple entries in the Unified Voter Register: findings of analyses P0-P5; Checking the accuracy of the data on the basis of a sample in 10 local self-government units: overview of the process and preliminary results (all 5 October 2026); Decisions in the Unified Voter Register, December 2011 - October 2026: monthly overview; Changes in the Unified Voter Register in 2026: from 1 January to the closing of the voter register on 9 October 2026 (both 13 October 2026); Consistency of the data from the UVR, the export by polling station and the extracts from the voter register for the parliamentary elections on 25 October 2026 (18 October 2026); Supplement to the official note of 18 October 2026: spaces before and after surnames in the UVR, by local self-government unit (19 October 2026); Places of registration of voters for voting abroad and designated polling stations (19 October 2026); Organised entries in the Unified Voter Register (20 October 2026). On 16 October, the Commission stated that the views of its members expressed in public appearances are not findings of the Commission until it has taken a position on them, and that it would submit the report on the audit to the National Assembly.

Four of these ten notes were described in the previous report: on access to data, on the description of the data in the register, on multiple entries, and on checking the accuracy of the data on samples in ten municipalities. They were followed by notes on decisions in the register from 2011 to 2026, on changes in 2026 up to the closing of the register, on the consistency between the register, the export by polling station and the extracts, on spaces in surnames, on the places of registration for voting abroad and on organised entries.

The first group of findings, which CRTA described in its previous report, concerns the quality of the data: CRTA, First interim report on long-term observation, 9-30 September 2026 (Belgrade: CRTA, October 2026), 45-46; Commission for the audit of the voter register, Description of the data in the Unified Voter Register, snapshot as at 2 October 2026. All figures in the paragraph are from that note; the estimate by the Statistical Office of the Republic of Serbia relates to 31 December 2025 and to residents born in 2007 or earlier (116 voters per 100 adult residents). 248,042 records in six municipalities in Kosovo that since 2012 have been covered neither by elections nor by the public data, without a legal basis known to the Commission; 1,466 voters over the age of 100, six times as many as would be expected to be alive; 488,095 voters without the mandatory data on place of birth; 79 records with the same personal identification number; addresses kept as free text, with no link to the residence records or the Address Register. Above all, outside Kosovo the register contains around 879,000 more voters than there are adult residents. The difference is most likely made up of citizens who live abroad and have not de-registered their permanent residence, and of persons who, after acquiring citizenship, formally registered a permanent residence in Serbia but live in their country of origin, above all in Bosnia and Herzegovina and Montenegro. This surplus is a possible source of manipulation, and its composition can be established only through a full audit, with systematic access to the source and auxiliary records, which the Commission was not given.

The second group of findings concerns multiple entries of the same person under different personal identification numbers, which the law prohibits. Law on the Unified Voter Register, “Official Gazette of the RS”, Nos. 104/2009, 99/2011, 44/2024 and 96/2025, Article 1(4). The figures are from the note Multiple entries in the Unified Voter Register, snapshot as at 18 August 2026; none of the cases has been checked individually in the source records. In 1,025 cases, two records match on all five personal attributes: first name, parent's name, surname, and date and place of birth. When a difference in one attribute is allowed, the analysis singles out 4,662 groups of records that probably relate to the same person and a further 3,301 groups that require additional checks. These 8,988 probable duplicate entries in the voter register are the most conservative estimate: the method captures only records that differ in at most one attribute, so if other criteria are cross-checked there may be considerably more duplicates. Checks on samples confirmed the scale: in Smederevska Palanka, among 399 randomly selected voters, two were found to be entered twice and two to have died before the local elections in March but to have been deleted only after them, and in both of the municipalities checked, at least one finding, most often incomplete citizenship records or an address that does not match the residence records, applies to 22 to 26% of the voters in the sample. Commission for the audit of the voter register, Checking the accuracy of the data on the basis of a sample in 10 local self-government units. Samples of around 400 voters per municipality, stratified by polling station; preliminary results, not verified with the municipal officials who maintain the register.

On the day the register was closed, the Office for IT and eGovernment announced that the voter register matches the source records by “at least 98 per cent”, and pro-government media reported this as proof that the register is in order. The Commission, at whose request the platform was built, had by that day seen neither the platform nor the results. With its announcement, the Office stepped out of its role as a technical implementer and offered the public an incomplete result that answers none of the questions that prompted the demand for an audit.

On 9 October, the Office for IT and eGovernment announced that, at the Commission's request, it had set up a platform that compares the register with the MUP residence records and the registers of births and deaths. According to the announcement, the data as at 1 September match by at least 98%, consistency above 99% is expected after checks by the competent institutions, and in the pilot test in Sevojno the 1.75% of data that did not match was reduced by further analysis to a single voter. Office for Information Technologies and eGovernment, “Platform for the analysis, control and assessment of the quality of data in the Unified Voter Register presented”, 9 October 2026, accessed 20 October 2026, https://link.crta.rs/17x On the same day, pro-government media reported the announcement under headlines stating that the register was “accurate and in order” and that “an end has been put to all speculation”. “An end has been put to all speculation: Confirmed that the voter register is in order and precise”, Večernje novosti, 9 October 2026, https://link.crta.rs/17w; “The voter register is accurate and in order! Digital platform buries all speculation”, Kurir, 9 October 2026, https://link.crta.rs/17z; Informer, 9 October 2026, https://link.crta.rs/17v; all accessed 20 October 2026.

The Commission had requested the platform back in March, as an environment in which it would cross-check the data itself, and it received a solution in which the cross-checking is carried out by the Office, as an imposed intermediary. By 9 October it had no insight into either the platform or the results; the presentation had been scheduled for that very day. Commission for the audit of the voter register, Official note on the process of securing access to data: the request for an analytical environment was sent to the Office for IT and eGovernment on 25 May 2026; on 2 October the presentation of the platform was announced for 9 October. The announcement offered the public a conclusion before it had been seen by the body for which the analysis was done and which alone has the statutory authority to assess the accuracy of the register. On 16 October, the Commission stated that the Office's findings were not the result of a consistent application of the methodology the Commission had established, that the findings of other state bodies are not findings of the Commission, and that it would submit its report to the National Assembly. Commission for the audit of the voter register, “Statement concerning the platform for the analysis, control and assessment of the quality of data in the Unified Voter Register”, 16 October 2026, accessed 20 October 2026, https://link.crta.rs/18a Further cooperation between the Commission and the Office has thereby been called into question.

More important than the procedure is what the 98% figure measures. First, at the level of the whole voter register, “two per cent” is 130,000 voters, enough to cross the electoral threshold in parliamentary elections. Second, the result does not cover voters from Kosovo at all and does not check the citizenship records. Third, the consistency test, as the basic step and only the first in a series of diagnostic steps in an audit, looks only at whether records currently match between the source records and the register: whether there is a corresponding record for the voter in the residence records, in the register of births and in the citizenship records, and no record in the register of deaths. That test does not reveal many other problems that bear on the accuracy of the register: the status of voters whose death has not been recorded, multiple entries, materially incorrect entries, incorrect addresses. A voter formally registered at an address where they do not live; a voter who died abroad or in Kosovo, from where data on deaths do not reach the civil registry in Serbia; a person entered twice under two personal identification numbers - all these cases pass the consistency test. A particular problem is that the second direction of the check - persons who meet the conditions but are not in the register - is not mentioned in the announcement at all. Without checking the second direction, the completeness of the register cannot be assessed. In addition, this test says nothing about what the voter register will look like as early as the following day. According to the findings of members of the Commission, the voter register “absorbs” more than 60,000 changes a month on average.

Extracts from the voter register: a check on a sample

Members of the Commission reviewed a sample of the printed extracts submitted to the Republic Electoral Commission (REC). At 304 polling stations the extracts are consistent with the register: no voters have been added or left out. The review did, however, reveal an error because of which more than 30,000 voters are not in the extracts where the polling board would look for them.

An extract is the list of voters for a single polling station, in which the voter enters a signature when voting, the last link between the database and the polling station. In its previous report, CRTA described how, on 6 October, a proposal that the extracts be submitted to the Commission in electronic form was rejected, with members from the ranks of the majority putting forward a compromise proposal that the Commission review a sample of the extracts at the REC. CRTA, First interim report on long-term observation, 9-30 September 2026 (Belgrade: CRTA, October 2026), 46-47. That is what was done: on 18 October, at the REC, members of the Commission manually compared, on a sample basis, the printed extracts with the electronic export of the closed register by polling station that the Commission had received from MDULS. Commission for the audit of the voter register, Consistency of the data from the UVR, the export by polling station and the extracts from the voter register for the parliamentary elections on 25 October 2026 (Belgrade, 18 October 2026). Ten members and deputy members appointed on the proposal of all three groups of nominators took part in checking the extracts. For 6,448,822 voters the personal identification numbers match in both exports; 1,697 voters from six Kosovo municipalities have not been assigned to any polling station.

First, the personal identification numbers of all voters in the register export were compared with the export by polling station. Then, from all 8,392 polling stations, 304 polling stations were selected at random, proportionally across Belgrade, the administrative districts, abroad and penal institutions, and at each of them three voters at predetermined serial numbers and the total number of voters were checked. At the 912 positions checked, the voters from the export were found, all discrepancies in serial numbers were explained by differences in ordering described below, and the total number of voters matches the decision on closing the register. In the extracts reviewed, therefore, no voters have been added or left out. The check has two limitations: it relates to the extracts at the REC, not to those that will be at the polling stations, and it would detect the substitution of one voter by another only if this shifted the order at the positions checked.

The discrepancies observed have a technical cause with serious consequences. When there is a superfluous space before a surname in the register, the voter appears at the very end of the alphabetically sorted extract. When the space comes after the surname, the voter is moved to the end of the sequence of voters with the same surname. The data are correct, but the voter is not where the board would look for them. A space before the surname affects 478 voters, and a space after the surname 30,553, at more than half of the polling stations in the country. Commission for the audit of the voter register, Supplement to the official note of 18 October 2026: spaces before and after surnames in the UVR, by local self-government unit (Belgrade, 19 October 2026); based on the MDULS export by polling station at the closing of the register. In Novi Pazar 7,462 voters are affected (7.9%), in Čukarica 7,420, in Novi Kneževac every tenth voter, and at the only polling station for the relocated municipality of Istok more than half. Such a distribution across whole municipalities indicates the way data are entered in those administrations rather than random errors.

The consequence is practical: a voter without the voting notice, which carries the serial number, may be turned away because the board cannot find them in the expected place. According to CRTA's findings, at the 2017 presidential election, voters at 16% of polling stations reported that they were in the electronic voter register but were missing from the extract at the polling station. According to the report, this irregularity affected 0.2% of voters. Members of the Commission proposed that, before the elections, the REC notify the local election commissions to stress to the polling boards that, when they cannot find a voter, they should check the end of the sequence for the same surname and the last page of the extract, and that MDULS should in future hand over data for printing cleaned of superfluous spaces and with a verified correct ordering of voters by surname before numbering. By the time this report was finalised, the REC had not sent such a notice.

Two figures from the Ministry of the Interior on the number of valid identity cards, which differed by 236,000 over four months, reached the public two weeks before the elections and caused alarm of a kind that few figures on the voter register have ever caused. MUP then stated that the first figure had been wrong, but the admission of the error did not dispel suspicion: after eight months in which the institutions hindered the audit more than they enabled it, the public is more inclined to believe that the state is hiding voters than that it made an error in counting documents.

The number of valid identity cards held by adult citizens is the closest available measure of the set of persons who should be in the register, so a member of the Commission twice requested this figure from MUP. According to the reply from June, on 18 May there were 5,988,749 valid identity cards held by adult citizens; according to the reply of 9 October, on 28 September there were 6,225,029. Ministry of the Interior, act 01 No. 021-1113/26-4 of 4 June 2026 (in response to the Commission's request 30 No. 208-63/26 of 18 May 2026) and act 01 No. 021-282/26-154 of 9 October 2026 (in response to request 30 No. 208-195/26 of 28 September 2026), according to MUP's letter to the Commission 01 No. 021-282/26-164 of 15 October 2026. Minors with a valid identity card numbered 111,229 and 119,048 respectively. The difference of more than 236,000 documents in four months had no counterpart in the register, in which around 10,000 new voters were entered over the same period, and MUP replied to the first query that there was nothing contentious about the increase. When members of the Commission raised the difference at the session on 13 October and in public, headlines followed about 235,000 new identity cards before the elections, along with a demand that MUP respond before the elections. “Number of valid identity cards up by more than 200,000: ‘Where did these people come from and who are they?’”, N1, 15 October 2026, https://link.crta.rs/18b; “Serbia's Ministry of the Interior should explain how the number of identity cards rose by 235,000 in four months”, Nova.rs, 14 October 2026, https://link.crta.rs/18c; both accessed 20 October 2026. The growth of the register by around 10,000 is cited by MUP in its letter of 15 October.

On 15 October, MUP informed the Commission that on 18 May there had in fact been 6,211,724 valid identity cards held by adult citizens, and that the query run by the person processing the request had by mistake not included 222,975 permanent identity cards, which have no expiry date recorded. Ministry of the Interior, Office of the Minister, letter to the Commission 01 No. 021-282/26-164 of 15 October 2026. Permanent identity cards are issued to citizens over the age of 65. Owing to limited access, members of the Commission were unable to verify this reply from MUP, but some members noted that the corrected figure, an increase of around 13,000 documents, is consistent with regular issuance and with the changes they see in the voter register. The public reaction to the correction pointed to the dramatic state of trust in the competent institutions: the question was not whether MUP had made a mistake, but which of the two figures was false. “Rašuo: Police claim a technical error led them to provide a wrong figure on the number of valid identity cards”, Insajder, 18 October 2026, accessed 20 October 2026, https://link.crta.rs/18d. For five months the Commission worked with a figure that a state authority had issued as accurate, confirmed as uncontested when asked again, and corrected only once it became public. The same authority and the same information system are the source of the residence data that are transferred into the register and that, according to the sample findings, most often diverge from it. Commission for the audit of the voter register, Checking the accuracy of the data on the basis of a sample in 10 local self-government units: in Aranđelovac and Smederevska Palanka, most of the discrepancies between the register and the residence records concern the address. When a query can omit 223,000 records without anyone noticing, an assurance that the data are accurate is not enough for anyone whose duty it is to check them, particularly since, by the time this report was finalised, MUP had provided neither the Commission nor the public with more detailed data that would further allay the suspicion.

The corrected figure also opens up the comparison the Commission originally sought. The register was closed with 6,450,523 voters, while adult citizens with a valid identity card numbered 6,225,029: around 225,000 fewer. The register may also include citizens whose identity card has expired, but that does not explain a difference of this order of magnitude. Since an identity card ceases to be valid when a death is entered in the residence records, Law on Identity Cards, Article 20(1)(2) and 20(3), according to MUP's letter of 15 October 2026. a voter without a valid document is either a citizen who has not renewed it for years, most often because they live abroad, or a person whose death has not been recorded; among the oldest voters, members of the Commission found more than 180 persons without a single valid identity card. “Commission on the voter register: We do not know how many have died”, Bujanovačke, 8 October 2026, citing a statement by a member of the Commission to N1, accessed 20 October 2026, https://link.crta.rs/18e. The difference of 225,000 is not the number of non-existent voters, but it is the order of magnitude of what an audit would have to explain, and could easily explain, if the Commission had the access to all records relevant to the voter register that the law prescribes.

A note by members of the Commission for the audit of the voter register on organised entries in the voter register describes two ways in which the register was shaped ahead of earlier elections: moving voters already in the register to a municipality holding local elections, and entering new voters, predominantly born abroad, in groups, at a small number of addresses and ahead of elections. The analysis confirmed that both patterns existed and that they were most pronounced ahead of the elections in December 2023. The findings strongly confirm CRTA's assessment in its final report on the 2023 Belgrade elections: that the election result did not reflect the electoral will of citizens and that the manipulations identified affected the final distribution of seats, and thus possibly also who holds power in Belgrade. The 2024 law restricted the first pattern, but not the second. The voter register for 25 October 2026 still contains voters who were entered en masse at numerous addresses in order to influence elections.

Members of the Commission based the analysis on the decisions on changes to the voter register, which show every change in a voter's status since 2011, with its date and legal ground. Moving between municipalities and being entered in the register for the first time are regular occurrences; they become a sign of organised activity when they cluster: several persons at the same address within a short period, in the days before elections, persons who de-register or are deleted after the elections, or who, by all appearances, do not live at the registered address. Such patterns were looked for around each of the 598 local elections held from 2012 to 2023, and a selected sample of addresses was ranked by likelihood and checked in the source records. Commission for the audit of the voter register, Organised entries in the Unified Voter Register (Belgrade, 20 October 2026), data as of 9 October 2026. Basis: all decisions for the 246,566 voters who moved between municipalities at least twice and for the 387,271 persons entered for the first time since 2012. The note follows on from the note Decisions in the Unified Voter Register, December 2011 - October 2026: monthly overview (13 October 2026). CRTA's assessment of the Belgrade elections: CRTA, Elections 2023: final report (Belgrade: CRTA, February 2024). Two caveats apply. The indicators capture only entries that cluster above set thresholds, so everything that follows is a lower bound: there may be considerably more organised entries that blend in with regular ones. And “moving” here means a voter being transferred in the register on the basis of a registered permanent residence; whether the voter actually moved can be established only through a large-scale check of the residence records.

The first pattern, registering permanent residence before local elections and de-registering after them, changes not the number of voters but the municipality in which a voter votes. Of the 598 local elections, twenty stand out, and four fifths of the excess falls on five municipalities that held early elections on 17 December 2023. Ahead of the City Assembly elections in Belgrade, 1.7 times as many voters from other municipalities registered as in comparable periods, while national elections, which this kind of moving cannot affect, show no excess. The transitional rule of the 2024 law returned almost 90% of such voters before the local elections in June 2024, and after those elections almost all of them re-registered at the same addresses. Law on Amendments to the Law on the Unified Voter Register, “Official Gazette of the RS”, No. 44/2024, Articles 2 and 6(2). According to the note, 52,017 of 59,207 voters were returned; 3,947 were not, of whom 2,040 because of a gap in the rule and 1,684 with no explanation in the data. Around the local elections of 2025 and 2026 the pattern no longer stands out.

The second pattern, first entries bearing the hallmarks of organised entry, adds voters, has the same effect at national and local elections, and has not been addressed by the law. The findings single out around 2,100 addresses with around 19,000 persons entered, more than 7,000 of them in Belgrade, and around 1,100 more villages and streets without a house number. Of the voters entered in groups or immediately before elections, 89% were born abroad, three quarters of them in Bosnia and Herzegovina; almost half subsequently requested to vote abroad, and one in ten was deleted ex officio at least once, often in groups, several months after the elections. The pattern begins in 2021, peaks in 2022 and 2023, and continues in 2026, with less clustering at a single address. In the register for 25 October, 9,507 of these persons are entered at the addresses singled out; at 25 polling stations, 21 of them in Podrinje, they make up 5% or more of voters.

Confirmation came from a flat-by-flat check, or house-by-house in the case of family houses. For twenty addresses in five Belgrade municipalities and in Mali Zvornik, members of the Commission compared residence registrations by flat and cadastral data on floor area and ownership with the decisions in the register. The finding is the same everywhere: dozens of voters registered in flats of 23 to 130 square metres, with different surnames and places of birth that do not correspond to a single household, entered on the same day by consecutive decisions and in the weeks before elections; in Belgrade, most of them with a request to vote abroad, and in Mali Zvornik with group deletions ex officio after the elections, carried out at some addresses but not at others with the same pattern. Around 500 voters are registered at these sample addresses, and a large share of them can also vote on 25 October. The flat-by-flat check was based on data from the residence records and the real estate cadastre available to members of the Commission; personal data and data on owners are not part of the note. The note does not claim that the persons entered voted, or how; it shows that the order of magnitude of the entries singled out in Belgrade is comparable to the number of votes that would have changed the distribution of seats in the 2023 City Assembly elections.

Three conclusions follow for the assessment of the 2026 elections. First, the 2023 patterns have been confirmed on the basis of records kept by the competent authorities - manipulation of permanent residence in order to influence elections did take place. Second, although the six-month deadline significantly reduced the effect of moving within the country ahead of local elections, all first entries remained outside this rule: they are possible right up to the day the register is closed, and without proof of actual residence. Third, the same signals are present in 2026 - most of the voters entered for the first time whose permanent residence was registered in earlier electoral cycles in order to influence elections are in the voter register and will be able to vote in the elections on 25 October.

The register was closed with 6,450,523 voters, 49,642 fewer than for the 2023 elections. Even so, in the election year it grew, instead of shrinking as it had the previous year, and the whole difference, at most 14,100 voters, arose in the last 90 days before the closing. In that period the register was filled by new voters who had not previously been in any municipality, by voters who vote outside their place of permanent residence and, in Kosovo, by voters moved by decisions of the administration. Taking into account new entries and the situation inherited from 2022 and 2023, there are 13,900 voters at singled-out addresses showing patterns of organised entry.

The note by members of the Commission on changes in the register from 1 January up to the closing is based on 755,796 individual decisions issued in 2026 and on monthly statistics since 2011. Commission for the audit of the voter register, Changes in the Unified Voter Register in 2026: from 1 January to the closing of the voter register on 9 October 2026 (Belgrade, 13 October 2026), data as of 10 October 2026. All figures in this section are from that note, excluding the six municipalities in Kosovo in which elections are not held. The last 90 days: 12 July - 9 October. CRTA's previous report measured growth from February (+6,994 by 28 September); the note measures it from 23 December 2025. From 23 December 2025 up to the closing, the register grew by 1,202 voters; had the decline of 2025 continued, it would have been around 12,900 smaller. The difference compared with a year without elections is at most 14,100 voters, and at least 8,700 if the same months are compared, and all of it arose in the last 90 days, when entries exceeded deletions by 11,943. Growth before the closing is not new, but it is greater than in 2020, 2022 and 2023, and it does not stem from a backlog in deleting deceased persons.

First entries in the broad sense, persons who since 2012 had never been in the register in any municipality and who were entered on the basis of a registration of permanent residence or admission to citizenship, are the main source of growth. There were 21,315 of them in 2026, half in the last 90 days. In the three-month window before the closing there were 2.1 times as many per day as in the preceding twelve months, against 1.4 to 1.7 times at the elections from 2014 to 2023, with the exception of 2016. The share of first entries made at the person's own request jumped from 2.6% to 40.9%, partly because, after the elections are called, citizens have the right to request entry (Law on the Unified Voter Register, Article 14). In Kosovo, 91% of this year's first entries were made in the last 90 days; outside Kosovo, the largest excess is in Bujanovac, Prijepolje, Mali Zvornik and Vranje. One in seven of those entered for the first time in the last 90 days subsequently requested to vote abroad. The six-month deadline does not apply to first entries, so they remain possible right up to the closing.

In the last 90 days, compared with the 2022-2023 period, patterns of organised entry were most pronounced in Bujanovac, Gnjilane, Priština, Vitina and Obilić, but also in Vranje, where 218 voters were entered for the first time at 33 singled-out addresses, more than 180 of whom were born in North Macedonia. Of these, 138 registered to vote abroad immediately after being entered. In terms of scale, however, the main risk for the elections of 25 October is not a new wave of entries but the situation inherited from the 2021-2023 period. There are 13,900 voters at singled-out addresses showing patterns of organised entry, 4,413 of whom (32%) have registered to vote abroad.

In 2026, 7,380 temporary residences of internally displaced persons from Kosovo were deleted, the most since 2018; 3,381 in September alone, three times more than in any earlier month. A voter whose temporary residence is deleted remains in the register but is returned to the extract for their municipality of permanent residence in Kosovo, and so changes their place of voting, and if they do not receive the decision, they do not know this. Not one decision was issued at a voter's request, five municipalities issued 43% of the deletions made since August, and in Vrnjačka Banja the decisions were not delivered to persons with permanent residence in Kosovo, so they did not even have the possibility of appeal. Ministry of Public Administration and Local Self-Government (MDULS), Manual for the implementation of the Law on the Unified Voter Register (Belgrade, 2019), Chapter V, 4.2.3; replies of municipalities to letters from members of the Commission, according to the note. The five municipalities: Novi Pazar, Subotica, Kragujevac, Vrnjačka Banja and Prokuplje; the series of deletions came after the MDULS webinar for municipalities on 2 September and the measures taken by the administrative inspectorate.

Voting outside the place of permanent residence reached an unprecedented scale: 102,147 voters were assigned to vote abroad, 2.6 times more than in 2023, and 26,843 were entered to vote at a temporary residence in the country, twice as many as in 2023, more than a quarter of them in the 23 municipalities in Kosovo that take part in the elections. Number of voters abroad: MDULS table of assignment to polling stations, 17 October 2026 (see section 7); there were 105,513 entries for voting abroad, but no polling station was designated for around 3,000 of them. Law on the Unified Voter Register, Article 7(1).

In the month from the calling of the elections to the closing, the municipal officers who maintain the voter register issued a third of all this year's decisions, four times more per day than in the first half of the year, with the fastest growth in changes that citizens do not initiate: corrections of technical errors, certificates of electoral right, deletions of the temporary residence of displaced persons, and deletions on account of loss of legal capacity. In the thirty days from the calling of the elections to the day before the closing, 240,062 decisions were issued, 31.8% of the annual number (8,002 per day against 2,027). Corrections of technical errors rose 35-fold (Preševo: 76% of all those in the country, for 15% of its voters), certificates of electoral right 7.8-fold (Ražanj: 639, of which 513 on the day of the closing, even though the Law on the Election of MPs, Article 71, does not require them even for candidate registration), deletions of the temporary residence of displaced persons 5.5-fold, deletions on account of loss of legal capacity 4.4-fold (1,134 in the year, twice as many as in the whole of 2025), and changes to personal data 3.3-fold; deletions on account of death and change of address did not rise. The note also singles out smaller numbers that call for individual checking: 13 persons for whom a new decision was issued after they had been deleted on account of death, 1,393 entries of persons living abroad on the basis of their last permanent residence, almost all of them in the last 90 days, and around 600 persons recorded as first entries who had already been in the register. The competent authorities have offered no explanation for any of these findings ahead of the elections.

In its previous report, CRTA left open the question of where the around 23,000 voters not covered by the decision on polling stations abroad would vote. The answer came through the extracts: 102,147 voters were assigned to 93 polling stations, but in such a way that exercising the right to vote has been made seriously more difficult for three groups: more than 13,000 voters were assigned more than 50 kilometres from the place they gave in their registration, almost 300 voters from cities with an open polling station were sent, without explanation, to another city or country, and almost 11,000 will have to vote partly at night. Particularly disadvantaged are 4,800 voters from as many as 19 cities in which more than 100 voters registered, thereby meeting the statutory requirement, yet no polling station was opened. To exercise their right, between them they will have to travel a distance equivalent to 15 trips to the Moon and back.

Members of the Commission analysed the table of the Ministry of Public Administration and Local Self-Government (MDULS) listing the 102,147 voters assigned to polling stations abroad, as of 17 October. Commission for the audit of the voter register, Places of registration of voters for voting abroad and designated polling stations (Belgrade, 19 October 2026). The analysis was carried out on data stripped of voters' personal data. The figure of 102,147 refers to voters with a valid decision who were assigned to a polling station; for around 3,000 voters with a valid decision no polling station was designated, and of the 106,306 requests around 700 were not accepted. CRTA's previous report used the number of requests. The place where a voter wishes to vote is entered in the request as free text, so for the analysis it was reduced to a single standard name: 512 places in 55 countries, against 93 polling stations in 37 countries.

The first group consists of voters assigned far from the place of registration. Of these, 13,485 (13.2%) were sent to a polling station more than 50 kilometres away in a straight line, 5,878 more than 200, and 2,213 more than a thousand kilometres away. Half of the distant voters came from ten cities, among them Vicenza, Bratislava, Vancouver, Miami, Barcelona, Geneva and Herceg Novi; in most of them there was a polling station in 2023, which was discontinued by the decision of 5 October. Republic Electoral Commission, Decision on the designation of polling stations abroad, 02 no. 013-319/26-3 (5 October 2026). Distances in the note were measured in a straight line between the centres of the places; the distance by road is as a rule greater. Voters from Vancouver were assigned to Toronto, 3,355 kilometres away, from Miami to Washington, from Barcelona to Madrid. From sixteen countries without a polling station, 1,208 voters were sent to another country, and for 473 voters there was a closer polling station in the same country: 176 voters from Houston were sent to Los Angeles, even though Chicago is 700 kilometres closer. Particularly affected are voters from 19 places in which the statutory minimum of at least 100 registered voters was met - no polling stations were opened in them, so 4,824 voters were assigned to 13 other polling stations. To exercise their right to vote, they will have to travel a total of 11,693,558 kilometres in both directions, which corresponds to more than 15 trips to the Moon and back.

The second group consists of 291 voters from 48 cities that do have a polling station, who were nevertheless assigned to another city, 44 of them to another country, from Dublin to London and from Luxembourg to Brussels. Of 266 voters from Dublin, 246 vote in Dublin and 20 in London, with the same details in their requests. Within the country this is impossible, because every address belongs to the area of exactly one polling station; abroad, no such rule was applied.

The third group was created by the decision of the Republic Electoral Commission (REC) that voting at all polling stations abroad should run from 07:00 to 20:00 Serbian time, which CRTA has already assessed as exceeding the REC's powers. Republic Electoral Commission, Decision on the determination of the times at which voting begins and ends at polling stations abroad (5 October 2026); CRTA, Voting abroad: the REC has abandoned a decade-long practice and imposed night-time voting on the diaspora, media statement (Belgrade, 5 October 2026). At 13 polling stations, with 10,851 voters, voting takes place partly between 22:00 and 06:00 local time: in Los Angeles voting begins at 23:00 on Saturday, on the east coast of the United States and in Canada at 02:00, and in Australia and New Zealand it ends on Monday morning. Voters who have already been assigned too far away are affected a second time by this decision.

The Law on the Election of MPs makes no distinction between polling stations in the country and abroad: a polling station must be accessible to voters, an area is designated for each one, and a polling station may also be designated for fewer than a hundred voters if distance would make voting at another one considerably more difficult for them. Law on the Election of MPs, Article 57(1), (3) and (4); Article 91(4). The consequences of an error in assignment are more serious than within the country: a voter entered to vote abroad is removed from the extract for their place of permanent residence and can vote only where they have been assigned, and this information cannot be changed until the elections are over. Law on the Unified Voter Register, Article 19(3); Instruction on the implementation of the Law on the Unified Voter Register, point 8. Members of the Commission proposed that the Commission take a position on these decisions; as of the cut-off date for this report, the Commission had not done so.

Ahead of the 2023 elections, CRTA recorded the case of voters from Novi Sad who, without their knowledge, had been entered in the register as voters voting abroad, in Banja Luka. Ahead of the elections on 25 October, the same thing happened to two voters from the same group, even though they had filed criminal complaints over the previous abuse. Impunity for one abuse has produced the next.

Ahead of the elections held on 17 December 2023, a group of voters from Novi Sad discovered that the voter register recorded them as voting abroad, in Banja Luka, even though they had never submitted such a request. With CRTA's legal assistance, four of them filed criminal complaints for abuse of office in January 2024 against the responsible officials, including the Consulate of the Republic of Serbia in Banja Luka. CRTA, records of legal assistance to citizens; criminal complaints filed in January 2024. Criminal Code, “Official Gazette of the RS”, Nos. 85/2005 to 94/2024, Article 359 (abuse of office). To date, the complaints have had no outcome.

Ahead of the elections scheduled for 25 October 2026, the same pattern was repeated for two voters from the same group: they were again entered as voters voting in Banja Luka, without a request and without their knowledge. They learned of the change only when they checked the register themselves, which is the only way a voter can find out that they have been moved, because the decision entering the information on voting abroad is sent to the address abroad given by the person who submitted the request. They did not manage to have the entry reversed before the register was closed.

Although we do not know the actual scale of this phenomenon, it is certainly significant for what it shows. A request to vote abroad can still be submitted in someone else's name, without effective verification of the applicant's identity or the authenticity of the signature, through the diplomatic and consular mission that receives the request and the municipal administration that issues a decision on it. Law on the Unified Voter Register, Article 16; Instruction on the implementation of the Law on the Unified Voter Register, point 8: the request is submitted through the diplomatic and consular mission, and the decision is issued by the municipal administration of the voter's last permanent residence. A voter who has once been the victim of such abuse and has reported it is not protected against its repetition, because the complaint has led neither to anyone being held accountable nor to a change in the procedure, and the consequence is immediate: a voter entered to vote abroad is removed from the extract for their place of permanent residence and cannot vote in Novi Sad. Law on the Unified Voter Register, Article 19(3). The case raises two questions that go beyond a single group of voters: the effectiveness of criminal-law protection of the right to vote, and the accountability of the bodies that act on such requests without verification.

From 6 October to the cut-off date for this report, CRTA received a further 498 citizen reports concerning the voter register. After the register was closed, their content also changed: citizens are now photographing the voting notifications that arrive at their addresses for persons who do not live there or who have died, and dilapidated buildings in which no one lives but to which notifications still arrive. The number and nature of the reports are in themselves a measure of trust in the register.

In its previous report, CRTA described 341 messages from citizens received by 6 October, predominantly about incomplete address data and about the difference between the number of voters at an address and the number of people who live there. CRTA, First interim report on long-term observation, 9-30 September 2026 (Belgrade: CRTA, October 2026), 47-48: 341 messages, making up 190 cases for checking, from March to 6 October 2026. After the register is closed, municipal administrations deliver to every voter a notification of the day and time of the elections, with the number of the polling station and the voter's serial number in the extract. Instruction on the implementation of the Law on the Unified Voter Register, point 11: the notification is printed and delivered by the municipal or city administration in which the voter is entered in the extract from the voter register. This notification marks the first time the register physically arrives at an address, and citizens have recognised it as evidence: among the 498 reports received since 6 October, the most frequent are photographs of notifications for persons who do not live at the address and whom the residents do not know, followed by notifications for deceased members of the household or neighbours, as well as photographs of dilapidated residential and non-residential buildings to whose addresses notifications nevertheless arrive. The reports have not been checked individually and do not show how many such entries there are in the register; what they show is that when the register arrives on their doorstep, citizens recognise in it people who are not there.

The influx was so great that CRTA, “owing to the large number of reports reaching us over the past week”, was unable to reply to every message individually.

More than 800 reports in seven months, almost five hundred of them in two weeks, are an indicator of sorts of low trust in the voter register. Citizens would not photograph letterboxes and send the photographs to an observation organisation if they believed that the competent body would correct the error itself, or that there was any body at all to which they could turn with a prospect of a reply. In the absence of an audit, they have become the only mechanism for checking the register at the level of the address, with no powers, no access to the records and no answer from the institutions to the question of who the people entered in their flats are.