ANALYSIS | 25 YEARS SINCE THE OHRID FRAMEWORK AGREEMENT
Twenty-five years since the Ohrid Framework Agreement: the framework is in place, but the full picture is still missing
On the twenty-fifth anniversary of the Ohrid Framework Agreement in North Macedonia, the gap is no longer between the text of the Agreement and the legal order. It lies between the legal order and administrative and budgetary practice. The same pattern recurs across each of the Agreement’s five pillars: a legal instrument is created, but the funding, sanction or verification mechanism is not. This is stated in an analysis by expert Lura Pollozhani, prepared for the Metamorphosis Foundation.
How was the Agreement implemented over the past 25 years?
Laws were created and amended during the first decade; later, they began to be challenged.
The Agreement that ended the conflict
On 13 August 2001, after a six-month conflict in North Macedonia that cost around 250 lives and displaced between 140,000 and 170,000 people, the Ohrid Framework Agreement was signed.
From a political agreement to a constitutional norm
At the constitutional level, the Agreement accomplished what it was signed to do. Fifteen amendments, promulgated on 16 November 2001, translated it into a binding legal norm, while Amendment XXXIV of 2019 incorporated the Agreement itself into the Preamble of the Constitution. Twenty-five years later, the question is no longer whether a legal basis exists. The question is why that legal basis is not producing the expected result.
From creation to contestation
Laws were created and amended during the first decade; later, they began to be challenged.
Decentralisation: still at the 2011 level
Rapid growth until 2011, followed by stagnation
Rapid growth until 2011
Fiscal decentralisation officially began in 2005 and progressed rapidly. The total revenue of local self-government units rose from 1.9 per cent of gross domestic product in 2005 to 5.8 per cent by 2011. Growth then stopped. Estimates for 2023 put it at around 5 to 5.5 per cent—essentially the same level as more than a decade earlier—while the EU-27 average is 11.2 per cent.
More responsibilities, an even wider gap
The real gap is even wider than this simple comparison suggests, because municipalities in North Macedonia are responsible not only for primary education, as most local authorities in the European Union are, but also for secondary education.
Expired deadlines, unmet goals
This is not an external assessment. It follows from measurements against the goals set by the Government itself. The 2021–2026 Programme for Sustainable Local Development and Decentralisation identifies local public revenue as a share of gross domestic product as a central indicator, with a baseline of 5.40 per cent for 2019, a target of 5.80 per cent for 2024 and 7.00 per cent for 2026. The 2024 target was not met. The deadline for the 2026 target expires this year. The cause of the stagnation is structural. The Constitution guarantees local self-government and expands municipal competences, but sets no funding threshold. The fiscal weight of decentralisation therefore remains a matter decided when the annual state budget is adopted. The Government programme clearly shows that grants accounted for close to 60 per cent of local revenue between 2013 and 2019, meaning that growth did not come from strengthening municipalities’ fiscal capacity, but from central-government transfers. On the expenditure side, 54 per cent of municipal budgets goes to salaries—the highest share in the Western Balkans—leaving minimal room for genuine local decision-making.
Equitable representation: slow progress and regression
Representation increased slowly; the target percentage was then lowered and has still not been reached.
Real growth, but below the required level
In 2001, only 7.5 per cent of public-administration employees were Albanian, compared with 84.5 per cent who were Macedonian. This was one of the direct causes of the conflict. The subsequent progress in representation is real: 16.9 per cent in 2010, 18.4 per cent in 2014 and 21.70 per cent according to the 2025 Register of Public Sector Employees.
Representation was not achieved as envisaged
This pace, however, reveals another problem. Between 2024 and 2025, representation rose from 21.39 to 21.70 per cent—a very small increase. At this rate, the envisaged level of representation was not achieved within the stated timeframe.
The target percentage was lowered by a unilateral decision
Meanwhile, the target percentage for representation also changed. In 2024, the Minister of Public Administration altered the employment targets: for Albanians, from 29.52 to 24.30 per cent, and for Macedonians, from 54.21 to 58.44 per cent. The decision was taken unilaterally at ministerial level, outside any procedure requiring the consent of communities not in the majority.
One census, two bases for calculation
To allow readers to judge for themselves, the origin of the two different figures must be explained. According to the final results of the 2021 Census published by the State Statistical Office, the total enumerated population is 2,097,319, of whom 54.21 per cent are Macedonians and 29.52 per cent are Albanians. The resident population is 1,836,713, of whom 58.44 per cent are Macedonians and 24.30 per cent are Albanians. Both figures therefore come from the same census. The only difference is whether the calculation includes those who self-enumerated abroad, among whom Albanians account for 66.36 per cent of the non-resident population. The choice of census base is not a technical matter. It lowers the representation target by more than five percentage points. Even after that reduction, the target has not been met.
Equitable representation of the communities has not been achieved
The public focus on the Albanian community obscures the other side of the problem. No community meets its own representation target: Turks stand at 2.06 per cent against a target of 3.86; Roma at 1.16 against 2.53; Serbs at 0.96 against 1.30; and Vlachs at 0.41 against 0.47 per cent. Moreover, between 2024 and 2025, all communities below the 20 per cent threshold recorded a slight decline.
The average conceals the full structure
Looking only at average representation does not reveal its full structure. Across the ministries taken together, Albanian representation stands at 26.03 per cent, above the established target. But the figures fall sharply in institutions of central government and oversight: 11.33 per cent in the Government Service, 12.73 per cent in the Office of the President, 14.45 per cent in the Government, 16.12 per cent in the Public Prosecutor’s Office, 17.65 per cent in the courts and 18.03 per cent in the Constitutional Court.
Lower representation where decisions are made
The same disparity appears within the ministries. Representation is lowest in Sport, at 12.20 per cent; Defence, at 13.79 per cent; and Digital Transformation, at 14.47 per cent. It is highest in the Ministry of Inter-Community Relations, at 52.47 per cent; Health, at 43.85 per cent; and Economy, at 39.86 per cent.
Representation depends on the institution, not on an overall state policy
Two observations follow naturally. Under-representation remains pronounced in the defence sector, a fact of symbolic significance given the origins of the 2001 conflict. The Ministry of Inter-Community Relations, with 52.47 per cent Albanian staff compared with 34.57 per cent Macedonian staff, reproduces a familiar criticism: institutions created to protect equitable representation risk remaining isolated instead of influencing policy as a whole.
The new law: an instrument without sanctions
The law replaces the logic of quotas with planning, reporting and coordination, but institutions remain unsanctioned.
The Balancer was abolished; there is still no new law
The mechanism that enabled the increase in representation, known as the Balancer, was abolished by a Constitutional Court decision on 9 October 2024. Almost two years later, in June 2026, the law intended to replace it was still only in parliamentary procedure.
The absence of quotas is not merely a technical gap
The Draft Law on Fair and Adequate Representation replaces the logic of quotas with planning, reporting and coordination. Qualifications and competence are the primary criteria, while membership of an under-represented community may be considered only when candidates have equal qualifications. The absence of quotas is not a technical gap, but a deliberate choice justified by the Venice Commission’s opinion that a merit-based system should remain the foundation of public-sector employment.
Institutions are not sanctioned, but citizens are
The problem lies elsewhere. The law contains no sanctions for institutions that fail to comply with it. If a body does not submit data by the deadline, the only consequence is that the Coordinating Body informs the Government through a special report. The only specific sanction is directed at the citizen: if an applicant abuses the declaration of community affiliation, the applicant is rejected from that recruitment procedure and may not apply for the following three months. This asymmetry—in which the punitive measure affects the applicant while the institution faces no consequences—is the clearest indicator of the law’s actual enforcement weight.
A law without a budget and with an unclear basis
The financial explanation is equally significant. The Government states that the law requires no additional funds for implementation and that training will be financed from existing budgets. A law that creates a new coordinating body, a planning methodology, annual reporting and an educational programme without any dedicated funding continues the existing practice. Nor does the law resolve the issue that arose when the ministry changed the target employment percentages, because it does not specify whether the calculation should be based on the resident population or on the population including the diaspora. The personal name recorded on an identity card is envisaged as an objective verification criterion—an unsuitable solution for people from mixed marriages and one that conflicts with the guarantee of free self-identification.
The law is still in parliamentary procedure
At the time of publication, the law had still not been adopted. According to media reports, it passed its first reading in June 2026, underwent an amendment debate in which 31 amendments were submitted, and was placed on the agenda for a second reading in July. On 4 June 2026, the Committee on Inter-Community Relations decided, by 11 votes to two, that the law should be adopted under the so-called ‘small Badinter’ rule—that is, by a majority of the MPs present who do not belong to the majority community. MPs from the Albanian opposition demanded the ‘large Badinter’ rule—a majority of the total number of MPs who do not belong to the majority community. This parliamentary procedural detail is highly significant. It is the most recent illustration of what the analysis below finds about the Badinter mechanism itself.
Language: a right that stops at the service counter
Language rights are measured at the municipal service counter, in the courtroom and in the document a citizen needs to obtain.
The right that affects citizens every day
Language is the only one of the pillars that affects citizens every day. It is not measured in budget percentages, but at the municipal service counter, in the courtroom and in the document that needs to be obtained. Two issues define the current unresolved situation. The first is the case concerning the constitutionality of the Law on the Use of Languages, which has been pending without a decision before the Constitutional Court since 2019, with a second postponement in December 2024. The second emerged in spring 2026, when law students from three universities organised protests demanding that the bar examination also be available in Albanian. The Government declared the matter closed, treating the examination as a professional assessment rather than an administrative procedure governed by the Law on the Use of Languages.
Studying in Albanian, but no professional certification in the language
The practical result is a broken chain: a person may complete their entire course of study in Albanian, but cannot obtain professional certification in the same language.
The right changes with one’s address
Restrictions on language use are also evident in communication with central institutions, where the law permits this only for residents of municipalities in which Albanian is an official language—a narrower interpretation than the constitutional guarantee itself. At the local level, in 2014 there were 31 units with more than one official language, eight of which had voluntarily added the language of a community below the 20 per cent threshold. Nevertheless, no municipality had voluntarily introduced Albanian where the threshold was not met. The European Charter for Regional or Minority Languages, signed in 1996, remains unratified.
Education: a decade without change
Ten years with the same language structure in the school system
A decade with almost the same division
In primary and secondary education, the language structure has remained practically unchanged for a decade. At the end of the 2022/2023 school year, of 182,245 primary-school pupils, 63.6 per cent studied in Macedonian, 32.4 per cent in Albanian and 3.8 per cent in Turkish. In 2015, the corresponding figures were around 64.5, 32.2 and 3 per cent. In secondary education, of 66,678 students, 67.3 per cent studied in Macedonian and 28.6 per cent in Albanian, compared with around 67 and 29 per cent a decade earlier.
A language division that reproduces itself
The linguistic division of the school system is not a temporary standstill that is gradually being overcome. It is a stable structure reproduced from one generation to the next.
Inter-ethnic contact that is largely theoretical
The consequences can also be measured through real-life contact. According to data cited in the 2015 Review, fewer than 10 to 13 per cent of Macedonian and Albanian pupils in primary education, and around 37 per cent of those in secondary education, have at least a theoretical possibility of being in the same building at the same time. Even when schools are attended by pupils from different communities, they often operate in separate shifts according to language.
A plan with a budget, but without measuring outcomes
Pre-school education remains the weakest link. According to data presented in June 2026, more than 79,000 children up to the age of six remain outside pre-school education, while almost 5,000 children waited for a kindergarten place during the previous year. There has also been regression—that is, movement in the opposite direction. The European Commission against Racism and Intolerance criticised the country, noting that integration among communities had been left without a concrete action plan or funding after it was incorporated into the general education strategy. The 2026–2032 Education Strategy is accompanied by an Action Plan with deadlines, responsible institutions, indicators and budget lines, but two limitations remain: interculturalism is treated as a cross-cutting theme rather than a separate priority area, while success is measured only by the percentage of schools carrying out activities, not by any real increase in inter-ethnic contact.
Badinter: a mechanism manipulated by the majority itself
Three formulas with different thresholds and scopes do not provide equal protection.
The veto of non-majority communities
Double-majority voting was the Agreement’s most original solution. It requires laws on culture, language, education, personal documents and symbols to be adopted with a majority that also includes a majority of MPs present who do not belong to the majority community, effectively giving them a veto.
Three formulas, unequal protection
In practice, the mechanism is not one formula but three different formulas with different thresholds and scopes, and this room for interpretation is being abused. The list of areas requiring double-majority voting in Parliament does not match the corresponding list at municipal level. Education is the clearest example: it is protected by the procedure when decisions are made at central level, but not when they are made by a municipality, even though education competences were transferred to municipalities as early as 2005. Budget decisions, including capital investment, remain entirely outside the Badinter mechanism, even though the distribution of funds among municipalities is one of the country’s greatest problems.
Eleven years without correction
The 2015 review of the Badinter mechanism proposed three major corrections and recommended establishing a group of experts. Eleven years later, none of the recommendations has been implemented and the group has not been formed.
The recurring pattern
Five pillars, one recurring diagnosis.
Five pillars, one diagnosis
Decentralisation has its own financing law, but no constitutional threshold. Equitable representation is set to have its own law, but without sanctions or a budget. Language has its own law, but also an unresolved court case dating from 2019. Education has a funded action plan, but success is measured by the number of activities rather than real contact. The Badinter mechanism has constitutional status, but is manipulated by the very majority it was intended to constrain.
Regression, stagnation and a lack of measurement
Some areas show a regressive trend, especially equitable representation and the use of languages. In others, the problem is not formal backsliding, but stagnation and the absence of measurable mechanisms for financing, monitoring and accountability. Measurement and evaluation themselves are difficult: there has been no comprehensive official assessment of the Agreement’s implementation since 2015, meaning that some of the most frequently cited data today are eleven years old.
The norm exists; implementation does not
The Agreement was incorporated into the Preamble of the Constitution in 2019. This means that guaranteeing it is no longer a matter of a particular government’s political will, but an obligation of every institution. Twenty-five years after the signing, the normative framework is not what is missing. What is missing are meeting the threshold, allocating the budget, respecting deadlines and imposing sanctions for non-implementation.
