What makes territorial autonomy last? Åland to Hong Kong

Territorial autonomy, the grant of self-government to a part of a state's territory without independence, is one of the oldest tools for managing national, linguistic and island differences. Some arrangements have lasted a century with little violence. Others have collapsed into secession crises, been hollowed out by the centre or become way-stations to independence. This study compares eight cases, Åland, South Tyrol, Greenland, Bougainville, Catalonia, Scotland, Hong Kong and, through the comparative literature, the autonomies of the South Caucasus, and asks what distinguishes durable arrangements from fragile ones. The evidence points to five factors: an external anchor or guarantee, entrenchment that the centre cannot change alone, a clear and regularly renegotiated division of powers and money, power-sharing inside the autonomous unit where it is itself diverse, and the absence of a strong grievance that autonomy is a substitute for, rather than an expression of, self-determination. Autonomy is neither a reliable cure for secessionism nor its cause; its effect depends heavily on design and timing.

Key findings

• The most durable arrangements in the sample, Åland (from 1920–1921) and South Tyrol (1946, 1972), combine an international agreement, entrenchment against unilateral change and detailed protection of language and land.
• Åland's Autonomy Act can be amended only by the Finnish Parliament in constitutional order and with the consent of the Åland Parliament. South Tyrol's 1972 statute moved most powers from the region to the provinces; the province retains almost 90% of taxes levied in its territory.
• Greenland shows a staged path: home rule in 1979 (73% yes in a referendum), self-government in 2009 (76% yes), and a legal route to independence decided by the people of Greenland. Fiscal dependence on Denmark remains the main brake on independence.
• Bougainville's 2001 peace agreement produced autonomy and a referendum, in which 98.31% voted for independence in 2019. The result was tabled in the Papua New Guinea parliament on 27 August 2026, but no ratification vote had been held by late October 2026, and the leaders' forum moved its recommended independence date from 2027 to 2030.
• Catalonia and Scotland show that negotiated autonomy in democracies can coexist with strong independence movements. Their trajectories diverged because Spain's constitutional framework allowed no legal referendum, while the UK agreed one in 2014 but later, according to its Supreme Court, kept the power to hold another at Westminster.
• Hong Kong shows the limit of autonomy without independent guarantees: the 1984 Joint Declaration promised a high degree of autonomy for 50 years, but the 2020 National Security Law, the 2021 electoral overhaul and the 2024 Article 23 ordinance concentrated power in Beijing well before 2047.
• Cross-national studies find that autonomy tends to prevent conflict where there is no conflict history, but post-conflict autonomy on its own may be "too little, too late"; regional parties, boundary design and the existence of a "core" region condition the effect.

Research question and scope

What makes territorial autonomy last? By "last" we mean that the arrangement continues to function as autonomy, neither collapsing into violent conflict nor being hollowed out by the central state nor becoming a transitional step to secession against the will of one of the parties. Independence by agreement is not necessarily a failure, and we treat Greenland and Bougainville, where a legal route to independence is built into the arrangement, as a distinct category.

We focus on asymmetric autonomy, where one or a few regions have powers others do not, rather than on full federations, which a companion study covers. The cases are:

• Åland (Finland), a demilitarised Swedish-speaking island autonomy.
• South Tyrol (Italy), a German- and Ladin-speaking province with a bilateral international anchor.
• Greenland (Denmark), an Arctic autonomy with a statutory route to independence.
• Bougainville (Papua New Guinea), a post-conflict autonomy with an independence referendum.
• Catalonia (Spain) and Scotland (United Kingdom), devolved regions in democracies with strong independence movements.
• Hong Kong (China), a special administrative region whose autonomy was promised by treaty.

We draw on comparative work by Ruth Lapidoth, Hurst Hannum, Yash Ghai, Marc Weller and Stefan Wolff, Svante Cornell and quantitative studies of autonomy and conflict.

Background

Lapidoth defined autonomy as a means of diffusing powers to preserve the unity of a state while respecting the diversity of its population. Hannum's study Autonomy, Sovereignty, and Self-Determination framed it as a way of accommodating competing claims without either side abandoning its principles. Ghai's edited volume Autonomy and Ethnicity (2000), with chapters on Canada, India, China, South Africa, Ethiopia, Papua New Guinea, Spain and others, stressed that autonomy is one of the most sought-after and most resisted devices of conflict management.

Two opposite worries run through the literature. One holds that autonomy is a concession that satisfies moderate demands and keeps the state together. The other, put most sharply by Svante Cornell in 2002, holds that autonomy can create the institutions, elites and identities that make secession more likely. Cornell's evidence came from Georgia, where of five ethnic minority populations, the two that had autonomous status, the Abkhaz and South Ossetians, were the only ones to engage in armed conflict with the government.

Stefan Wolff and Marc Weller, in Autonomy, Self-Governance and Conflict Resolution (2005) and later work, argued that the debate is too binary. In many recent settlements, territorial self-governance is the centre of a "complex power-sharing" package that also includes representation at the centre, human rights guarantees, international involvement and dispute-resolution procedures. Whether autonomy works depends on the whole package.

Method and sources

We use a structured comparison of cases chosen for variation on outcome (durable, transitional, contested and eroded) rather than for representativeness. For each case we identify the founding instrument, the guarantee mechanism, the scope of powers and fiscal arrangements, the internal power-sharing rules and the trajectory to 2026, using primary documents where available (the Joint Declaration, Spanish court rulings and laws, UK Supreme Court records, the Åland Parliament's own account of its autonomy) and secondary sources for the rest. We then set the cases against three cross-national studies of autonomy and conflict.

The comparison has limits. It includes famous cases, which tend to be either unusually successful or unusually troubled. It cannot separate the effect of autonomy from the effect of wealth, geography or the democratic quality of the host state, which vary across the cases.

Finding 1: durable autonomies are anchored outside the host state's ordinary politics

Åland

After Finland's independence in 1917, the Swedish-speaking population of the Åland Islands pressed for reunion with Sweden. Finland responded with an Autonomy Act in 1920 and the dispute went to the League of Nations, whose Council decided in 1921 that the islands would remain Finnish but that Finland must guarantee the population's Swedish language, culture and traditions. A convention signed on 20 October 1921 by Sweden, Finland and several other European states confirmed the islands' demilitarised and neutral status.

The arrangement has since been renewed twice, by new Autonomy Acts in 1951 and 1991; the current Act entered into force on 1 January 1993. Two design features stand out. First, the Act can be amended only by the Finnish Parliament in the order required for constitutional amendments and only with the consent of the Åland Parliament. Second, it protects the language and land base of the community through the right of domicile (hembygdsrätt), which is needed to vote for or stand in the Åland Parliament and to own rural real estate. People who move to Åland can acquire it after five years of residence if they are Finnish citizens with sufficient Swedish. When Finland joined the European Union, Åland held its own referendum on 20 November 1994, which approved membership with 73.64% of the vote, and a protocol in Finland's accession treaty preserved its restrictions on land ownership and certain business rights.

The arrangement is not frozen. A parliamentary committee chaired by former Finnish president Tarja Halonen worked on a new Autonomy Act from 2013, reporting in 2015 and 2017; joint working groups of the Ministry of Justice and the Government of Åland continued the work from 2023, and a group appointed in January 2025 was due to complete a draft by May 2026. A population of about 30,000 has thus kept the same basic settlement for a century while renegotiating its details.

South Tyrol

South Tyrol's autonomy also has an international root: the agreement signed in Paris on 5 September 1946 by Austrian foreign minister Karl Gruber and Italian prime minister Alcide De Gasperi. It promised the German-speaking population equal rights, schooling in their mother tongue, bilingual public offices and names, and regional legislative and executive autonomy.

The first autonomy statute of 1948 disappointed German speakers because it placed the province within a larger region, Trentino-Alto Adige, with an Italian-speaking majority. Austria took the question to the United Nations in 1960; a terrorist bombing campaign began in the early 1960s; negotiations resumed and produced a package of measures in 1969. The second autonomy statute of 1972 transferred most legislative and executive powers from the region to the two provinces. The province of Bolzano retains almost 90% of the taxes levied in its territory. In 1992 Austria and Italy formally ended their dispute, on the basis that the package had been implemented.

Inside the province, autonomy is combined with group-based power-sharing. Seats in the provincial government must reflect the proportional strength of the German and Italian language groups in the provincial parliament, Ladin representation is guaranteed separately, and public posts are allocated in proportion to the language groups. Stefan Wolff describes South Tyrol as a case of complex power-sharing in which territorial autonomy, group rights and an international anchor reinforce each other. In the language-group census of 2024, the provincial statistics institute ASTAT counted 68.61% of valid declarations for the German group, 26.98% for the Italian and 4.41% for the Ladin, against about 69.4%, 26.1% and 4.5% in 2011. The declarations cover Italian citizens only; with a growing number of foreign residents, the groups' shares of the whole resident population are lower, a reminder that migration is changing the society the 1972 settlement was designed for.

Why external anchors matter

In both cases, the original settlement was a bargain between states, monitored by an international body (the League of Nations for Åland) or by a kin state (Austria for South Tyrol). That anchor did three things. It made the host state's commitment more credible, because breaking it would have international costs. It gave the minority an external channel for complaints that did not require secession. And it slowed down change, because changes had to be agreed by more parties.

Finding 2: some autonomies are designed as staging posts

Greenland

Greenland's autonomy was built in stages. In a referendum on 17 January 1979, 73.06% of voters approved home rule, which took effect on 1 May 1979. Greenland then used its new powers to leave the European Economic Community, which it did in 1985. In a second referendum on 25 November 2008, 76.22% of voters, on a turnout of 71.96%, approved a broader self-government arrangement. The Act on Greenland Self-Government came into force on 21 June 2009. It recognised Greenlanders as a people under international law, made Greenlandic the official language, allowed Greenland to take over further fields such as police and courts, and set out a procedure under which a decision on independence is taken by the people of Greenland, followed by negotiations with the Danish government and the consent of the Danish parliament.

Money is the binding constraint. Section 5 of the 2009 Act fixes Denmark's annual block grant at DKK 3,439.6 million at 2009 price and wage levels, indexed each year. Under section 8, if Greenland receives revenue from mineral resources, the grant is reduced by half of the revenue above DKK 75 million a year (that threshold is also indexed). Independence thus depends on whether Greenland can replace the grant with its own revenues.

Recent events show how external pressure can change the politics of autonomy. In the run-up to the election of 11 March 2025, US President Donald Trump's stated interest in acquiring Greenland dominated debate; a January 2025 poll found 85% of Greenlanders opposed joining the United States. The Democrats (Demokraatit), who favour gradual independence, won 30.26% and their leader Jens-Frederik Nielsen formed a broad coalition. In this case outside pressure appears to have strengthened cohesion within the Danish realm rather than accelerated separation, at least in the short term.

Bougainville

Bougainville's autonomy came out of a civil war that lasted from 1988 to 1998. The Bougainville Peace Agreement of 2001 provided for an autonomous government, weapons disposal and a deferred referendum on independence, with the final decision reserved to the national parliament of Papua New Guinea. Anthony Regan's study Light Intervention (2010) credits the process to a combination of local ownership, a modest international presence and a deliberately slow, staged design.

The referendum was held from 23 November to 7 December 2019. Of 207,213 registered voters, 87.38% took part; 176,928 voters (98.31%) chose independence and 3,043 (1.69%) greater autonomy. Negotiations followed. In 2021 the two governments agreed a timetable for independence by 2027, and the Era Kone Covenant set out a ratification process with an independence window of 2025 to 2027. The national parliament and the Autonomous Bougainville Government have disagreed on whether ratification requires a two-thirds or a simple majority. In March 2025 the Bougainville Independence Leaders Consultation Forum recommended 1 September 2027 as the date of independence; a Bougainville Constituent Assembly was established in November 2025; and in July 2026 the forum moved its recommended date to 1 September 2030. On 6 August 2026 the Joint Supervisory Body signed its final consultation documents. On 27 August 2026 the minister for Bougainville affairs, Manasseh Makiba, tabled the referendum result, the joint consultation report and a "Melanesian Framework" in the national parliament; under that framework an affirmative vote would lead first to a transitional phase of self-government, not immediate independence, and the two governments had not agreed on what happens if parliament rejects the result. Parliament adjourned after debate without setting a date for a vote. When debate resumed in October 2026, Makiba told MPs that the decision required a three-quarters majority; Prime Minister James Marape said the date of the vote would be set in consultation with the Bougainville leadership. No vote had been held by late October 2026.

What staged autonomy shows

Greenland and Bougainville turn the usual worry about autonomy on its head. In both, a legal route to independence is part of the bargain. That may make the arrangement more stable, because the question of final status is channelled into a procedure rather than left to confrontation. The risk is a long limbo: if the agreed procedure stalls, as in Bougainville, frustration can grow on both sides.

Finding 3: democratic devolution can coexist with strong independence movements, but the route to a vote matters

Catalonia

Catalonia's autonomy rests on the Spanish Constitution of 1978 and its Statute of Autonomy. A new statute, approved by the Spanish and Catalan parliaments and by referendum in Catalonia, became Organic Law 6/2006. In its judgment 31/2010 of 28 June 2010, on an appeal by 99 deputies of the Partido Popular, the Constitutional Court struck down or reinterpreted parts of the statute and held that the preamble's references to Catalonia as a nation had no legal effect. Support for independence grew in the following years.

On 7 September 2017 the Constitutional Court suspended the Catalan law convening an independence referendum. The vote went ahead on 1 October 2017: about 90% of votes cast supported independence, on a turnout of 43.03%, with many opponents abstaining. After a declaration of independence by the Catalan parliament later that month, the Spanish government used Article 155 of the Constitution to take over the Catalan administration and called regional elections. Several leaders were prosecuted; others left Spain.

A partial reconciliation followed. Organic Law 1/2024 of 10 June 2024 granted an amnesty for acts connected with the independence process. In judgment 137/2025 of 26 June 2025 the Constitutional Court upheld most of the law while declaring specific provisions unconstitutional, and further rulings followed later in 2025.

Scotland

Scotland's devolution was approved in a 1997 referendum and created by the Scotland Act 1998. In 2014, after an agreement between the UK and Scottish governments, Scots voted on independence; 55.3% voted No. After the 2016 Brexit referendum, in which Scotland voted to remain in the EU, the Scottish government sought a second vote. On 23 November 2022, in a reference by the Lord Advocate ([2022] UKSC 31), the UK Supreme Court held that a Scottish Parliament bill for an independence referendum would relate to matters reserved to Westminster and was outside the Scottish Parliament's competence.

Comparing the two

Both cases are democratic, both have broad devolved powers, and in both the independence movement grew during the period of autonomy. The difference lies in the procedure for a final-status vote. The UK, which has no entrenched written constitution, agreed a vote once. Spain's constitution, as interpreted by its Constitutional Court, does not allow a unilateral regional referendum on secession. In Scotland the result was a legal vote whose outcome both sides accepted; in Catalonia, a confrontation followed by prosecutions and then amnesty. Neither case shows that autonomy caused independence sentiment; both show that the absence of an agreed procedure for final-status disputes raises the cost of disagreement.

Finding 4: autonomy without independent guarantees can be withdrawn

Hong Kong

The Sino-British Joint Declaration, signed on 19 December 1984, provided that the Hong Kong Special Administrative Region would enjoy a high degree of autonomy, except in foreign and defence affairs, and that its basic policies would remain unchanged for 50 years. These commitments were written into the Basic Law, under which Hong Kong was handed over on 1 July 1997. In 2003 a government bill implementing Article 23 of the Basic Law, on national security, was shelved after mass protests.

From 2020 the arrangement changed fundamentally. The National Security Law of 30 June 2020 was imposed by the national legislature. On 11 March 2021 the National People's Congress voted 2,895 to 0, with one abstention, to overhaul Hong Kong's electoral system: the Legislative Council grew from 70 to 90 seats, directly elected geographical seats fell from 35 to 20, the Election Committee gained 40 seats of its own, and a committee was created to vet candidates, with decisions that cannot be appealed. On 19 March 2024 the Legislative Council passed the Safeguarding National Security Ordinance by 89 votes to 0, completing the Article 23 legislation; it took effect on 23 March 2024. The UK government said that the law undermined the Joint Declaration.

Why Hong Kong differs

Hong Kong had an international treaty, but the treaty partner had no practical means of enforcement and the interpretation of the Basic Law rested with the national legislature of the sovereign state. Its autonomy also covered a political system in which the central government had a strong interest. Åland and South Tyrol, by contrast, combined international commitments with democratic host states, domestic entrenchment requiring the autonomy's consent, and narrower stakes for the centre.

Comparison of cases

Finding 5: what the cross-national evidence says

Autonomy prevents conflict more than it ends it

Cederman, Hug, Schädel and Wucherpfennig (2015), analysing ethnic groups worldwide since the Second World War, found that decentralisation has a significant conflict-preventing effect where there is no prior conflict history. In post-conflict settings autonomy concessions on their own may be "too little, too late", but they can help when combined with power-sharing at the centre. Using colonial history as an instrument, they suggest naive analysis underestimates the pacifying effect of decentralisation.

Regional parties and boundaries condition the effect

Dawn Brancati (2006), in a study of 30 democracies from 1985 to 2000, found that decentralisation reduces ethnic conflict and secessionism directly but can increase it indirectly by strengthening regional parties, which reinforce regional identities and mobilise for secession. A 2026 study by Andreas Juon, using spatial data on regional boundaries in 132 multi-ethnic countries from 1946 to 2017, found that aligning administrative boundaries with ethnic settlement patterns helps defuse separatist conflicts, while minorities split across regions or outnumbered within them are more likely to keep mobilising.

Autonomy can feed separatism under some conditions

Cornell's argument from the Caucasus remains a warning: autonomy gives minority elites state-like institutions, resources and a platform. Where the minority's goal is independence and the host state is weak or hostile, autonomy can be a stepping stone. The Georgian autonomies ended in war; Hong Kong's erosion went the other way, with the centre reclaiming powers.

What the evidence does not show

• It does not show that any single factor guarantees durability. Åland and South Tyrol combine several favourable features, including small populations, wealth and democratic host states, and their success cannot be attributed to any one of them.
• It does not show that autonomy causes independence movements. Independence support in Catalonia and Scotland grew during autonomy, but there is no counterfactual showing it would have been weaker without it.
• It does not settle Bougainville's outcome. The national parliament's decision on ratification was pending when this study was written, and the timetable has moved more than once.
• It does not make Hong Kong typical. It is a case of autonomy in a one-party state with a strong interest in its political system; it says little about autonomy within democracies.
• The cross-national findings are averages. They hold across many groups and decades and allow exceptions in both directions.

Implications for institution design

1. Entrench against unilateral change. Require the consent of the autonomous unit for amendments, as Åland does, or place the arrangement in an international agreement with a body able to hear complaints.
2. Write the money down. Durable arrangements specify fiscal transfers and revenue shares. Fiscal ambiguity becomes a recurring source of conflict.
3. Protect the community's base. Language rights, land rules and schooling were central to Åland and South Tyrol. Design them with an eye to how the population will change through migration.
4. Share power inside diverse autonomies. Where the autonomous unit is itself mixed, as South Tyrol is, proportional representation in government and public employment protects the local minority.
5. Provide a procedure for final-status disputes. Greenland's statutory route and the UK's 2014 agreement contrast with Spain's confrontation in 2017. A clear, lawful procedure lowers the stakes of disagreement, even if it is demanding.
6. Combine autonomy with representation at the centre. The quantitative evidence suggests autonomy works best when the group also has a voice in national institutions.
7. Do not rely on promises the guarantor cannot enforce. Hong Kong shows that a treaty without enforcement and with interpretation left to the sovereign offers little protection.

Open questions

• Will Åland's new Autonomy Act, if adopted, show how a century-old arrangement can be modernised without reopening its foundations?
• How will South Tyrol's power-sharing rules adapt to a growing share of residents who belong to none of the three recognised language groups?
• Can Greenland build a revenue base that makes its statutory independence route realistic, and how will great-power pressure affect that choice?
• What happens to a post-conflict autonomy, such as Bougainville's, when a referendum result is not ratified on the agreed timetable?
• Is there a design that can give autonomy credible protection inside non-democratic states?

Sources

1. Lapidoth, R. (1997). Autonomy: Flexible Solutions to Ethnic Conflicts. Washington, DC: United States Institute of Peace Press.
2. Hannum, H. (1996). Autonomy, Sovereignty, and Self-Determination: The Accommodation of Conflicting Rights (revised edition). University of Pennsylvania Press.
3. Ghai, Y. (ed.) (2000). Autonomy and Ethnicity: Negotiating Competing Claims in Multi-Ethnic States. Cambridge University Press.
4. Weller, M. and Wolff, S. (eds) (2005). Autonomy, Self-Governance and Conflict Resolution. Routledge.
5. Wolff, S. (2009). Complex power-sharing and the centrality of territorial self-governance in contemporary conflict settlements. Ethnopolitics 8(1), 27–45.
6. Wolff, S. (2008). Complex power sharing as conflict resolution: South Tyrol in comparative perspective. In J. Woelk, F. Palermo and J. Marko (eds), Tolerance through Law: Self Governance and Group Rights in South Tyrol. Martinus Nijhoff.
7. Cornell, S. E. (2002). Autonomy as a source of conflict: Caucasian conflicts in theoretical perspective. World Politics 54(2), 245–276.
8. Cederman, L.-E., Hug, S., Schädel, A. and Wucherpfennig, J. (2015). Territorial autonomy in the shadow of conflict: too little, too late? American Political Science Review 109(2), 354–370.
9. Brancati, D. (2006). Decentralization: fueling the fire or dampening the flames of ethnic conflict and secessionism? International Organization 60(3), 651–685.
10. Juon, A. (2026). Unfulfilled aspirations: autonomy, regional boundaries, and self-determination conflicts. Journal of Conflict Resolution (online first).
11. Ålands lagting. Ålands lagting: the Parliament of Åland and ongoing revision of the Autonomy Act.
12. Wikipedia. Åland (used for the 1994 EU referendum, right of domicile and population figures).
13. Wikipedia. South Tyrol (used for the 1972 statute's fiscal terms).
14. Wikipedia. 2008 Greenlandic self-government referendum, 1979 Greenlandic home rule referendum and 2025 Greenlandic general election.
15. Regan, A. J. (2010). Light Intervention: Lessons from Bougainville. Washington, DC: United States Institute of Peace Press.
16. Wikipedia. 2019 Bougainvillean independence referendum (results and 2021–2026 ratification process).
17. Blades, J. (2026). PNG's parliament Bougainville vote still many months away. RNZ Pacific / Asia Pacific Report, 28 August 2026; and PINA (2026). PNG Parliament urged to balance Bougainville independence with national unity, 22 October 2026.
18. Denmark (2009). Lov om Grønlands Selvstyre (Act No. 473 of 12 June 2009), sections 5 and 8.
19. ASTAT (2024). Risultati censimento linguistico 2024. Istituto provinciale di statistica, Bolzano, 6 December 2024.
20. Tribunal Constitucional de España (2010). Sentencia 31/2010, de 28 de junio, on the 2006 Statute of Autonomy of Catalonia. BOE no. 172, 16 July 2010.
21. Spain (2024). Ley Orgánica 1/2024, de 10 de junio, de amnistía para la normalización institucional, política y social en Cataluña, with the BOE's record of Constitutional Court rulings including STC 137/2025.
22. Wikipedia. 2017 Catalan independence referendum.
23. Supreme Court of the United Kingdom (2022). Reference by the Lord Advocate of devolution issues under paragraph 34 of Schedule 6 to the Scotland Act 1998, [2022] UKSC 31.
24. Governments of the United Kingdom and the People's Republic of China (1984). Joint Declaration on the Question of Hong Kong.
25. Wikipedia. 2021 Hong Kong electoral changes and Safeguarding National Security Ordinance.