One Country, Too Many Nations

After World War II, the United States spent roughly $150 billion in today's dollars to help rebuild 16 European economies through the Marshall Plan.¹ Most of those countries became stable democracies. Twenty years and $2.3 trillion later, the United States walked away from Afghanistan as the central government collapsed in 11 days.²

Why did the same playbook work in one place and fail in another? Most explanations point to the timeline, the corruption, or the geography. Those answers are not wrong. They are incomplete. The political system the U.S. helped build in Afghanistan was structurally incompatible with the country it was governing. A highly centralized state was bolted onto an ethnically polarized, geographically fragmented society. By 2021, the foundation gave out.

The standard explanations for Afghanistan's collapse are well-rehearsed. Chatham House calls it an overambitious project with unrealistic timelines, corrupt local governance, and weak regional cooperation.³ The George C. Marshall Center cites rugged terrain, tribal loyalties, and historical resistance to central authority.⁴ Both are right. Neither answers a more basic question: why did the political institutions chosen for Afghanistan in 2001 look more like France's than Switzerland's?

This piece argues the failure was structural, and the structure was a choice. The post-2001 system concentrated power in Kabul under a single president with the constitutional authority to appoint provincial governors, judges, and security chiefs.⁵ That works in a homogeneous country. Afghanistan is not one.

Alberto Alesina and Enrico Spolaore's framework in The Size of Nations sets up the trade-off plainly. Larger states get more economies of scale: bigger markets, cheaper defense, spreading the cost of public goods across more taxpayers. They also get more heterogeneity, which means more people who disagree about how the state should be run. Beyond a certain point, the cost of forcing one set of policies on a diverse population outweighs the benefits of being a single country.⁶

The standard fix is federalism. Push contentious decisions like language, education, religious courts, and local policing down to subnational units. Keep the federal layer for things that genuinely need scale: currency, defense, foreign affairs. Spolaore frames federal systems as "shared rule and self-rule," a way to keep a single state intact while giving disaffected groups a stake in the system.⁷

Two related concepts matter for what comes next. Ethnic fractionalization measures how many distinct groups exist in a country. A country can be highly fractionalized but still peaceful, if no group is large enough to dominate. Ethnic polarization, introduced by José Montalvo and Marta Reynal-Querol, measures something different: whether two or three large groups face off against each other. It's polarization, not fractionalization, that drives civil war.⁸

There has been no national census in Afghanistan since 1979, so all population shares are estimates.⁹ The most commonly cited ranges:

The geography matters because it isn't random. Each major group has a homeland. Pashtuns dominate the south and east, around Kandahar and along the Pakistan border. Tajiks are concentrated in the northeast (Panjshir, Badakhshan) and in major cities. Hazaras live in the central highlands, an area sometimes called Hazaristan, and are mostly Shia. Uzbeks and Turkmen are in the northern plains. Smaller groups (Aimaq, Baluch, Nuristani, Pashai) cluster in their own pockets.⁹

The Hindu Kush mountain range carves the country into valleys and plateaus where central authority has, for centuries, been more theory than practice. This is what Jared Diamond, and more recently Fernández-Villaverde and co-authors, call "fractured land."¹⁰ Rugged terrain hinders central state formation. The same logic that explains why Europe became 40-plus states while China became one applies inside Afghanistan: the country's own topography pushes it toward decentralization.

A foreign saying captures the pattern: Afghanistan is "easy to invade, but hard to conquer." The British, the Soviets, and eventually the Americans all discovered they could take Kabul. They could not take the valleys.

Despite all of this, the political system designed at the Bonn Conference in December 2001 concentrated power. The Bonn Agreement formally called for "broad representation of Afghanistan's different ethnic, regional, and political groups."¹¹ In practice, no mechanism was built to enforce that. The 2004 constitution gave the president the authority to appoint provincial governors, judges, and security chiefs. National elections, not federal devolution, were the only legitimacy mechanism.

This is in a country where one ethnic group has held political power, almost without interruption, since 1747. Pashtun kings, autocrats, and presidents have run Afghanistan for roughly 274 of the last 279 years. The only two exceptions were brief and ended violently. Habibullah Kalakani, a Tajik, ruled for nine months in 1929 before being executed.¹² Burhanuddin Rabbani, also Tajik, was president from 1992 to 1996 during the civil war that brought the Taliban to power.⁹

Under Hamid Karzai (2004–2014) and then Ashraf Ghani (2014–2021), centralization deepened. The 2004 constitution let the president appoint not just cabinet ministers but provincial governors, district officials, and even schoolteachers, an arrangement scholars later identified as the single biggest design flaw in the post-Taliban state.⁵ Ghani relied on a tight circle of mostly Pashtun advisors and systematically marginalized non-Pashtun warlords and provincial leaders who had played central roles in the early post-Taliban government. Tajiks, Uzbeks, and Hazaras who held key positions in the early 2000s found themselves frozen out by the late 2010s.¹³

The dynamic was more specific than insurgent opportunism. The Taliban is, at its core, a Pashtun ethnic project. Among 1,212 senior and mid-level Taliban officials tracked by the Middle East Institute after the 2021 takeover, roughly 90% are ethnic Pashtuns; Tajiks are 5.3% and Uzbeks about 3%.¹⁴ Non-Pashtuns did not enable the Taliban's rise. They were on the receiving end of it. What gave the Taliban its opening came from inside the Pashtun political class itself, in two ways.

The first was structural. Ghani's government had spent years actively sidelining the non-Pashtun figures who had once been the country's main defense against the Taliban. Senior advisor Ahmad Zia Massoud, an ethnic Tajik, was fired. First Vice President Abdul Rashid Dostum, the most powerful Uzbek leader, was pushed into self-imposed exile. Hazara leader Mohammad Mohaqiq lost his role in government coordination. Northern strongman Atta Noor, the Jamiat power broker who had governed Balkh province for over a decade and successfully resisted the Taliban in the 1990s, was forced out as governor in a drawn-out 2017 standoff with Kabul.²⁰ By 2021, the warlord networks that had held the north for twenty years had been hollowed out by the very government they were supposed to defend. When Ghani called on Noor and Dostum to organize a last stand in August 2021, both men's militias were routed within days and both fled across the border into Uzbekistan.²¹

The second was ideological. Inside the Pashtun political class, a strain of ethnic nationalism saw a Taliban return as preferable to a multi-ethnic republic with empowered Tajiks, Uzbeks, and Hazaras. Many Pashtun officials in Ghani's government had little objection to the Taliban's program of Pashtun primacy, even where they disagreed with the movement's tactics or theology.²² Some quietly accommodated the handover. When provincial capitals started falling, the army on paper had more troops, more equipment, and more training than the Taliban. What it didn't have was a Pashtun elite committed to defending an inclusive constitution or a minority population that believed the state in Kabul represented them.

Other multi-ethnic states made different institutional choices, with different results.

India: the success story

After independence in 1947, India adopted a federal constitution. In 1956, the States Reorganization Act redrew internal state boundaries along linguistic lines.¹⁵ Tamils, Punjabis, Marathis, Bengalis, and many others got states where their language and culture were the majority. The fact that no single Indian group exceeds about 40% of the population helped. But the institutional design did the work. Decentralization didn't weaken India. It gave each group a stake in the system.

Iraq: a partial success

After the 2003 U.S. invasion, Iraq adopted a federal constitution in 2005 that recognized the Kurdistan Regional Government as a semi-autonomous region with its own government, parliament, and armed forces.¹⁶ This kept the Kurds inside the Iraqi state without forcing them into a unitary structure. Iraq still went through a sectarian civil war and the ISIS insurgency, but the Kurdish question, the most likely candidate for triggering a breakup, was managed institutionally rather than militarily.

Ethiopia: the cautionary tale on the other side

After the Derg regime fell in 1991, Ethiopia adopted ethnic federalism with a constitutional right to secede.¹⁷ It worked for a while. Then each ethnic region built up its own institutions, identities hardened, and inter-regional rivalry replaced national identity. The Tigray war is the latest chapter.¹⁸ Federalism without a unifying overlay can entrench the divisions it was meant to manage.

Yugoslavia: the cautionary tale at the limit

Marshal Tito's federation of six republics held together for 45 years through authoritarian glue. Once that glue dissolved with the end of communism, the republics seceded one by one, often violently.¹⁹ Federal lines became international borders. The lesson is not that federalism failed. The lesson is that federalism alone, without a national identity strong enough to compete with ethnic ones, is fragile.

The takeaway for Afghanistan isn't "copy India" or "do what Iraq did." It is more basic. Of the four serious comparators, three were federal and one had to choose federal lines under duress. None of them tried to govern a polarized, mountainous, multi-ethnic society from a single capital with a single president holding most of the levers. Afghanistan did. The result is in the history books.

The Taliban regime that took over in 2021 has doubled down on the same structural mistake, with worse intent. It is more centralized than the republic it replaced, more exclusionary, and more violent in enforcing uniformity on a country that has never accepted it. The pattern is the same. Only the cruelty has scaled.

A stable Afghanistan, if there ever is one, will not look like the 2004 constitution. It will not look like the Taliban's emirate. It will need a structure that admits what every Afghan already knows: this is not one nation. It is a federation of nations that has been forced into a single political unit, and any system that pretends otherwise is building on the same fractured ground that has buried every previous attempt.

That lesson travels. The trade-off Alesina and Spolaore described, between the gains from being a single country and the costs of governing diverse populations as if they were one, is not unique to Afghanistan. It shows up wherever an outside power tries to rebuild a state without first asking whether the state's design fits the country underneath it. Afghanistan is the cleanest negative case in the modern record. It would be useful to remember that the next time someone proposes a $2 trillion solution for a $20 question.