To contain the impact of the withdrawal of international forces1, the Afghan government has, at US behest, established militias in most Afghan provinces. With decisive backing from US Special Forces, militia units have been formed in Wardak, Logar, Ghazni, Paktya, Paktika and Kunar Provinces. This strategy was previously tested in Iraq by the same US general, David Petraeus, who in 2011 led ISAF (the International Security Assistance Force) in Afghanistan. In Iraq, such militias, generally from Sunni tribes, were—probably somewhat excessively—credited with stabilizing the security situation.

In Afghanistan, the establishment of militias hearkens back to a long and rather disastrous history. In the 1980s, militias created at Soviet initiative had in fact proved fairly effective in countering the insurrection in some regions (e.g. in Jalalabad, Faryab, Jawzjan and Helmand, among others). In the early 1990s, however, after the fall of the USSR, their growing autonomy from central power in Kabul was a key factor in the fall of Najibullah in 1992.

Analyzing the current dynamics of the militia phenomenon seems especially important at a time when signs of fragility in the Afghan army are multiplying. Analyzing the militias allows examination of the central state’s ability to control paramilitary groups on the periphery; to extricate itself from dynamics of ethnic polarization; and to confront the Taliban insurrection. Using Kunduz as a case study, this article highlights the dynamics at work in the implementation of the militia-based strategy. It argues that short-term militia successes over the Taliban insurrection have come at the price of a further erosion of state institutions—thereby actually, in the long-run, strengthening the armed opposition. On the one hand, the militias appear to strengthen the autonomy of local commanders vis-a-vis Kabul, and lead to a deinstitutionalization of the security forces. On the other, in practice, they promote an ethnicization dynamic that encourages the local population to rally to the Taliban—inasmuch as the latter have built relatively reliable judicial institutions as compared to the Afghan government’s.

Kunduz Province, in Afghanistan’s North, gathers important features key to interpreting national-level trends at the local level. Firstly, the province is ethnically diverse, shared between Pashtuns, Tajiks, Hazaras and Uzbeks2. In the 1980s, but especially in the 1990s, political parties exploited these ethnic identities—even if political affiliations at the time were complex3. Thus Uzbeks mainly belonged to Hezb-i Islami (rather than to Jumbesh-i Milli, as is often the case in the rest of the country.) In Kunduz, Jamiat-i Islami, which in Afghanistan is mainly composed of Tajiks, also drew Pashtuns and Uzbeks to its ranks. Secondly, several political faultlines divide the province: These include, within Jamiat-i Islami, the divide between the networks of Ahmed Shad Massoud (Badakhshan, Takhar, Panjshir) and those linked to Ustad Ata Noor in Mazar-i Sharif; and, at the end of the 1990s, between the Taliban and Commander Massoud. Moreover, Kunduz was at the time one of the key areas featuring a strong international jihadi presence, especially through the Uzbekistan Islamic Movement and Pakistanis evacuated by Islamabad in 2001, through a deal with Washington. Finally, Kunduz is an area where two key dimensions of Taliban strategy are especially visible: on the one hand, the extension of their ethnic base to Uzbeks, Turkmens and Tajiks; on the other, their ability to progressively stifle the governmental apparatus and, in particular, to break the territorial continuity between the North-West (Shiberghan, Mazar-i Sharif) and the North-East (Takhar, Panjshir), i.e. the two regions in which the government retains a relative measure of control.

The Taliban made remarkable breakthroughs in Kunduz between 2008 and 2012. The establishment of militias helped halt their progression between 2012 and 2014. From 2014, however, the Taliban encircled Kunduz city, taking it over for a few days in September 2015. Kunduz was eventually reconquered by the Afghan armed forces with the support of US Special Forces—but the state remained confined to the city limits. The Taliban de facto govern most districts of Kunduz.

To aid understanding of the militia system in Kunduz and its effects on the Afghan state, this article first provides a description of the militia groups (whether legal or not) active in the province during 2014. It then outlines a typology of the various militias—anarchic, ethnic and communitarian—with respect to the ethnic and political frameworks found in the rest of Afghanistan. Finally, it demonstrates how the militia system is part and parcel of the periphery’s process of growing autonomy with respect to the central state.

The ALP (Afghan Local Police) and Illegal Groups

The legal framework for the militias is the Afghan Local Police (ALP) program. This was partly tested in Kunduz, where the province’s first unit was trained in 2011. In 2014, of the 7,000 members of the province’s security forces, nearly 1,200 were militiamen within the ALP; around 2,000 were military (from the Afghan National Army, ANA). The remainder consisted of policemen, customs officials and members of the intelligence services. In practice, pro-government armed groups number far more than those troops registered within the ALP. Due to the limited headcount permitted (1,200 across the province), most Kunduz commanders can register only a part of their troops in the ALP. They thus retain the rest of their armed men outside any legal framework. At the national level, the permitted headcount is of 30,000, with an upper limit of 300 per district—but in practice, the number varies. In many districts, the true figures are 150-200.

One reason for the ALP’s establishment was its low cost per capita as compared to policemen or soldiers: 9,000 Afghanis/month (under $200/member), equipment included. The Afghan Interior Ministry ensures their financing, equipment and salary-payments through a specific budget line—one that is actually directly funded by the US. Since funding levels are highly inadequate, ALP training tacitly implies, at the local level, the population financing its own security. Moreover, local recruitment in principle depends on local notables who submit a list of candidates from their community to the Interior Ministry for approval. In practice, this list generally derives from the networks that local commanders have at their disposal within state institutions to impose their own men. Finally, the ALP, even while registered within the Interior Ministry, remains autonomous in its day-to-day operation.

Further, the ALP framework acknowledges the militia phenomenon by conferring legal status on some share of men in arms. By limiting the authorized headcount, however, it expels the remainder of these paramilitaries outside the legal framework. Running parallel to the ALP, unregistered militia groups remain, in greater numbers and playing a near-analogous role. In 2014, there were around 300 troops in each of Dasht-i Archi and Qala-i Zal and nearly 1,800 in Chahardara—as against a theoretical contingent of 300 ALP in these districts. Some 3,000 troops are, moreover, assumed to be active in Khanabad District—where there is no ALP unit. Commonly, commanders register part of their troops with the ALP—as many as the authorized limit—and thus operate a unit combining registered and unregistered fighters. Ultimately, the ALP have brought about the establishment of a militia system on the boundary between legality and illegality, half-legal, half-not, in a grey zone whose effects in the war depend on local frameworks.

Three Militia Models

A key difficulty in conceptualizing the militia phenomenon in Afghanistan is that the militias (both ALP and informal) are both the product of, and actors in, local scenarios. In Kunduz Province, ethnic identities play a decisive role on two levels. Firstly, the militias mobilize and establish polarization strategies on an ethnic basis—with the support of Western forces that are deeply steeped in a (historically false) vision of Afghanistan as a tribal and ethnic mosaic without a state4. By contrast, the Taliban, whom the government and commanders strive to confine to a Pashtun identity, have systematically recruited and deployed their administrative institutions—in particular judicial ones—to position themselves outside of ethnic divisions. Secondly, the degree of ethnic homogeneity is decisive in understanding the effects of militias’ identity-based strategies in local frameworks—which are largely defined by the borders of districts that determine funding from and nominations by the state and the Western coalition.

Three militia models can thus be identified, according to the degree of political fragmentation and ethnic homogeneity involved. While militias built on a community-based model (i.e. the ethnic homogeneity of the district) generally lead to circumscribing Taliban entrenchment, models of ethnic polarization (e.g. ethnic heterogeneity, or a militia recruiting among the ranks of a minority) and anarchy (e.g. a number of uncoordinated militias, a range of patronage and ethnic networks) exacerbate ethnic conflicts and, in the long term, favour the insurrection. A large proportion of the militias in Afghanistan were recruited according to the logics of these latter two models, explaining—beyond a rise in local conflicts—the insurrection’s progress since the Western withdrawal.

The Community Model (Qala-i Zal)

Qala-i Zal is the only ethnically homogenous district in Kunduz Province. The population is almost entirely Turkmen, and the district is largely composed of desert, with irrigated lands in the East. Qala-i Zal is thus the only district in which it has been possible to apply a community-based strategy in militia recruitment, with the aim of mobilizing the entire district as a unified politico-military player ranged against the Taliban. Mohammad Nabi Gechi established himself as the District commander by creating his own militia. Made up of 300 men, it was first funded by the German army within a forerunner program to the ALP, the Critical Infrastructure Police. Its members were thereafter trained by US Special Forces.

Nabi Gechi was, as a result, credited with having repelled the Taliban who had taken over most of the district in 2012. When, however, the Germans ended their militia-support program in September 2012, Nabi Gechi failed to obtain recognition of his troops within the ALP. The ethnic homogeneity and economic marginality of the district spared him the inter-commander clashes that affected the remainder of the province. Nabi Gechi’s militia thus remained unchallenged, and funds itself at the expense of the local population.

The Anarchic Model (Khanabad)

The context in Khanabad is determined by three factors. Firstly, peopled by Pashtuns, Hazaras and Tajiks, the District is defined by a genuine ethnic mosaic, inherited from the district-formation period during the 20th century, built on population groups who arrived from the east and centre of Afghanistan. At the village level, however, the population is rather homogenous and, with a few exceptions, interethnic marriages are rare. Voting patterns are very closely tied to ethnic belonging, given the political work of village notables. Secondly, the land is irrigated, making it a wealthy district with two or three rice harvests per year. Further, the journey to Kunduz can be made in good conditions (under half an hour from the town centre), allowing for the swift sale of perishable goods. The district thus represents an economic asset, allowing militiamen to survive through extortion on the roads or raising taxes. Thirdly, the district was, from the 1980s to 2001, dominated by the family of Commander Amir Mohammed, killed by Abdul Rashid Dostum’s men in the 1990s, whose patron was Commander Abdul Rassul Sayyaf5. His family remains influential, and is riven by a vendetta between Amir Mohammed’s brother and son. Political positioning in Khanabad derive from these local rivalries and are relatively unstable.

The militias became an important player in 2012, when US Special Forces set about significantly weakening the Taliban, with their own withdrawal in mind. The militia system is primarily composed of illegal groups: about 3,000 armed men according to concurring sources, whereas the ALP can boast only 300—a tenth of militia manpower. Over 100 commanders head these groups, whose size varies from a few troops to several hundred. Nearly every village is divided between several commanders, who are usually representatives of an important local family. Political tensions are driven, not by ethnic dividing lines—the villages, as noted above, being rather homogenous—but to ultra-local rivalries regarding land, marriage or blood-debts inherited from the 1980s.

These militias are grouped into two alliances around two dominant figures of the province, Mir Alam and Mohammed Omar “Pakhsaparan” (“the Wall-breaker”), according to political and ethnic factors. On the one hand, Mir Alam is the historical representative of Jamiat-i Islami, while Mohammed Omar is tied to Sayyaf’s party, the Ettehad, which dominated the district during the 1990s. On the other, Mir Alam is Tajik, and commanders drawn to him are primarily Tajik and Uzbek—while Mohammed Omar prefers to lean on Pashtun and Hazara networks, both often tied to Hezb-i Islami at the local level6. These two factors should not, however, suggest strictly constructed and delimited formations. Much tactical manoeuvring is involved, featuring alliance-reversals and generalized mutual jockeying-for-power. Mir Alam thus also exploits intra-Pashtun conflicts to create allies for himself within that community.

The Ethnic Polarization Model (Dasht-i Archi and Chahardara)

In the two districts of Dasht-i Archi and Chahardara, the establishment of the ALP, and of the militias more broadly, has provoked growing intercommunity tension. In these districts, the establishment of the ALP, designed to curb abuses, has clearly not been enough to prevent a locally-dominant community from oppressing and marginalizing another, by using those armed groups that are in principle designed to fight the Taliban.

Chahardara District is Pashtun-majority, with Tajik, Uzbek and Turkmen minorities. Since the 1990s, partisan affiliations and macro-ethnic divisions have tended to coincide: Pashtuns siding with Hezb-i Islami and Ettehad-i Islami; Tajiks with Jamiat-i Islami; and Uzbeks and Turkmen with Jumbesh-i Melli. In 2011, the ALP selection process in the district was strongly influenced by Mir Alam, resulting in the Pashtuns being excluded. The ALP, who represents only a small part of the 2,000 men in arms in this district, have thus been recruited among ex-Jamiat-i Islami networks (essentially Tajiks) and Jumbesh-i Melli ones (essentially Uzbeks and Turkmen).

No militia is dominant, and frequent clashes occur between the principal commanders. The militias are, moreover, often accused of pillage and theft from the local population. The Taliban have, in response, been able to entrench themselves locally— in particular among the Pashtun majority. Moreover, the Coalition’s military operations have drawn hostile reactions from the district’s Pashtuns. On several occasions, non-Pashtun commanders from neighbouring districts have been called in to participate in these operations—for instance, the Turkmen Nabi Gechi. The Taliban-militia confrontation thus has a distinctly ethnic character.

The situation is similar in Dasht-i Archi District. The context here is determined by the opposition between Uzbeks (30%) and Pashtuns (70%). The former are gathered upstream, providing them access to control of the water supply and to the best lands (rice farming). The Pashtuns are on lower ground, on inferior-quality land (wheat farming). In recent years, the Uzbeks have monopolized all the administrative posts in the district, controlling the ALP selection process. Thereafter, a few dozen Pashtuns were recruited into the ALP based on the US desire to balance out their ranks. These, however, remain marginal. US Special Forces and private security contractors in a base in the district have largely leaned on the Uzbek militias, of which the ALP can boast only half the numbers. The Taliban consequently became, as in Chahardara, the only political alternative for the Pashtun population.

The militias’ role in the local power system

The militias reinforce and formalize an accelerated transformation of power relations in Afghanistan, providing local authorities with the means to emancipate themselves from central power. Mir Alam is now the Kunduz strongman—especially given that he can count upon hundreds of fighters across the province. A Northern Alliance commander in Kunduz against the Taliban at the end of the 1990s, Mir Alam acquired a central position in the province in 2001, thanks to the patronage of ex-Vice-President Fahim, who controlled nominations in the North-East following a deal with Hamid Karzai. He was also supported by US Special Forces to fight the Taliban. Thanks to this backing, in 2007 he managed to avert the nomination of Juma Khan Hamdard7 as Governor and, in 2008, placed one of his commanders, Majid, at the head of the local branch of the intelligence services. In 2009, the latter was replaced by Mohammad Daud, Mir Alam’s brother-in-law. His men could thus extort money from the province’s inhabitants with impunity. With Fahim’s death, and the election of Ashraf Ghani, Mir Alam lost part of his stranglehold on the administrative apparatus: the Governor and Mayor are now openly hostile to him. Both of these, however, have displayed their impotence, leaving him the province’s most powerful commander regardless. The informal policing power that he exercises affects all actors in the province.

At the provincial level, the balance of power between the army and police on the one hand, and the militias on the other, thus clearly favours the latter. Policemen number no more than 200 in Khanabad, 200-300 in Dasht-i Archi and 300 in Chahardara. The ANA has a battalion in Dasht-i Archi, but no troops in Khanabad. Moreover, in some cases, the police sides with one militia against another. In Khanabad, for instance, Manana Hazara, the Chief of Police, formerly of Hezb-i Islami, supports Omar Pakhsaparan against the troops of Qadirek, an ALP commander. In early 2014, the police attacked and killed several of Qadirek’s men who had blocked the road between Aliabad and Khanabad to extort travellers. The ANA does not in general intervene in local conflicts, limiting itself to taking part in large-scale “cleaning-up” operations with the ANP and militias. The latter are thus associated with these operations, sometimes with disastrous results. In Dasht-i Archi, for example, in the winter of 2014, the Police Commander of the province, Andarabi, launched an operation with the help of 200-300 troops from a Khanabad militia led by Yusuf Arjal, a Tajik. These troops pillaged, killed and raped Pashtun civilians, leading to Khalil Andabari’s being sacked, and a stronger Taliban presence.

The Kanam massacre (in Khanabad District, a suburb of Kunduz) is another infamous case. The village lies in an area of Taliban approaches towards the town. In 2012, the Taliban killed some of Mir Alam’s soldiers; Mir Alam sent them Qadirek, an ALP commander, whose reprisals killed around 45 civilians. The case rallied local civil society, but the police arrested only second-rank figures. Moreover, three of those publicly involved in the case were assassinated by Qadirek’s men. In August 2014, the Taliban killed Qadirek and took over this village close to Kunduz with the local population’s support, easing their subsequent operations against the town.


The Taliban’s progress in Kunduz marks the failure of a security-first approach premised on increasing the number of armed groups. Faced with the prospect of an inevitable fall in Western aid, the militias’ role is all the more destabilizing that the setting-up of a diarchy to run the state, between President Ashraf Ghani and “CEO” Abdullah Abdullah, has paralyzed state institutions and limits control over military units that often operate on the edge of brigandry8. While the central administration gradually loses control over the management and resources of whole swathes of the country to local potentates, the army gradually finds itself in competition with armed groups based on ethnic and clientelist dynamics.

The three militia models described above (communitarian, ethnic and anarchic), and the same process of state-weakening, are replicated across the other Afghan provinces. The militia system of which Kunduz was the prototype has been imitated throughout the country. Especially complex and highly unstable local frameworks have emerged according to the anarchic model (Ghazni, Wardak, Helmand, Nangarhar). In these, crisscrossing alliances and confrontations between militias, police and the army may initially have slowed Taliban progression or expelled them from the area—but at the cost of the withdrawal of state institutions. In other cases (Uruzgan, Paktika), identity-based divides are reinforced by abuses by militias made up of the minority ethnic group in the area. Ultimately, this setup favours the Taliban, who become able to exploit local discontent. Moreover, the ALP’s future is dependent on continued US funding. Should this be cut short, extortion and rivalries could increase, gifting the Taliban a yet more favourable situation from which to portray themselves—as they did in the mid-1990s—as an alternative to disorder. The current collapse of the security system in Kunduz could thus prefigure the situation to come in other provinces.

  1. Most of the analysis here derives from the authors’ fieldwork in Kunduz in April 2014 and previous years (in 2009 in particular).
  2. The province’s highly diverse ethnic composition is due to the central government’s settlement policy in the 1920s, when the Kunduz marshes were drained. At the time, the Kingdom transferred population groups from the whole country, sometimes forcibly, to settle them on new lands. The same process occurred in Lashkar Gah, in Helmand Province, when a major hydroelectric dam was built there in the 1960s.
  3. During the war against the USSR in the 1980s, the insurrection split into political parties. In the early 1990s, as political and ethnic dividing lines coalesced during the civil war, these parties became increasingly ethnically homogeneous. Specifically, Jamiat-i Islami established itself as the representative of Tajiks; Hezb-i Islami became majority Pashtun; and Jumbesh-i Milli largely comprises Uzbeks. In some regions—and in Kunduz in particular—this dynamic nonetheless remained contained by local rivalries, partisan loyalties or economic interests.
  4. To the contrary, the history of state-formation since the end of the 19th century demonstrates its central role in Afghanistan and its interlocking with tribal institutions. Moreover, the war that began in 1979 considerably weakened tribal elites, targeted by the USSR and rivalled by rising actors (commanders and clerics).
  5. Abdul Rassul Sayyaf is an 1980s warlord, who built his network of commanders essentially from Gulf funding, which in particular explains his staunchly anti-Shia positions. (These became more moderate on the subject after 2001.) Renowned for his rhetorical ability and corruption, he was among the key figures of Parliament in the 2000s. A candidate in the Presidential elections, he rallied Abdullah in the second round.
  6. The Hazaras are generally Shia. Their a priori surprising affiliation to Hezb-i Islami, a Sunni Islamist party composed mainly of Pashtuns, has also been documented in other regions, in particular in the south of Ghazni.
  7. Originating from Balkh Province, Hamdard is Pashtun and formerly of Hezb-i Islami, Jamiat-i Islami’s historical rival. His nomination was thus—probably rightly—seen as a means for Karzai to regain a minima control of the province.
  8. Following the Afghan presidential election, the two candidates came to a deal under US auspices. This involved the creation of a new “head of government” position, the Chief Executive Officer, splitting the executive’s powers into two. In particular, nominations to administrative positions, whose role in the Afghan political system is decisive, must now be approved by both heads of the executive branch. The result has since been paralysis in many senior civil service posts.