The international effort to build a state in Afghanistan carried tremendous financial and human costs. The US spent 2.3 trillion dollars, and around 176,000 people, mostly Afghans, were killed . If we consider development assistance alone, Afghanistan received 145 billion dollars (or about 4,000 USD per Afghan), which, in real terms, amounts to substantially more than was spent under the Marshall Plan. Nonetheless, more than half of the country’s population – an estimated 22.8 million people – now face life-threatening food insecurity as the economy crumbles, while many of the human rights advances achieved during the last 20 years are being quickly reversed . An unprecedented international effort to modernize Afghan institutions has ended politically almost where it started: with Afghanistan under the control of a brutal Taliban theocracy.
It is hard to imagine a scenario that more starkly calls into question whether stable democracy is possible in Afghanistan. Indeed, it makes a case that any such effort, no matter how it is executed, may not be worth the phenomenal financial and human costs. Such pessimism is reinforced by the fact that Afghanistan has managed only two peaceful transitions since 1747: in 1901 when Habibullah Khan inherited the throne and in 2014 when Ashraf Ghani was elected .
A substantial body of political economy research – much of it written post 9/11 – argues that insurgent conflicts, like that in Afghanistan, are best understood as violent contests for state control [4, 5, 6, 7, 8, 9]. As such, success is much more a question of politics and popular support than one of military superiority. For peace to endure, it must be palatable to any potential spoilers, and, correspondingly, provide them with an acceptable degree of political voice and power. If the design of the state is fundamentally out of sync with underlying social and political power dynamics, it has little chance for success. In such a scenario, the state both has limited incentive to invest in capacity  and cannot navigate the traditional forces which block reform and oppose the development of a modern state .
This article contends that the reason democracy failed to take root in Afghanistan is because of three specific design choices – which we call the three sins – that ensured Afghanistan’s de jure political institutions did not cohere with the underlying allocation of de facto political power. Collectively, these three sins put a political solution that might achieve a broad enough consensus to work out of reach with disastrous consequences for the Afghan state.
First, the Taliban were explicitly excluded from peace negotiations and constitutional deliberations, restricted from any political participation by a provision in the constitution, and otherwise disallowed from any form of non-violent participation. This served to disenfranchise a significant percentage of the population, not least the confederation of Pashtuns who came to support the Taliban. Adopting such a stance ignored the fact that globally, the most successful peace agreements are those that allow for insurgent participation [12, 13] and made Afghanistan’s institutions fundamentally exclusionary. The only avenue to gain voice for the Taliban and their supporters was to sustain violent conflict, seek support from Pakistan, and push for more favorable political terms or the outright victory they ultimately achieved.
Second, political parties could not develop because of the decision to create large voting districts along with single non-transferable votes. This system is basically not used anywhere else in the world precisely because it drives political division, as We detail below. Restrictions on listing party affiliation on the ballot also did not help. The electoral system, in this sense, was fundamentally divisive and created a winner-take-all system. Predictably, this led to a patronage-oriented politics that entrenched existing elites  and excluded groups that shared pro-growth agendas, such as the growing urban middle class or new business elites, from becoming a viable political force.
Finally, the constitution gave vast powers to the executive, such as appointing all provincial and district governors. The highly centralized system is as far from the de facto federal nature of Afghan tribal authority as can be imagined.
Several experts have already pointed out that these design flaws encumbered Afghanistan’s political development [15, 16, 17, 18]. Our observations are not novel. We argue, however, that it is important to consider these design features together. They created an exclusionary, divisive, and centralized set of formal institutions that were fundamentally disconnected from Afghanistan’s political realities.
This, in turn, created two downstream issues for Afghanistan’s political and economic development. First, there was limited room for healthy politics and for effective state-building. Elites excluded from power opposed the state and sabotaged its development, both from within and from without. The crescendo of violence from the Taliban, the fact that the outcome of every election from 2009 onward remained deeply contested, and a series of failed power-sharing arrangements, all provide evidence that many of Afghanistan’s elite never bought into the mission of the state. Afghanistan also suffered frequent opposition from within, with Ministers refusing to pursue the agenda of the President, ministries frequently working at cross-purposes, and a broad range of actors deciding to loot rather than to build the state. Unprecedentedly large flows of foreign assistance exacerbated these issues, creating the widely discussed focus on capturing rather than building the Afghan state.
Second, because this dysfunction was woven into the country’s institutional fabric, international actors – even when they worked to improve the situation in Afghanistan – could only work at the margins. International forces were restricted to winning hearts and minds through the provision of local development projects [5, 6, 19] or revising tactics to minimize civilian casualties . While these pursuits were necessary for the US-led coalition to succeed in Afghanistan, they could not be sufficient. Matters as fundamental as institutional reform and bringing in the Taliban were basically off the table until it was too late.
State building, at its core, requires identifying a domain in which there is sufficient agreement on the core mission as a precondition for building state capacities . The US did not, and perhaps could not, create a coherent long-term strategy focused around this mission. Instead, it spent colossal sums trying to fix problems at the margins. H.R. McMaster, whose involvement in the Afghan war culminated in serving as U.S. National Security Advisor, famously argued that Afghanistan was not a 20-year war but rather a one-year war fought 20 times over. Much of those 20 years was spent working on problems at the fringes, and not at the core, of Afghanistan’s political issues.
Many of the fundamental issues that would plague the US-created Afghan government began at the Bonn Conference. As Surhke  describes, when the architects of the Afghan state convened in Bonn on November 27, 2001, they created a system with no room for the Taliban and one that naturally led to a highly centralized presidential system. This was for at least three reasons.
First, the goal of the conference was not to create a viable long term political solution. It was to quickly create a palatable successor regime to the Taliban. Then–Secretary of State Colin Powell and the U.S. military, wanted terms for an interim governing arrangement before the U.S.-backed Northern Alliance militias captured Kabul (in part, to avoid the bloody settling of ethnic vendettas). The much longer project of building consensus around a set of institutions – a hallmark of successful constitutional processes  – was simply incompatible with US timelines. This was left to the future.
The UN, which was put in charge of the Bonn negotiations to provide an international stamp of approval, allowed only four narrow political factions to be represented. It created an iterative structure that included a timeline for progressively wider elections and the eventual 2004 constitution. Much of the hard work of negotiating a peace process was left to the future, but was made impossible by the fact that elites had control from the outset . Hamid Karzai guided the country toward a highly centralized state with an electoral system that de facto prevented the emergence of alternative coalitions or parties [15, 16] as we describe below.
Second, the US severely underestimated the Taliban’s degree of grassroots support, its importance to the Pakistani military, and therefore its potential to reconstitute itself. Had it appreciated this, a much better option would have been to ensure that the Pashtun confederations that had always supported the Taliban were genuinely bought into Afghanistan’s political institutions.
Third, in 2001, there was little debate that western liberal democracy would inevitably be the preeminent model of political organization [23, 24, 25]. Even before 9/11, the neoconservative movement took this to its most extreme – if markets and politics globally should be fashioned in America’s image, then why not intervene to accelerate that process wherever possible? Military intervention was added to the set of instruments acceptable to neoconservatives to propagate the western liberal model after 9/11.
Confidence in the US’s ability to quickly build democracies, of course, proved misguided. In the immediate aftermath of 9/11, the existing literature on the fundamentally political nature of counterinsurgency and civil war, such as Galula  and Popkin  was largely forgotten. It would not be rediscovered and used to guide military policy until the publication of the US Army Field Manual on counterinsurgency . And the remarkable political economy literature on conflict, governance, and development was created largely in response to the need to understand state formation and provide solutions for policymakers engaged in the colossal undertaking of building modern states in poor war-torn countries .
It was challenging to forecast how long or costly engagement in Afghanistan would be, or even to think about how to pursue this effectively. Correspondingly, there was limited appreciation of the political complexity of insurgent conflict and of the potential to be drawn into a quagmire. The example of Iraq is instructive. In October 2002, Nobel Laureate William D. Nordhaus produced one of the only independent and professional attempts to forecast the costs of a potential invasion in Iraq. He provided two estimates. If the occupation was short and favorable the war would cost $121 billion. If it was prolonged and unfavorable it would cost $1.595 trillion. Nordhaus’s estimates vastly exceeded the official estimates from the US government and were viewed as outlandish, even though he pointed to the frequent failures to estimate the eventual costs of wars, including the Vietnam War, which cost 11 to 15 times the original estimate. In practice, Iraq greatly exceeded Nordhaus’s maximum estimate. The most recent estimates place the costs of the Iraq war at just over $2 trillion, before including future veterans’ care.
Several commentators have pointed to the exclusion of the Taliban from politics as the ‘original sin’ in the Afghan war . However, there were a set of related issues that also shackled political and economic development. This section considers a broader set of three original sins.
The exclusion of the Taliban from the original negotiations at the Bonn Conference was the original, most damning sin. Successful peace arrangements often include provisions that allow all parties to participate [12, 21]. For example, of the 110 conflicts that were settled between 1975 and 2005, 33 of the 42 that permitted rebel participation endured for five years, while only 30 of the remaining 68 survived that long . While this research came of age after 2001, it is not altogether surprising that if large groups or powerful actors are entirely excluded from a political system, they will resort to violence to force their way in.
Early in the war, conceding a role for the Taliban was anathema to America, who had the veto power to block their inclusion. George W. Bush conflated Al Qaeda and the Taliban in a speech on the evening of 9/11 , and pursuing terrorists with the full military might of the US was the order of the day. The authors had a particularly memorable conversation in Afghanistan, in 2009, with a US Army Colonel who was at Central Command in Orlando during the planning phases for the Afghan invasion. He described a stressful round-the-clock planning process that involved running through nightmare scenarios like Al Qaeda obtaining a nuclear weapon from the Pakistani arsenal and the escalation of a broader war that might draw in Pakistan. Tellingly, the intellectual exercises of either not invading at all or of only running a limited ‘over-the-horizon’ counter-terrorism mission aimed at killing Osama bin Laden were apparently not discussed at Central Command in October 2001. Not invading was not an acceptable option. Nor, as Lakhdar Brahimi, the lead UN negotiator in charge of the Bonn proceedings explained, was involving the Taliban in any successor regime. Signaling that an attack on the American homeland would carry major and lasting consequences even for the Taliban – who were only tangentially involved in the 9/11 attacks, but did refuse to give up bin Laden – was the paramount consideration.
The Afghan constitution enshrined an electoral system that combined multi-member districts and a single non-transferable vote (SNTV). This system is exceedingly rare and used by almost no successful democracies . It is only used, or only has been used in, Jordan, the Pitcairn Islands, Vanuatu, Japan from 1948 to 1993 (with the important caveat that constituencies were limiting to having very few representatives), and Taiwan from the 1960s to the 1990s. Both Japan and Taiwan abandoned the problematic system because it led to factionalism and created incentives for patronage.
Afghanistan only holds elections for the president and for the lower house (Wolesi Jirga). The Wolesi Jirga comprises 250 seats spread across 34 constituencies (provinces). When Afghanistan decided on its electoral system, the interim government felt that any constituency other than the 34 provinces would be unacceptable (although the country does have 421 districts contained within those 34 provinces that could plausibly constitute single-member electoral units).
The problem with the combination of multi-member districts and SNTV are carefully described by Reynolds and Carey in two pieces [15, 16]. At its most basic, members from the same party or political alliance are forced to run against one another. Moreover, if a candidate receives more votes than needed to enter office, they cannot transfer these excess votes to their allies. And so, they do not form coalitions. In contrast, other multi-member systems that allow pooling within party lists avoid this issue. Under an SNTV system with large multi-member districts, unless a political alliance perfectly anticipates its voter support, and nominates a number of candidates in line with that support, and controls its voters such that it distributes support across them evenly, support for the alliance will not translate into votes for that alliance. It is easy to devise scenarios where parties can receive a substantial majority of votes and still receive a minority of seats. Therefore, there is very limited incentive for political coordination. It is every candidate for themselves.
This system led to a number of costly outcomes. First, it made it virtually impossible for parties or other political coalitions to emerge around a shared political agenda. Consequently, the same tribal groups that fought during the bloodiest period of Afghanistan’s constant 40-year-long episode of instability had no incentives to form broader political coalitions. Nor could new political actors, such as new business-oriented urban elites, easily create new pathways to political power.
Second, the system created incredible incentives for election fraud, which is documented in our study of the 2010 Wolesi Jirga election reported in Callen and Long . There we precisely measure how much specific candidates were able to inflate their vote totals during the aggregation process, and found consistently that a small set of powerful candidates engaged in dramatic vote inflation.
Widespread fraud not only undermines the key role that elections play in both allowing voters to select competent politicians and in providing performance incentives to incumbents who know they will someday face re-election . It also erodes the social contract. In Berman, Callen, Gibson, and Long , we find clear evidence that reducing election fraud causally increased popular support for the Afghan government. It also increased citizens’ willingness to cooperate with the state in basic and fundamental ways, such as being willing to pay taxes or providing critical intelligence regarding anti-state actors to state forces.
The Taliban understood the vital nature of free-and-fair elections with broad-based participation for legitimizing the state. On election days, the Taliban committed about ten times as many attacks as they would on a normal day. Moreover, a remarkable study using fine-grained data on attacks and travel routes to polling centers shows clearly that the Taliban sought to disrupt voting, by, for example, attacking travel routes while simultaneously minimizing civilian casualties by attacking in the morning .
Reflecting the fading legitimacy of the Afghan state and disaffection with the system, turnout in Afghan elections dropped successively and very severely, from 9,716,413 voters (83.66% turnout) in 2004 to 1,823,948 voters (18.87% turnout) in 2019.
Third, the system is incredibly complex for voters. Ballot papers were often several pages long and included the names of hundreds of candidates.
Fourth, voters share only a very broad geography with all of their elected representatives. Afghanistan’s provinces are both large and incredibly diverse. Allowing candidates to run at-large in a province almost ensures that some regions and groups will have no elected political representation.
Indeed, the decision to have large district magnitude, SNTV, and electoral rules prohibiting party affiliation on the ballot were implemented precisely to prevent the formation of parties and to preserve the power of the executive. While the UN argued for a proportional representation system, through a series of machinations, Hamid Karzai controlled the entire process leading up to the constitutional convention in 2004, and it was quickly ratified. The first draft was devised by a nine-member committee appointed by Karzai between October and March 2003, and from April to December 2003 a further 35-member all-Afghan constitutional commission selected by Karzai finalized the draft. They presented it to the Loya Jirga in December 2003. The highly controversial constitution did not specify the electoral system, though intimated it should be some form of list proportional representation (PR).
The precise details of the system were to be worked out by the Afghan government in cooperation with the Joint Election Management Body (JEMB) and the UN Assistance Mission in Afghanistan (UNAMA). They agreed to a closed-list PR system using multi-member districts based on Afghanistan’s 34 historic provinces.
However, Karzai tasked a young assistant with making the case for closed-list PR to his cabinet. The assistant did not understand the system, its logic, and did not make a compelling case . This, combined with popular distrust of political parties due to the chaotic nature of multi-party politics in the 1960s and the subsequent Communist Party rule and Soviet occupation (1978–89), and a belief that creating single member districts from Afghanistan’s traditional 34 provinces was not logistically or politically feasible, led to the eventual adoption of SNTV with multi-member districts.
Afghanistan is highly polarized: while the country has never had a census, around 40% of its population are Pashtuns, 30% are Tajik, 10% are Hazara and 10% are Uzbeks. These groups are also religiously divided. The Pashtuns and Uzbeks are predominantly Sunni, while the Tajiks and Hazara are predominantly Shia. Historically, Pashtuns have dominated politically but Pashtun regimes in Kabul have been forced into de facto federal arrangements because of the large non-Pashtun populations in Afghanistan’s north and west. Correspondingly, prominent Tajik leaders, as well as political scientists working on Afghanistan, advocated for a federal system.
There are several logics regarding why exclusionary, divisive, and centralized winner-take-all political systems, in ethnically, religiously, and culturally divided societies become extractive and do not develop politically or economically [32, 33]. The characterization from Padró-i-Miquel  describes Afghanistan well. In the presence of entrenched social cleavages, especially when succession protocols are weak (like highly controversial elections), rulers can gain the support of a sizeable share of the population even while pursuing policies focused on personal enrichment. The rationale is that citizens will have a preference for rulers from their own group, even if they are corrupt, because they will be better off than they would be if a similarly ineffective and venal ruler from another group took power. In such an equilibrium, political order and continued opportunities for the ruler to enrich themselves, are maintained through in-group patronage, rather than through pursuing inclusive and effective reforms.
While it is clear that the design of Afghanistan’s institutions severely undermined the country’s chances, Afghanistan will always be beset by major obstacles. Any account of why Afghanistan failed is incomplete without acknowledging these.
First, the challenging reforms required to create stable and inclusive political institutions requires complete sovereignty. So long as the dispute between India and Pakistan persists, this is all but impossible. Pakistan simply will not risk the possibility of a stable and autonomous ruling regime in Afghanistan that could some day refuse to support Pakistan in its dispute with India.
This is because Pakistan’s powerful military believes it needs ‘strategic depth’ in case of an Indian land invasion. It also trains proxy terrorists groups in Afghanistan, which it views as key to its military strategy against India. Reportedly, when George W. Bush offered Pervez Musharraf, then Pakistan’s prime minister, a large aid package to cut ties with the Taliban, a large number of senior military officers resigned in protest.
Second, many of those involved in the creation of Afghanistan’s institutions in the run up to 2004 criticized the US decision to subsequently invade Iraq. The US took its best and brightest and focused them on what they thought was a more important objective: replacing Saddam Hussein and building a democratic ally in the Middle East. The shift in focus by the US State Department and Department of Defense came at a critical moment, when the Afghan constitution was on its way to ratification. Perhaps this is why the US allowed Karzai to essentially dictate the terms and choose a set of institutions that did not acknowledge existing power dynamics and that entrenched existing divisions.
Third, there is no shortage of foreign benefactors in Afghanistan’s neighborhood who see no issue with autocratic governments and who are not especially concerned with the welfare of Afghans. Beyond Pakistan, which seeks a sympathetic regime above all else, Russia and especially China, which is allied with Pakistan on the issue, see benefit in seeing the US humiliated. China also seeks free access to Afghanistan’s mineral wealth. Afghanistan’s collapse was a major geopolitical victory for Russia and for China. It also underscored the case that autocrats are making the world over: that Western liberal ideas are antiquated and should no longer be viewed as the objective of political development.
Last, if indeed it is true that no democracy in Afghanistan can exist that does not possess two features: (1) maintaining its 34 districts as the fundamental political unit and; (2) letting voters vote for a candidate and not a party, then the SNTV system with multi-member districts may be the only option. If this is so, given that the system is both conceptually and empirically known to be highly ineffective, this would argue that Afghanistan is exceptionally unsuited to democracy. However, these two requirements seem artificial.
The American quagmire in Afghanistan carried tremendous costs, most especially for Afghans who saw their long struggle for a brighter future crushed by a brutal theocracy. From 2000 to 2019, GDP per capita increased from around 320 to around 555 US dollars (in constant 2015 dollars). Male primary school enrolment increased from 40% to being near total, and female primary enrolment increased from 0% to 90%. In 2022, the country stands on the brink of collapse.
Given the degree of investment and the implications for human welfare, it is deeply important to acknowledge the reality that one size does not fit all countries. Despite this, there was little consideration as to how the design of Afghanistan’s political institutions and constitution should reflect its unique character. We contend that three design choices caused the country’s institutions to be fundamentally out of sync with its political realities. We emphasize that, because these were choices, it is wrong to accept that there is no solution that could have possibly worked to create democracy in Afghanistan.
This article benefited from conversations with colleagues and collaborators working in and interested in Afghanistan including many devoted civil servants in the Government of the Islamic Republic of Afghanistan and the International Security Assistance Force. We thank Eli Berman, Tim Besley, Elizabeth Callen, Ali Cheema, Mick Cox, Joseph Felter, Ruixue Jia, James D. Long, Jacob N. Shapiro, Mohammad Isaqzadeh, Nargis Nehan, Scott Guggenheim, Tarek Ghani, Gerard Padró-i-Miquel, Rohini Pande, and Noam Yuchtman.
The authors have no competing interests to declare.
Brown University Watson Institute International & Public Affairs: Costs of War [Internet]; [posted 2022; cited 2022 Mar 9]. Available from: https://watson.brown.edu/costsofwar/.
UN World Food Programme: Half of Afghanistan’s population face acute hunger as humanitarian needs grow to record levels [Internet]; [posted 2021 October 25; cited 2022 March 9]. Available from: https://www.wfp.org/news/half-afghanistans-population-face-acute-hunger-humanitarian-needs-grow-record-levels.
Besley T, Persson T. The logic of political violence. The Quarterly Journal of Economics. 2011; 126(3): 1411–1445. DOI: https://doi.org/10.1093/qje/qjr025
Berman E, Shapiro JN, Felter JH. Can hearts and minds be bought? The economics of counterinsurgency in Iraq. Journal of Political Economy. 2011; 119(4): 766–819. DOI: https://doi.org/10.1086/661983
Berman E, Callen M, Felter JH, Shapiro JN. Do working men rebel? Insurgency and unemployment in Afghanistan, Iraq, and the Philippines. Journal of Conflict Resolution. 2011; 55(4): 496–528. DOI: https://doi.org/10.1177/0022002710393920
Weinstein JM. Inside rebellion. Cambridge: Cambridge University Press; 2006. DOI: https://doi.org/10.1017/CBO9780511808654
Popkin SL. The rational peasant. Oakland: University of California Press Books; 1979. DOI: https://doi.org/10.1525/9780520341623
Besley T, Persson T. State Capacity, conflict, and development. Econometrica. 2010; 78(1): 1–34. DOI: https://doi.org/10.3982/ECTA8073
Matanock AM. Bullets for ballots: Electoral participation provisions and enduring peace after civil conflict. International Security. 2017; 41(4): 93–132. DOI: https://doi.org/10.1162/ISEC_a_00275
Matanock AM, Staniland P. How and why armed groups participate in elections. Perspectives on Politics. 2018; 16(3): 710–727. DOI: https://doi.org/10.1017/S1537592718001019
Callen M, Long JD. Institutional corruption and election fraud: Evidence from a field experiment in Afghanistan. The American Economic Review. 2015; 105(1): 354–381. DOI: https://doi.org/10.1257/aer.20120427
Carey JM, Reynolds A. The U.S. helped design Afghanistan’s constitution. It was built to fail. The Washington Post. [Internet]; 2021 September 8 [cited 2022 March 9]. Available from: https://www.washingtonpost.com/outlook/2021/09/08/afghanistan-constitution-failure/.
Dell M, Querubin P. Nation building through foreign intervention: Evidence from discontinuities in military strategies. The Quarterly Journal of Economics. 2018; 133(2): 701–764. DOI: https://doi.org/10.1093/qje/qjx037
Shapiro JN, Condra LN. Who takes the blame? The strategic effects of collateral damage. American Journal of Political Science. 2012; 56(1): 167–187. DOI: https://doi.org/10.1111/j.1540-5907.2011.00542.x
Commission on State Fragility, Growth and Development. Escaping the Fragility Trap [Internet]; [posted 2018 April; cited 2022 March 9]. Available from: https://www.theigc.org/wp-content/uploads/2020/10/Escaping-the-fragility-trap_Oct-2020.pdf.
Horowitz DL. Constitutional process and democratic commitment. New Haven: Yale University Press; 2021. DOI: https://doi.org/10.12987/yale/9780300254365.001.0001
Huntington SP. Democracy’s third wave. Journal of Democracy. 1991; 2(2): 12–34. DOI: https://doi.org/10.1353/jod.1991.0016
Fukuyama F. The end of history? The National Interest. 1989; 16: 3–18. DOI: https://doi.org/10.1080/07366988909450540
US Army Field Manual 3–24 Counterinsurgency Available at: https://armypubs.army.mil/epubs/DR_pubs/DR_a/pdf/web/fm3_24.pdf.
Berman E, Felter JH, Shapiro JN. Small wars, big data: The information revolution in modern conflict. Princeton: Princeton University Press; 2018. DOI: https://doi.org/10.23943/9781400890118
Berman E, Callen MJ, Gibson C, Long JD. Election fairness and government legitimacy in Afghanistan. Journal of Economic Behavior and Organization. 2019; 168: 292–317. DOI: https://doi.org/10.1016/j.jebo.2019.10.011
Condra LN, Long JD, Shaver AC, Wright AL. The logic of insurgent electoral violence. American Economic Review. 2018; 108(11): 3199–3231. DOI: https://doi.org/10.1257/aer.20170416
Acemoglu D, Robinson JA. Why nations fail: The origins of power, prosperity, and poverty. New York: Crown Publishers; 2012. DOI: https://doi.org/10.1355/ae29-2j
Acemoglu D, Johnson S, Robinson JA. The colonial origins of comparative development: An empirical investigation. American Economic Review. 2001; 91(5): 1369–1401. DOI: https://doi.org/10.1257/aer.91.5.1369
Padró-i-Miquel G. The control of politicians in divided societies: The politics of fear. Review of Economic Studies. 74(4): 1259–1274. DOI: https://doi.org/10.1111/j.1467-937X.2007.00455.x