Building state capacity in fragile states: Evidence from Afghanistan

Poverty and state fragility are increasingly connected: while fragile states comprise just 24% of the global population, they account for 73% of those living in extreme poverty—a figure projected to rise to 86% by 2030 (OECD 2022). It is clear that fragility undermines the development of the types of effective states that enable economic growth. Yet, why fragility hinders state development remains unclear. Our research examines one potential explanation: that extreme fragility, because it implies deep uncertainty about future “rules of the game”, discourages investments in state capacity. Risks of war, state collapse, and institutional upheaval (Fearon and Laitin 2003, Collier 2008, Acemoglu and Robinson 2012) may discourage costly and complex investments in state capacity, such as systems to effectively compensate civil servants, which yield benefits only if the state endures (Besley and Persson 2010, North and Thomas 1973).

Building such administrative capacity – starting from effectively identifying and paying civil servants – is crucial for fragile states transitioning from informal systems to enduring and structured organisations (Besley and Persson 2009, Cantoni et al. 2024). However, this task is especially daunting for states trapped in “fragility traps”, where such investments must be made not only in contested regions but also under the constant threat of collapse. We examine a flagship initiative by the Afghan government to make such an investment. The Mobile Salary Payment (MSP) reform aimed to build basic administrative capacity by leveraging digital technology to modernise the identification and payment of civil servants. This reform began in 2017 and continued until the Taliban's takeover.

Afghanistan needed major administrative capacity building

Decades of instability, starting with the Soviet invasion in 1979, left Afghanistan among the world’s weakest states, struggling to manage basic administration. Even with extensive international support from 2001 to 2021, substantial challenges persisted, including delays in salaries, leakage, and widespread “ghost workers”— fake employees on the payroll but not actually working for the state. Indeed, recent post-mortems on the Afghan government point to such phenomena as being at the core of the Afghan state’s eventual failure (Sopko 2025).

The education sector highlighted these challenges starkly. Our baseline audit found only 41% of teachers present at schools; just 30% had a government ID card, and 25% could not produce an ID number. And salary delays affected 60% of employees, with an average unpaid period of 3.5 months. Reflecting these issues, Afghanistan ranked 150th out of 176 countries for learning outcomes in 2020 (Angrist et al. 2021). These inefficiencies underscored the need to develop the basic administrative systems that underpin the state.

The Mobile Salary Payment Reform in Afghanistan

To attempt to increase administrative capacity, The Ministry of Education launched the Mobile Salary Payment (MSP) reform. The ambitious reform aimed to use biometric registries and mobile payments, partly inspired by the success of such digital systems in other developing countries (Muralidharan et al. 2016). But in Afghanistan, like other fragile states, such foundational reforms face the basic challenge that the state faces a high probability of either radically changing or completely collapsing before the reform yields fruit. This was starkly demonstrated when, just four years into the reform, the Taliban overthrew the government, dismantled the reform's systems, and restricted public education, especially for women. Such conditions facilitate examining how beliefs in state durability impact the success of state capacity-building efforts.

During the experiment, the Taliban posed an existential threat to the Afghan state, creating deep uncertainty about Afghanistan's future. Would it remain a flawed democracy or transform into the theocratic autocracy seen today? This period exemplifies a critical juncture, defined as moments of profound uncertainty about future institutions (Callen et al. 2024).1 Survey data from May 2018, at the onset of the MSP reform, highlight this uncertainty: 47.7% of respondents believed the National Army would defeat the Taliban, 19.5% anticipated a Taliban victory, and 31.7% were uncertain. This data, combined with our experiment, provide an opportunity to explore how confidence in the state's durability impacts investments in state capacity.

In this context, our study experimentally studies the MSP reform to provide evidence on two questions. First, can new technologies build capacity to register and pay employees and improve service delivery in fragile states? Second, do beliefs about institutional durability - and uncertainty about the same - shape such investments in state capacity? We view these questions as fundamental to making progress on the reasons why fragility stands in the way of political and economic development. 

Can digital systems build administrative capacity in fragile states?

The MSP reform aimed to improve payroll integrity and payment processes for teachers in three stages. First, a Mobile Network Operator registered teachers, collecting biometric data and national IDs to create mobile-money-enabled SIM cards. This step audited payrolls to identify discrepancies and verify actual employees. We refer to this step as “Registration”. Second, the government matched registration data with payroll records, enabling the Ministry of Education to decide on salary continuance and remove ghost workers. Third, verified employees received salaries via mobile wallets, requiring biometric authentication for cash withdrawals. We refer to this step as “Payment”.

We evaluate the MSP reform through a randomised controlled trial that ran from May 2018 to May 2020. The study included approximately 30,000 employees across more than 1,000 schools in  Kandahar, Nangarhar, and Parwan. Roughly a third of the rural districts we study were contested variously by the Taliban, the Islamic State, and other violent actors. We grouped the schools into registration zones and randomly assigned zones to one of three treatment groups by varying the timing of the “Registration” process and the “Payment” implementation.

Impact of civil service reform in Afghanistan

First, biometric registration streamlined payrolls and employee records, but they are not a cure-all and would require additional verification. Although our most conservative estimates indicate that 97.2% of employees successfully registered biometrically, further investigation revealed that some of these individuals were likely people who were not actually hired by the MoE but registered biometrically to receive salary payments. Indeed, by administering a forensic “litmus test” survey to assess employees’ knowledge of basic information that any active employee should know and by comparing the responses of true employees with the responses of other biometrically registered individuals, we estimate that 8.3% to 14.1% of registered individuals were likely ghost workers—critical statistics previously unknown to donors funding salaries. This exercise made employees legible to the state, improving state oversight.

Second, the MSP reform reduced salary delays from 59% to 35% across all employees, driven by significant improvements in first-mile delays—those delays occurring during the central salary approval process in Kabul. These first-mile delays fell substantially in treatment zones after one year. However, last-mile delays (those due to local bank branches and mobile money agents) initially rose due to insufficient cash-out agents, but normalised after two years. In urban areas (where the government had uncontested control and the MNO had more agents), treated employees reported better payment experiences and a reduction in any delays by 24.9 percentage points (from a base of 34.5% in the urban control group). By contrast, in rural areas, treatment still elevated delays by 19.5 percentage points (from a base of 19.8% in the rural control group), even after two years. Despite this divergence in salary experience, after two years, a majority of employees in both urban and rural areas supported expanding the reform nationally.

Third, the MSP reform improved state performance by improving employee retention rates and student learning. High turnover is costly, particularly for educational institutions, where losing skilled teachers has been shown to undermine educational quality (Zeitlin 2021, Akhtari et al. 2022). The reform reduced exits among new MoE employees by 27.5 percentage points from an initial first-year turnover rate of 55% and cut new hiring by 4 percentage points compared to an 18% control group baseline.

At the same time, in an assessment that we conducted with students, learning scores improved by 0.147 standard deviations in treated schools, driven by gains in both maths and reading across the sample.

Fourth, the MSP reform fostered financial inclusion. Peer-to-peer mobile money transfers increased by 27.2 percentage points relative to 6.7% usage in the control group, and employees increased their mobile money balances by 0.4 percentage points more of their salary monthly for every month on the platform, totaling 2.8 percentage points after seven months.

Finally, the success of the MSP reform was linked to citizens' confidence in state durability. Teacher registration for the biometric system increased by 0.994 percentage points for every 1 percentage point rise in citizens' belief that the Taliban would lose. This relationship was strongest in districts where consensus on the Taliban’s defeat was high, aligning with theories connecting beliefs about future stability to engagement with the formal state (Callen et al. 2024). This is well illustrated in the graph below, which shows the interaction between “the marginal effect of people's confidence in a Taliban defeat on employee registration” (y-axis) and “how divided their district is about who will win the war” (x-axis). The graph shows that as districts become more divided on who will win the war (higher x-axis values), the relationship between biometric teacher registration and citizens’ belief in a Taliban defeat weakens.

Figure 1: The relationship between confidence and registration by the degree of consensus

Notes: This figure depicts the relationship between the share of citizens in a district who believe the Taliban would lose the war in 2018 — the outset of the MSP reform — and the share of eligible teachers who appeared to give their biometric data in person (y-axis) against the degree of disagreement in who will win the war in the district (x-axis).

Policy lessons on state capacity in fragile contexts

The Mobile Salary Payment (MSP) reform was an ambitious effort to help build administrative capacity in one of the world’s most fragile states. Our analysis of this exercise brings forth two main lessons regarding what helps to build capacity and to improve state performance, and, more conceptually, why fragility can stand in the way.

First, biometric registration and digital payment systems can enhance administrative capacity, even in fragile settings. The reforms helped identify employees, reduce payroll delays, cut turnover, and improve student learning, while also fostering financial inclusion through mobile money adoption. By the end of the reforms, most employees supported scaling it nation-wide. However, these systems are not a complete solution. Biometrics could not fully eliminate ghost workers, and digital payments relied heavily on efficient third-party providers like mobile network operators. As with Afghanistan’s broader state-building efforts, the reform’s progress came only after significant implementation challenges and ultimately collapsed with the Afghan state in 2021. These results show that while technology is not a quick fix, it can enhance administrative capacity and state performance when given time and applied in areas under stable state control.

Second, the MSP reform revealed the critical role of local beliefs about institutional durability—citizens’ confidence in the government’s survival. While technology provided the tools to improve state performance, the reform’s effectiveness varied significantly based on whether citizens believed in the state’s future. It was most effective in districts where citizens believed the Afghan Army would prevail against the Taliban, especially where this belief was widely shared. This finding underscores that escaping the fragility trap requires creating consensus and confidence in the state’s endurance. When citizens doubt the future of their government, they are less likely to invest in or cooperate with reforms. This insight is actionable: governments in fragile states must consider public confidence in the state’s survival when designing capacity-building initiatives. It also highlights the importance of measuring such beliefs as a critical input for policy.

References

Acemoglu D, and J A Robinson (2012), Why Nations Fail: The Origins of Power, Prosperity, and Poverty, Crown.

Akhtari M, D Moreira, and L Trucco (2022), "Political turnover, bureaucratic turnover, and the quality of public services," American Economic Review 112(2): 442–449.

Angrist N, S Djankov, P K Goldberg, and H A Patrinos (2021), "Measuring human capital using global learning data," Nature 592(7854): 403–408.

Besley T, and T Persson (2009), "The origins of state capacity: Property rights, taxation, and politics," American Economic Review 99(4): 1218–1244.

Besley T, and T Persson (2010), "State capacity, conflict, and development," Econometrica 78(1): 1–34.

Blumenstock, J, M Callen, A Faikina, S Fiorin and T Ghani (2024), “Strengthening Fragile States: Evidence from Mobile Salary Payments in Afghanistan”, Working Paper.

Callen M, J L Weigel, and N Yuchtman (2024), "Experiments about institutions," Annual Review of Economics.

Cantoni D, C Mohr, and M Weigand (2024), "The rise of fiscal capacity: Administration and state consolidation in the Holy Roman Empire," Econometrica.

Collier P (2008), The Bottom Billion: Why the Poorest Countries Are Failing and What Can Be Done About It, Oxford University Press.

Fearon J D, and D D Laitin (2003), "Ethnicity, insurgency, and civil war," American Political Science Review 97(1): 75–90.

Muralidharan K, P Niehaus, and S Sukhtankar (2016), "Building state capacity: Evidence from biometric smartcards in India," American Economic Review 106(10): 2895–2929.

North D C, and R P Thomas (1973), The Rise of the Western World: A New Economic History, Cambridge University Press.

OECD (2022), States of Fragility 2022, OECD Publishing, Paris. https://doi.org/10.1787/c7fedf5e-en.

Sopko J F (2025), "Afghanistan's reconstruction: An audit of U.S. efforts," The New York Times, January 2. https://www.nytimes.com/2025/01/02/opinion/afghanistan-audit-reconstruction-us.html.

Zeitlin A (2021), "Teacher turnover in Rwanda," Journal of African Economies 30(1): 81–102.