"Caught in a Trap: Simulating the Economic Consequences of
Internal Armed Conflict" with Hannes Mueller
[pdf]
While a year of internal armed conflict has modest effects on economic growth, conflict trap dynamics determine how these effects amplify over time. This paper studies the role of these dynamics in shaping the long-run economic costs of conflict. We develop a Markov model of the conflict trap, estimate it using cross-country panel data, and simulate the distribution of conflict trajectories and their associated growth paths. This allows us to study how growth effects compound with conflict trap dynamics. We can also analyze growth conditional on specific conflict histories. Our results show that conflict dynamics map modest growth effects into large long-run losses and account for substantial variation in outcomes. While a conflict year reduces GDP per capita growth by about 3 percentage points, the mean long-run loss is 17 percent. Losses are about 8 percent in countries we classify as peace-prone and rise to 31 percent in conflict-prone countries. Full recovery after conflict is rare, and more persistent conflict trap dynamics lead to higher growth volatility. Relapse risk is a key factor behind these outcomes. Policies that reduce recurrence can therefore yield disproportionately large benefits, especially in conflict-prone countries.
"A Framework for Fragility: Guiding Measurement for Policy" with Laura Mayoral, Hannes Mueller, and Christopher Rauh
[pdf]
[WIDER WP].
Fragility is a term widely used in research and policy, yet its conceptualization remains contested even within fields. We develop a general framework for understanding and measuring fragility. It applies across fields and is oriented toward policy use. The framework provides practitioners with a common structure for thinking about fragility, a set of candidate measures, and guidance for constructing them. Fragility is defined as the ease with which an object fails. We distinguish it from risk, resilience, and vulnerability and identify five types of measures: critical stress, failure condition, conditional probability, unconditional probability, and composite indices. We examine the use of these measures and identify the contexts in which they are most appropriate. Together, they imply a trade-off between causal interpretability—the ability to explain how failure occurs—and operational feasibility—the ease with which a measure can be constructed with reasonable validity. We then discuss the use of fragility measures in policy and develop a protocol for constructing fragility measures, which we apply to a case study of urban areas exposed to wildfire. Even small differences in failure definitions lead to different fragility measures and policy implications.
"A Probabilistic Measure of State Fragility"
[pdf]
[WIDER WP].
State fragility indices conflate fragility with its consequences, rely on conceptually ambiguous definitions, and aggregate indicators using often arbitrary weights and thresholds. This paper proposes a novel measure that addresses these limitations: the probability of state failure. We define state failure as the inability to perform core state functions and operationalize it through four observable outcomes: loss of territorial control, armed conflict, deterioration in public service delivery, and forced regime change. Using a machine-learning approach that mimics an early-warning system, we estimate failure probabilities for 170 countries between 2005 and 2024. We assess their predictive performance and compare them with five leading fragility indices. The proposed measure is forward-looking and exhibits strong predictive power, substantially outperforming the existing indices. Territorial control loss and conflict are the most predictable dimensions. Political conditions and forced displacement emerge as the main predictors of failure. The framework provides a conceptually clear and empirically useful measure for both academic research and policy analysis.
"Disaster Risk through Investors’ Eyes: a Yield Curve Analysis" [pdf]
This paper develops a model to estimate investors’ perceived probability of disaster from yield curve data. Disasters are extreme events like defaults or interstate wars with significant economic impact. By integrating an asset pricing model with government bond yields from Datastream, I provide daily estimates of the one-year-ahead disaster probability as perceived by investors for approximately 60 countries from 2000 to 2023. The use of yield curve data offers a high-frequency measure of disaster risk that can rapidly incorporate new information. Several facts indicate that the estimated probabilities have predictive value. Probabilities spike before disaster events, such as the debt restructurings of Greece, Sri Lanka, and Ghana, and the onset of the Russia-Ukraine war. They are also strongly associated with higher-risk credit ratings. In a forecasting exercise using machine learning, the estimated disaster probabilities enhance the predictive power of credit ratings. This demonstrates the informational value of bond market data in predicting events.
"Coordination across Tax Havens: A Global Games Approach"
[pdf].
This paper develops a coordination game to study tax evasion across multiple tax havens. I extend the global games framework—a class of coordination games with incomplete information—by introducing multiple tax havens in which investors coordinate. This allows for strategic interaction across jurisdictions, capturing cross-haven coordination effects. I analyze how policies that raise the cost of being a tax haven affect overall evasion. Targeting a single haven can backfire by concentrating evaders elsewhere, increasing overall evasion. In contrast, uniform interventions across jurisdictions are more effective. This mechanism helps explain the limited success of past international efforts and highlights the value of a harmonized approach, as intended under the Global Minimum Tax.