Economics of National Security

Economics of National Security

March 5-6, 2017
Martin Feldstein of Harvard University and Eli Berman of the University of California at San Diego, Organizers

Suleiman Abu Bader, Ben Gurion University of the Negev, and Elena I. Ianchovichina, the World Bank

Polarization, Foreign Military Interventions, and Civil Conflicts

In a behavioral model of civil conflict external military interventions alter the resources available to warring groups and their probability of winning. Such a model highlights the importance of distributional measures along with the effect of interventions for conflict incidence. Bader and Ianchovichina test the model empirically and confirm the finding in the literature that ethnic polarization is a robust predictor of civil wars. Furthermore, the researchers find that religious polarization is positively and significantly associated with civil conflict in the presence of non-humanitarian and non-neutral, external military interventions. This result is particularly pronounced in the Middle East and North Africa where religious polarization is found to lead to high-intensity conflicts in the presence of external interventions. The results are robust to different definitions of conflict, model specifications, and data time span.


Giorgio Chiovelli, the University of Bologna; Stelios Michalopoulos, Brown University and NBER; and Elias Papaioannou, London Business School

Land Mines and Spatial Development

Chiovelli, Michalopoulos, and Papaioannou explore the economic consequences of demining for both the affected districts and to the Mozambican economy as a whole. Their analysis proceeds in four steps. First, the researchers describe the self-collected and cross-validated data on the spatial distribution of land mines and unexploded ordnances (UXOs) at the end of the civil war (1992); they then provide fine geo-referenced information on land mine clearance operations. Second, they exploit time variation in the demining process — that appears to be non-coordinated and non-strategically planned — to assess its impact on economic activity, as reflected on satellite images on light density at night. The researchers estimate "difference-in-difference" specifications, that compare the evolution of local development in areas (Mozambican municipalities), where demining took place in a given period, to municipalities that either had no land mines at the end of the civil war or municipalities that were mined but had not been cleared. The analysis reveals small-to-moderate effects of the removal of land mines on local economic development. Third, they examine the economy-wide implications of land mine removals. This is crucial as there are spillovers, general equilibrium effects, from the removal of land mines in one district to others. The researchers' approach combines information on the evolution of land mine clearance with the transportation network (roads, railroads and rivers) at independence. They apply a "market access" approach that quantifies the aggregate effects of land mines (and their subsequent removals) on spatial development. The "market access" estimates reveal an economically large and statistically precise impact of land mines removal on aggregate economic development. Fourth, combining the spatial general equilibrium model of intra-country trade and the econometric estimates, they run counterfactual policy simulation that allows them to estimate the likely gains of demining if it was centrally coordinated and planned targeting the key colonial development corridors and the roads and railroads connecting the three main cities. The analysis reveals large losses from lack of central planning and strategic coordination among the demining operators.


Esteban Klor, Hebrew University; Sebastian M. Saiegh, the University of California at San Diego; and Shanker Satyanath, New York University

Crony Capitalism and the Targeting of Violence: Labor Repression during Argentina's Last Dictatorship

This paper studies whether crony governance affects the logic behind governments' targeting of violence, and how the deployment of violence allows politically connected firms to benefit from crony governance. Klor, Saiegh, and Satyanath address these issues in the context of the Argentine military junta that took power on March 24, 1976. Specifically, they examine the logic driving the choice of firm level union representatives who were subjected to violence following the coup. Using an original dataset assembled and digitized by the researchers, they find that political, business and social connections to the regime are associated with an increase of 2 to 3 times in the number of firm level union representatives arrested and/or disappeared. This is the case even after controlling for a battery of firms' characteristics that capture alternative explanations for the targeting of violence. The effect is particularly pronounced in privately-owned (as opposed to state-owned) firms, suggesting that the correlation is driven by cronyism for financial gain rather than ideology or information transmission. They also show that connected firms benefited from violence against union representatives by subsequently having less strikes and a higher market valuation. The findings highlight the pervasiveness of ties to the government, even in cases where one of the main stated goals of the regime is to curb cronyism.


Kerwin Kofi Charles, the University of Chicago and NBER; Konstantin Kunze, the University of California at Davis; Hani Mansour and Daniel I. Rees, the University of Colorado at Denver; and Bryson Rintala, U.S. Air Force Academy

Taste-Based Discrimination and the Labor Market Outcomes of Arab and Muslim Men in the United States

Between 2001 and 2014, more than 6,500 American soldiers died while serving in Afghanistan and Iraq. Drawing on data from the Defense Manpower Data Center, which contains information on each of these soldier's home state as well as the exact date on which he or she died, Charles, Kunze, Mansour, Rees, and Rintala estimate the relationship between home-state fatalities and the labor market outcomes of first- and second-generation Arab and Muslim men working in the United States. Because home state does not influence when, where, or how the U.S. military deploys its soldiers, news of a soldier's death can be thought of as producing a temporary, state-specific shock to the degree of prejudice faced by Arab and Muslim men working in the United States. The researchers find that home-state fatalities are essentially unrelated to wages and employment status. However, they are negatively related to hours of work, especially among Arab and Muslim men in occupations that require intense interactions with customers or clients. The researchers argue that these results are consistent with customer taste-based discrimination but inconsistent with statistical models of discrimination.


Samuel A. Bazzi, Boston University; Robert Blair, Brown; Christopher Blattman, The rgw University of Chicago and NBER; Oeindrila Dube, the University of Chicago and NBER; Matthew Gudgeon, Boston University; and Richard M. Peck, Jr, Innovations for Poverty Action

What Can Prediction Teach Us About Violence? Machine Learning Applications in Indonesia and Colombia

A large and growing literature aims to identify causal determinants of violent conflict. However, less is known about the predictability of conflict, and much less about what forecasting can teach us about violence. Bazzi, Blair, Blattman, Dube, Gudgeon, and Peck deploy six machine learning methods to forecast conflict in Indonesia with new geospatial data on violent events at both the subdistrict and village level, beginning in 2000. The researchers' preferred model, an ensemble of the other five, improves on prior attempts to predict local violence in other settings. Combining data from numerous administrative and survey-based sources, they find that lagged conflict, demographic characteristics such as the age distribution, education levels and infrastructure, and natural disasters are the most important predictors of future conflict. However, economic shocks associated with commodity price and weather volatility have limited predictive power. In ongoing work (to be incorporated into a new draft in February 2017), the researchers predict guerrilla and paramilitary attacks in Colombia from 1990-2005 using a similarly large municipality-level panel dataset. In both contexts, the exercise delivers several important findings about local-level violence. First, violence is persistent but cannot be characterized by a simple autoregressive process. Second, violence is highly predictable with a small subset of readily observable variables. Finally, the conflict literature may be focused on a narrow set of economic shocks that are causally well-identified but of limited value in a forecasting sense.


Benjamin Crost, the University of Illinois at Urbana-Champaign, and Joseph Felter, Stanford University

Export Crops and Civil Conflict

Many experts consider a move towards high-value export crops, such as fruits and vegetables, as an important opportunity for economic growth and poverty reduction, but little is known about the effects of export crops in fragile and conflict-affected countries. Crost and Felter exploit movements in world market prices combined with geographic variation in crop intensity to show that increases in the value of two major export crops — bananas and sugar — caused increases in conflict violence and insurgent-controlled territory in the Philippines. The results are consistent with a mechanism in which insurgents fund their operations by extorting large agricultural export firms.


Luke N. Condra, the University of Chicago; James D. Long, the University of Washington; Andrew C. Shaver, Princeton University; and Austin L. Wright, the University of Chicago

The Logic of Insurgent Electoral Violence

Competitive elections are essential to establishing the political legitimacy of democratizing regimes. Recognizing the symbolic value of voter participation, Condra, Long, Shaver, and Wright argue armed actors who lack institutional power try to undermine the state's mandate through electoral violence. The researchers study the logic of such violence, focusing on the crucial balance between disrupting elections and minimizing harm to potential supporters of the rebellion. They theorize when and where insurgents attack around elections. To test the observable implications of their theory, the researchers perform a quantitative assessment of insurgent electoral violence in Afghanistan. This assessment incorporates novel geo-referenced data on polling stations, data on the ethnic composition and locations of villages, and newly declassified microdata on insurgent activity. The results show that insurgents significantly increase the intensity of violence on election days during the hours before and as voting centers open, but harm relatively fewer civilians than non-election periods. Insurgents also deploy improvised explosive devices (IEDs) along roads connecting voters and polling stations, but rarely bomb these roads multiple times. The researchers instrument for the timing and spatial distribution of attacks using early morning wind conditions and nighttime cloud cover and show that insurgent electoral violence effectively undermines voter participation. The results provide important insights for safeguarding at-risk elections in emerging democracies.


Madeline Zimmerman, Harvard University

The Effect of U.S. Drone Strikes on Terrorism in Pakistan and Yemen

This paper examines the effect of U.S. Unmanned Aerial Vehicles, or drones, on terrorism in Pakistan and Yemen using the data from the Global Terrorism Database and the data on drone strikes from the Bureau of Investigative Journalism. Zimmerman analyzed the Pakistan data from 2007-2015 for all terror acts in the Federally Administered Tribal Regions (FATA) of Pakistan, and the Yemen data from 2011-2015 for terror acts committed by al-Qaeda in the Arabian Peninsula (AQAP) throughout Yemen's provinces. She found divergent effects for Pakistan and Yemen. In Pakistan, Zimmerman observed a fade-out effect, where drone strikes decreased both the weekly rate of terror attacks and the probability of a terror attack but only within one week of the drone strike. After one week, the probability of a terror attack increased. In Yemen, the researcher found evidence that within the first week of a drone strike, the probability of a terror attack increased. The increase in probability of terrorism was not persistent past the initial days of the strike. These results suggest U.S. drone use should be considered on a country-by-country basis and that the effects of drone strikes, whether they increase or decrease terrorism, do not seem to be longterm.