How do the characteristics and humanitarian impacts of U.S. counterterrorism strikes differ between Somalia and Yemen?

Introduction

Since 2002, the United States has waged a clandestine drone war in countries like Yemen and Somalia, often far from the public eye [1]. While these counterterrorism strikes aim to eliminate militant targets with minimal risk to U.S. forces, their humanitarian impacts remain a source of urgent concern. The cost to civilian lives can be high – for example, an investigation found that roughly one-third of those killed by U.S. drone strikes in Yemen in 2018 were likely civilians or pro-government allies [2].

This research addresses a critical problem: do the patterns and human costs of U.S. strikes differ between Somalia and Yemen, and if so, how? Understanding this is important both theoretically and practically.

Theoretically, comparing two distinct theaters of drone warfare can reveal how local context (from insurgent group dynamics to intelligence quality) influences strike outcomes. Practically, such analysis informs policy by identifying where drone operations are less effective in sparing civilians, guiding improvements to minimize collateral harm.

These strikes are carried out “out of sight,” but their consequences are very real [1]. Official accounts have often underestimated civilian casualties, prompting independent organizations to step in [2]. For instance, the U.S. government once claimed only 64–116 civilian deaths in all drone strikes outside warzones from 2009–2015, whereas independent monitors estimated several times higher [2].

Efforts to document the drone war’s toll have proliferated:
- Pitch Interactive’s Out of Sight, Out of Mind visualization illustrated every CIA drone strike and casualty in Pakistan [1].
- The Economist published infographics showing discrepancies between official and independent death counts.
- UCLA’s Drone Wars project created a comparative dataset across Afghanistan, Pakistan, Somalia, and Yemen using Bureau of Investigative Journalism (BIJ) data [3,4].

These initiatives underscore the need for rigorous comparative analysis — yet no study has systematically compared Somalia and Yemen in terms of strike characteristics and humanitarian outcomes. This paper fills that gap by leveraging detailed open-source strike records from both countries to quantitatively examine differences in civilian impact.

We explicitly test three hypotheses:

1.Hypothesis 1: Civilian Harm Difference

To test whether Somalia and Yemen differ in civilian casualty rates.

2.Hypothesis 2: Drone Effectiveness Across Countries

To test whether the impact of drone strikes differs between Somalia and Yemen, we estimate an interaction model:

3.Hypothesis 3: Reporting Uncertainty

To assess whether casualty reporting uncertainty differs between regions.

To investigate these hypotheses, we assemble a comprehensive dataset of U.S. counterterrorism strikes in Yemen and Somalia, drawn from independent monitoring organizations such as the Bureau of Investigative Journalism [1]. Because casualty counts are uncertain, our data use minimum–maximum ranges [1]. Given that casualty outcomes are over-dispersed count data, we employ negative binomial regression to model civilian casualties while controlling for strike features. This approach allows us to test whether “country” remains a significant factor in humanitarian outcomes once strike type and context are accounted for.

In the following section, we describe the data sources and methodology used in this analysis, before presenting results and implications for policy and scholarship.

Literature Review

Researchers and monitoring groups have spent many years examining how many people are killed in U.S. drone strikes, but most work focuses on one country at a time rather than comparing Somalia and Yemen directly.

Columbia Law School’s Human Rights Clinic, in Counting Drone Strike Deaths, shows that official U.S. numbers often underestimate civilian deaths. They recommend using casualty ranges (minimum–maximum) because information from the ground is often unclear [3].

The Bureau of Investigative Journalism (BIJ) collected open-source reports for every known strike in Yemen, Somalia, Pakistan, and Afghanistan. Their database records both minimum and maximum death counts and distinguishes civilians from militants when possible, noting that reports are often uncertain or contradictory [5].

New America’s Counterterrorism Wars project compiles strike data from Yemen and Somalia, listing total strikes and casualty ranges and explaining how they classify victims when reports are vague or disputed [6].

Together, these sources show that:
1. Independent groups usually find more civilian deaths than official U.S. reports.
2. Although detailed data exist for Yemen and Somalia, most previous analyses summarize each country separately rather than compare them statistically.

Our study fits into this work by using open-source strike records to conduct a direct, quantitative comparison between Somalia and Yemen. Using negative binomial regression, we test whether the countries differ in civilian casualty rates and the uncertainty of reported casualties, controlling for strike characteristics.

References

[1] Currier, Cora. “Everything We Know So Far About Drone Strikes.” ProPublica, 5 Feb. 2013.
[2] Woods, Chris. “Why White House Civilian Casualty Figures Are a Wild Underestimate.” Bureau of Investigative Journalism, 1 July 2016.
[3] Columbia Law School Human Rights Clinic. Counting Drone Strike Deaths. Human Rights Institute, Columbia Law School, Oct. 2012.
[4] Bergen, Peter, David Sterman, and Melissa Salyk-Virk. “The Drone War in Somalia.” New America, 30 Mar. 2020.
[5] Bureau of Investigative Journalism. “Our Methodology.” BIJ.
[6] New America. “America’s Counterterrorism Wars: Methodology.” Future Security Program.