Get the data: Drone wars

Get the data: What the drones strike

Damadola, PAKISTAN:  (FILES) In this picture taken 25 January 2006, a Pakistani tribesman sifts through the debris of his collapsed home after a 13 January, US air strike in the remote village of Damadola,in  Bajur Tribal Agency bordering Afghanistan.  Al-Qaeda deputy Ayman al-Zawahiri escaped death in a US missile strike last month on a remote tribal village bordering Afghanistan, Pakistani President Pervez Musharraf said. The January attack killed a close relative of al-Zawahiri and four other militants, Pakistan military spokesman Major General Shaukat Sultan quoted Musharraf as saying 12 February.  AFP PHOTO/Tariq MAHMOOD/FILES  (Photo credit should read TARIQ MAHMOOD/AFP/Getty Images)

A Pakistani tribesman sifts through the rubble of his house after an attack in January 2006
(Photo: Tariq Mahmood/AFP/Getty Images)

The Bureau is publishing, for the first time, data showing the types of targets that have been reportedly attacked by CIA drones in Pakistan.

The research is a joint project by the Bureau, Forensic Architecture, a research unit based at Goldsmiths University, London, and Situ Research in New York. The data feeds this interactive website mapping the strikes, the types of target attacked, and their relative scale.

Download the data here

This data reflects our understanding of the strikes as of May 2014. It will not be updated – so if new information emerges about the attacks, it will not be included in this dataset, although it will be included in our main drones databases.

Below is a brief methodology showing how we compiled the data and the definitions we used. There is a more detailed methodology for our broader research on drones here.

Related story – Most US drone strikes in Pakistan attack houses

How we compiled the data

We have used the index of reports the Bureau has compiled for each strike, extracting and recording data on what targets were reportedly hit. The targets were divided into domestic, public, religious, and commercial buildings, outdoor gatherings (such as meetings and funerals), and vehicles. Because of the very small number of public buildings, commercial buildings and outdoor gatherings that were hit, we have recorded these as ‘Other Buildings’.

Reporting is sometimes vague about the target, and sources sometimes directly contradict one other. The Bureau uses the minimum figure in a range for data analysis in its investigations, to reflect these uncertainties. For example, a strike on March 16 2011 killed at least three people but reportedly hit either a vehicle or a house. This has been recorded as 0-1 vehicles hit and 0-1 houses hit and is represented on the interactive map as ‘target unclear’.

This information about targets was combined with casualty estimates for the strikes. The locations were largely identified and plotted on the geo-platform using a CIA map of Waziristan, declassified in 2007. Latitudes and longitudes are based on reported locations, which are often only as specific as town or district, and so are not precise.


• Civilian
The Bureau classifies all individuals credibly reported as civilians as such. Where the dead are described as ‘tribesmen’, ‘locals’ or ‘people’, we believe this indicates possible civilian casualties and reflect this using the 0-X range.

The Bureau has recorded a number of female casualties in the drone war. It almost always classes women as civilians: in the FATA region of Pakistan, where the strikes take place, reports of female militants are exceedingly rare.

• Domestic building
Where drones attack buildings, these are often described as ‘compounds’ and sometimes even ‘militant compounds’. However, local sources confirm that these are typically domestic buildings that are often rented or commandeered by militant groups.

• Drone strike
A missile or set of missiles fired by a drone or drones at a single location. Where missiles hit more than an hour apart, we counted these as separate strikes. Where drones hit locations more than a couple of miles apart we also count these as separate strikes, even when they take place in quick succession.

• Other Buildings
A small number of strikes have targeted buildings that are neither domestic buildings nor madrassas and mosques. These include commercial buildings and disused government buildings.

• Religious (Madrassa/mosque)
A madrassa is a seminary – a religious school. These are usually residential facilities that educate children and youths. A very small proportion of all attacks have hit madrassas or mosques, but they have tended to have very high death tolls.

• Target Unclear
Reporting is sometimes vague about what was hit in a strike – and sometimes media reports contradict one another in terms of what type of target was attacked. In these cases, we have categorised the target as ‘unclear’.

• Vehicles
This category encompasses cars, pick-up trucks, four-wheel drives and motorbikes.


• Media sources
The most comprehensive public information on casualties generally lies in the thousands of press reports filed by reputable national and international media outlets. The bulk of our sources are in English, but some Urdu reporting has been used.

Further information on media sources

• Other sources
The Bureau has carried out several field investigations into possible civilian deaths. The data also incorporates the fieldwork of credible researchers (for example Stanford Law School and New York University School of Law) and evidence filed in legal cases brought in Pakistan and elsewhere on behalf of civilian drone victims. Leaked US intelligence reports, WikiLeaks diplomatic cables, sanctions lists and ‘most wanted’ lists, and jihadist forums and websites have also been used where relevant.

Further information on other sources

Additional details

In each category we aimed to note the nature and extent of the reported structural damage. We assigned a code for the degree of structural damage, where reported:

1. Minimal damage

2. Moderate damage

3. Severe damage

4. Completely destroyed

The damage was not consistently reported in each strike, and was not consistent in each report of each strike. In some cases, we have assigned more than one number to a strike. For example, when some sources reported the building was totally destroyed but others only reported severe damage the strike was assigned 3-4.

Where there are differences in reporting the number of missiles fired in a strike, we have similarly recorded these as a range.

We used the same technique to accommodate inconsistencies between sources in the number and types of targets hit in the strikes. On average one building or one vehicle were hit in each strike. However several strikes reportedly hit multiple targets. For example, drones destroyed a convoy of trucks in a strike on December 27 2010, killing 18-25 people. However sources reported either two or three trucks were hit. The Bureau has recorded this strike as hitting 2-3 vehicles.

Follow Alice K Ross and Jack Serle on Twitter. Subscribe to the Bureau’s drones podcast and newsletter.