Data on Kosovo – Other

The other data is in three files. All of the files are comma-delimited UTF-8 (like ASCII but including the characters to render Serbian names). The fields in each file are described below.

If you use these data, please cite them with the following citation, as well as this note:

“These are convenience sample data, and as such they are not a statistically representative sample of events in this conflict.  These data do not support conclusions about patterns, trends, or other substantive comparisons (such as over time, space, ethnicity, age, etc.).”

Human Rights Data Analysis Group. (2002). Database of NATO airstrikes, geographic coding, and KLA activity in Kosovo.

Geographic Coding

The first file is pcodes.csv. It contains 1588 records.

All places identified by interviewees were coded to specific geographic locations in a single coding scheme. A coding scheme uses a list of place names referenced to a unique id number and a location specified by latitude and longitude. Since many places have the name name, a place list is not uniquely identified by names. Instead, a code (the pcode) is assigned to each distinct place.

We began with the geographic structure described in Policy or Panic: The Flight of Ethnic Albanians from Kosovo and Political Killings in Kosova/Kosovo, March-June 1999, using 29 municipalities. These structures omitted many places introduced in the new data acquired for the present study. The place list available at the on-line Humanitarian Community Information Centre (HCIC) linked place names to grid positions on a detailed atlas. The HCIC list was used to standardize place names.

All place codes were coded for latitude and longitude. A first pass used the U.S. National Imaging and Mapping Agency’s (NIMA) Populated Place Locations list. The NIMA list was linked to the HCIC list using place names. When names were ambiguous, we hand-linked the codes using municipality names and checking places on the HCIC map and a commercial map. Using the HCIC list and map, as well as a commercial map, we developed computer routines that confirmed every place code’s latitude and longitude against the grid coordinates in the HCIC map. Locations which did not fall in their grid coordinate were hand-plotted and rechecked.

The pcodes are the standard geographic references used throughout the data in this project. In some cases, we used only the municipality codes (mcodes) which are referenced as integers between 1 and 29 inclusive. Note that the reference in pcodes.csv can be found by multiplying the mcode by 1000. Much of the analysis is done at the regional level (north, east, south, and west). The mapping of municipalities to regions can be seen in Figure 3, page 7 of the 3 January 2002 report given to ICTY.

This map includes only 29 municipalities. Some name lists include the municipality of Malisevo which did not exist during the first two quarters of 1999. During the time of the conflict in 1999, Malisevo was part of four other municipalities.

Several cities and villages have the same names as municipalities. Given one of these names, it was not always possible to determine whether the municipality or the village was being described. Sometimes, too, the same place name occurred in more than one municipality (e.g., Drenovac is a city or village name in four municipalities: Orahovac, Decani, Pristina, and Klina). Finally, there were cases where no place coding could be assigned (e.g., “in the mountains”).

Pcodes evenly divisible by 1000 are the municipality-level codes. The municipality of a pcode can be determined by examining the thousands’ part of the pcode. For example, the town of Alabak (pcode=18001) is in the municipality of Podejevo (pcode=18000).

Distances between locations were calculated using their latitude and longitude. These distances were used to help our coders to determine whether witnesses’ conflicting reports of locations plausibly referred to the same place. Locations fewer than 10 kilometers distant from each other were routinely treated as the same locations.

It contains the following fields.

Field name Field description
pcode The place code of this record. Note that these are unique.
sortname The name of the place in simplified latin characters
serbname The name of the place (encoded with UTF-8 characters).
latitude Degrees N latitude
longitude Degress E longitude

NATO airstrikes

Each record in nato_bombing_data.csv contains information about one reported airstrike; there are 364 reported airsrikes in this dataset. The dataset was derived from a draft version of the report later published by Human Rights Watch as Appendix A of “Civilian Deaths in the NATO Air Campaign” (February, 2000). This information includes the minor correctons described in the AAAS corrigendum given to the OTP on 31 October 2002. Note that only the airstrikes identified as occurring in Kosovo were included in this dataset; the attacks in Serbia and Montenegro were not relevant to this statistical analysis.

Field name
Field description
bdate The date on which the airstrike occurred.
muni One of the 29 municipalities described in the geographic data.
details Information about the source of information about this airstrike.

KLA activity

The information in kla_activity.csv was coded from a database of references to activity of the Kosovo Liberation Army (KLA) maintained by the Kosovo investigation team of the ICTY Office of the Prosecutor (OTP). This database was shared confidentially with AAAS as part of our preparation of our report and the testimony presented on 13-14 March 2002. Information confidential to the OTP pursuant to Rule 70 of the ICTY “Rules of Procedure and Evidence” was removed before this publication.

Each record in the KLA_activity.csv file represents a report of either Yugoslav government casualties sustained in armed confrontation between the KLA and Yugoslav authorities (type=k), or an exchange of fire between Yugoslav forces and KLA forces (type=b). Note that each record may contain more than one event. For example, the report of casualties in municipality 19 on 20 March 1999 involved three people wounded, and so for that record, count==3. Two records with type=b have count=0.5, which indicates that Serb authorities reported that they were shot at by KLA forces, but they did not return fire.

Field name
Field description
kdate The date on which the event occurred.
type Type of event: b=battle, k=casualty
muni One of the 29 municipalities described in the geographic data.
count Number of events of the type described above (there may have been more than one in each coded report)

Last updated: 1 November 2002

Back to main Kosovo Data page.

Our work has been used by truth commissions, international criminal tribunals, and non-governmental human rights organizations. We have worked with partners on projects on five continents.