Policy or Panic? The Flight of Ethnic Albanians from Kosovo, March-May 1999



This report describes the departure of ethnic Albanians from Kosovo from late March to mid-May 1999 during the armed conflict between the Federal Republic of Yugoslavia (FRY) and the North Atlantic Treaty Organization (NATO).1 The study is based on administrative records maintained by Albanian government border guards at Morina, other official records of refugee movements, and surveys conducted in refugee camps in Albania and Macedonia.2 Approximately half of all refugees who left Kosovo passed through Morina, and most of them were registered by the Albanian officals, making these border data central to the analysis of mass migration during the spring of 1999.

The study found that refugee flows out of Kosovo occurred in three distinct phases. During the beginning of each phase the flow of refugees was relatively light. Then the number of refugees leaving Kosovo would rise to a relative high point (a peak, group of peaks, or plateau) during the middle of the phase, before tapering off toward the end of the phase. During the first phase (24 March – 6 April), most of the refugees came from western and southwestern Kosovo. In the second phase (7 – 23 April), most of the refugees left their homes in the northern and central municipalities. During the final phase (24 April – 11 May), refugees came largely from the western and southern municipalities. By 11 May, more than 400,000 Kosovar Albanian refugees had entered Albania through the border crossing at Morina.3

This report compares the refugee flows to the location and timing of NATO bombings and mass killings allegedly carried out by Yugoslav forces. It also considers whether bombing raids or mass killings were the proximate motivations for refugees’ decisions to leave their homes. We conclude that bombing, mass killings, and mass migration were pervasive throughout Kosovo during this period, and that the three processes occasionally coincided. We found, however, that neither the bombing raids nor the mass killings occurred at times and places sufficient to be the primary motivation for Kosovar Albanians to leave their homes. Therefore, another process must have been at work to result in the exodus of hundreds of thousands of people in remarkably coherent waves.

Why did Kosovar Albanians leave their homes?

During the conflict many arguments were advanced about why the Kosovars were leaving their homes. Yugoslav officials contended that Albanian Kosovars were fleeing NATO bombs. Other analysts posited that people were escaping fighting between the Kosovo Liberation Army (KLA) and Yugoslav government forces. These explanations imply that Kosovar Albanians were motivated to leave their homes by their own fear or panic. Meanwhile, NATO officials argued that Yugoslav forces were subjecting Kosovar Albanians to a systematic policy of “ethnic cleansing,” a euphemism for the forced eviction of an ethnic group.

Human rights reports on Kosovo

Among the many reports by non-governmental and inter-governmental organizations about human rights violations and grave breaches of the Geneva Conventions during March-June 1999, see A Village Destroyed: War Crimes in Kosovo. New York: Human Rights Watch (Oct. 1999); “Ethnic Cleansing” in the Glogovac Municipality. New York: Human Rights Watch (July 1999). See also Kosovo Human Rights Flash series, 1-51 (February-July 1999). War Crimes in Kosovo: a Population-Based Assessment of Human Rights Violations against Kosovar Albanians. Boston/Washington DC: Physicians for Human Rights & Program on Forced Migration and Health, Center for Population and Family Health, The Joseph L. Mailman School of Public Health, Columbia University (Aug. 1999) [cited hereafter as PHR 1999]. Kosovo/Kosova: As Seen, As Told, An analysis of the human rights findings of the OSCE Kosovo Verification Mission October 1998 to June 1999. Warsaw, Poland: Organization for Security and Cooperation in Europe (Dec. 1999) [cited hereafter as OSCE 6 December 1999].

There are several potential methods for adjudicating between these contending explanations. One would be to ask the Kosovars themselves why they left. In its survey of Kosovar Albanian refugees in Albania and Macedonia, Physicians for Human Rights (PHR) found that Kosovars reported leaving their homes either because they were forcibly expelled by Yugoslav regular or paramilitary forces (68%), or because they were afraid of those same forces (23%). None of the 1,180 individuals interviewed reported that they were fleeing NATO bombing.4

Another method would be to examine the context of interpersonal and interethnic violence and civil war in an effort to determine what conditions might have driven the refugees from their homes. Human rights organizations have conducted extensive interviews with Kosovar Albanians to address this question. For example, the PHR study reported that between 48-60% of all survey respondents witnessed the destruction of documents, burning of homes, the looting or destruction of property, and/or robbery by Yugoslav regular or paramilitary forces.5 Numerous human rights reports quote Kosovar Albanians explaining how they were threatened in their homes by Yugoslav army or police officials, or how they were subjected to artillery fire. Many refugees described being transported to the borders in trains or buses, or being forced by Serbian authorities to move themselves on tractors, in private cars, and on foot.

Unlike previous analyses which have relied on refugee testimonies, this study deduces the causes of the refugee exodus by examining the statistical patterns of the exodus itself. By considering how many people left each municipality over time, and comparing those patterns to the times when NATO bomb attacks and alleged mass killings occurred, we draw conclusions about the plausibility of competing explanations for the migration.

This study breaks new ground for human rights analysis by using objective administrative data to evaluate – to corroborate or to refute – the claims made by witnesses and survivors, as well as to compare the claims of the various political actors involved in the conflict. The goal is to establish a solid empirical basis for legal, political, academic, journalistic, and other analyses of the mass migration of Kosovar Albanians in this period.

Empirical overview of Kosovar refugee flow

From 28 March to 28 May 1999, the Albanian government border guards in Morina maintained entry records under difficult conditions. During this period they compiled 690 pages of records in which over 19,000 groups of Kosovar Albanians were registered.6 According to these records, the number of refugees entering each day was similar to independent counts of refugees entering at Morina maintained by the United Nations High Commission for Refugees (UNHCR) and later by the Albanian government’s Emergency Management Group. When the estimates are not equal, the border guards’ numbers are always lower. Examining the different estimates, it is clear that on days when tens of thousands of Kosovar Albanians crossed the border, the guards’ registry system simply broke down as people streamed through the post, whereas on low-flow days, the estimates tend to coincide.7 The border guards managed to register groups including more than two-thirds of the more than 400,000 Kosovar Albanians who entered Albania through Morina.

Leaving home and crossing the border

On any given day, slightly more than half of the people who crossed the border at Morina had left their homes earlier that same day. However, the other half of people crossing the border were in transit for longer times, in some cases for several weeks. The transit process was accounted for in the discussion of the refugee movement. In Part II, the analysis of refugee flow during the three phases looks at the patterns of people departing their homes, not the patterns of people crossing the border.

Refugees interviewed in camps reported when they had left their homes and when they crossed the border: the difference between these two dates is their time in transit. By using methods described in Appendix A (Section A3) and the data on transit times, the number of people who crossed the border were transformed into estimates of the number of people leaving their homes, over time and for each city and village of origin.

With authorization from the Albanian Ministry of Public Order, representatives of this study copied the entire set of surviving border records in early June.8 Electronic images of the handwritten records were made in the nearby town of Kukës, and then brought to the Albanian capital of Tirana.9 The handwritten images were keyed into a simple database consisting of the name of the head of the household in the group, the home village or city of the group, the number of people in the group, and the date they entered Albania.10 All of the information was used for this study. A geographic code was assigned to each record in which the group’s home village or city was reported. Using geographic information systems (GIS) references, each record was mapped to a specific point in Kosovo. These records, along with various surveys and refugee camp listings, comprise the basic data for this project.

From late March – late May 1999, ethnic Albanians left their homes in Kosovo and entered Albania in three distinct phases. Approximately 95 percent of the Kosovar Albanian refugees who entered Albania did so during one of these three phases.11 The number of people crossing the border over time is presented in Graph 1.1.

Graph 1.1: Number of Kosovars entering Albania at Morina, by two-day period
Graph 1.1
Phase 1
Phase 2
Phase 3

Graph 1.1 represents the number of Kosovar Albanians entering Albania for each two-day period from late March to late May. The first phase rises from 24-25 March to a plateau then declines precipitously after 6 April. The second phase continues at a low level until mid-April when it rises to a peak at 15-16 April, declines to a second smaller peak, and then declines to its lowest level during the whole period. The third phase picks up on 24 April, rising to two peaks in early May that represent the last surges in refugee flow. After 11-12 May, refugee flow remained low until the end of the conflict in June.

The three phases detailed here also correspond to substantial changes in the refugees’ municipalities of origin. Refugee flows tended to be composed of people from similar regions during each phase, and so refugees from municipalities of the south and west are distinguished from refugees originating in municipalities of the north and east.12

Graph 1.2 shows the proportion of all Kosovars crossing the border at Morina who originated in the southern and western prefectures of Kosovo, over time. The data are again smoothed as in Graph 1.1 by adding days together into pairs, called two-day periods. By aggregating the data into two-day periods, there are fewer but more stable data points. Disaggregating the data to the daily level does not affect the analysis.

Graph 1.2: Proportion of Kosovar Albanians entering Albania who originated from municipalities in the south and west, by two-day period
Graph 1.2

The three phases seen in the wave pattern in Graph 1.1 appear in Graph 1.2 as distinct mixes of places from which people are emigrating. During Phase 1, the first wave of people came overwhelmingly from the south and west. In Phase 2, a second large wave of people came, but this time from the north, east, and central regions, shown by the relative low point in the middle part of Graph 1.2. Then in Phase 3, another group exited. As in Phase 1, refugees entering Albania during Phase 3 came largely from the south and central areas.

The phases are defined by periods of relatively high and low flow and by periods during which refugees came from distinct regions. The discussion in Part II will examine the three phases for each of twelve municipalities in the context of NATO bombing patterns and alleged mass killings.

Generalizing from the Albanian data to all Kosovar Albanians who left their homes

The data in this report are based primarily on the records maintained by the Albanian government border guards at Morina. However, it might be possible that the patterns of people leaving Kosovo and entering Macedonia, Montenegro, or Bosnia-Herzegovina were substantially different from the patterns of people entering Albania. That is, the findings in this report could be an artifact of the choice of data sources.

Sensitivity analysis

The statistical techniques used in this report to manage missing data and to transform border crossing data into home leaving data were subjected to sensitivity analysis (see Appendix A, Sections A5 - A7). This method tests both whether the substantive findings are robust to potential weaknesses in the data, and whether the kinds of methods used may have influenced the results. All of the alternative analyses led to findings substantially the same as those reported in the main body of the report.

The possibility that the findings were the result of using data only from Albania is examined in Appendix A (see Section A8). This analysis uses a number of sources of partial data on patterns of Kosovar migration to countries other than Albania. The conclusion is based on the following empirical observations.

  1. The findings for Albania (presented in Part II) are strong and clear.
  2. The data available for Macedonia yield findings consistent with the findings for Albania.13
  3. Preliminary evidence from refugees in Bosnia14 and Montenegro tends to support the Albanian analysis.
  4. An overall analysis that combines data from Bosnia, Montenegro, and Macedonia with data from Albania shows that the combined analysis is not substantially different from the analysis using the Albania data alone (see Appendix A, Section A8).

Based on these observations, we conclude that the findings presented in this report are generally applicable to the universe of all Kosovar Albanian refugees, within tolerances discussed in Appendix A.

Notes for Part I

1 On 27 May 1999, the Chief Prosecutor of the International Criminal Tribunal for Former Yugoslavia (ICTY) alleged that war crimes had been committed by five senior Yugoslav officials, including Slobodan Milosevic. The indictment is a legal instrument, not a finding of fact. The Chief Prosecutor herself noted that the accused are entitled to the “benefit of the presumption of innocence” until convicted (ICTY press release JL/PIU/404-E). Return to Text

2 Only the Albanian border records are used. Appendix A, Data and Methods, outlines a number of conclusions as to why these data may be generalized, within limits, to the entire population of Kosovar Albanian refugees during this period, including those who exited to Bosnia-Herzegovina, Montenegro, and Macedonia. An important caveat: internally displaced persons that remained in Kosovo are not part of the analysis because we have no data about them. Return to Text

3 Counting only refugees who entered Albania after 23 March, more than 97% had entered Albania by 11 May: only 3% entered during the final five weeks of the conflict. Return to Text

4 These proportions have a 95% confidence interval of approximately ±2.7% and 2.5%, respectively (PHR 1999: 40; the calculation of the confidence interval is ours). Return to Text

5 PHR 1999: p. 42. Proportions of Kosovar Albanians reporting that they witnessed the different listed violations vary within the stated range. Return to Text

6 See Appendix A, Section A1 for a detailed discussion of these data. Return to Text

7 See Appendix A, Section A2.1 entitled “Unregistered border crossers.” Return to Text

8 Three days from mid-May were lost; in discussions, the border guards recalled that these were low-flow days, which the EMG estimates confirm. From sequences of page numbers, we believe that about five or six other pages were lost from days in early and mid-April. See Appendix A, Section A2 for a discussion of how these missing data were managed. Return to Text

9 The original records were returned to the border guards in Morina. Copies of the electronic images and the resulting database were given to the Ministry of Public Order. Return to Text

10 Measures to protect data confidentiality have been taken at each step: only three people have had access to the border data since they were keyed, and just four more people were involved in the keying and coding. The data were encrypted (using PGP) before being transferred from the team in Tirana to AAAS in Washington, DC. Return to Text

11 Slightly more than 2% entered before the first wave; more people entered Albania in late 1998 and January to mid-March 1999 but were not systematically registered at the border. Return to Text

12 The list of municipalities, with their Serbian and Albanian names and the region in which they are located is presented in Appendix C. Return to Text

13 The data from Macedonia are from PHR’s survey; see Appendix A, Section A8.2. Return to Text

14 The Bosnian data are from a survey of Kosovar Albanian refugees in Bosnia-Herzegovina conducted by the Human Rights Center of the University of California-Berkeley in late-June early-July 1999. Return to Text

Previous   Table of Contents  


AAAS Science and Human Rights Program