Fiche du document numéro 8601

Num
8601
Date
Jeudi 31 mai 2001
Amj
Auteur
Fichier
Taille
442790
Pages
56
Urlorg
Titre
A Quantitative Analysis of Genocide in Kibuye Prefecture, Rwanda
Mot-clé
Source
Type
Langue
FR
Citation
A Quantitative Analysis of Genocide in Kibuye Prefecture,
Rwanda.
by
Philip VERWIMP

Development Economics

Center for Economic Studies
Discussions Paper Series (DPS) 01.10
http://www.econ.kuleuven.be/ces/discussionpapers/default.htm
May 2001

A Quantitative Analysis of Genocide in
Kibuye Prefecture, Rwanda

Philip Verwimp*
Ph.D Candidate
Fund for Scientific Research
Economics Department
Catholic University of Leuven, Belgium
Philip.verwimp@econ.kuleuven.ac.be

Abstract of the paper

This paper is a quantitative study of the genocide in the prefecture of Kibuye in Rwanda in 1994. We use an
original data base developed by the organisation of the survivors of the genocide (IBUKA) who collected the
data by house-to-house fieldwork. The data contain information on the age, sex, commune of residence before
the genocide, the professional occupation of the victims, the place and date of death and the weapon used to kill,
for a total of 59.050 victims of genocide. For one commune (Mabanza), we re-coded the data, present detailed
statistics and perform an analysis of survival chances. From the analysis, we derive that Tutsi from the sectors
of Mabanza commune whose Tutsi population did not (or only in limited numbers) go to the Gatwaro Stadium
had a better chance to survive the genocide in Kibuye. For the whole of the prefecture, we present an estimation
of the daily killing rate, estimations of the number of Tutsi killed in the major massacres and the weapons
used. For over 25.000 victims for which the data file has complete information, we present a logistical regression
explaining the use of either a traditional weapon or a fire-arm. The analysis shows that the probability to be
killed with a fire-arm depended on the commune of residence of the victim, the age of the victim, the number of
days after April 6 the victim was killed and on interaction effects between the latter two variables and the sex
of the victim.
* The author is a research scholar from the Fund of Scientific Research (Flanders, Belgium). The author owes
many thanks to the Fund for the grants that enabled two research stays in Rwanda in 1999 and in 2000. The
author expresses his gratitude to IBUKA and in particular to F.R. Ruvukanduvuga for the permission to use their
data file. It is worth emphasizing that the author can use the file without restrictions. Academic freedom in the
use of the data file for the analysis of the genocide were unconditional. I am indebted to L.Berlage, S.Cook,
S.Dercon, A.Desforges, W.Seltzer, and seminar participants in the Genocide Studies Seminar at Yale University
for insightful and critical comments to a draft version of this paper. I am grateful to F. for the skillful editing of
the paper. All responsibility for the paper remains with the author.

2

A Quantitative Analysis of Genocide in
Kibuye Prefecture, Rwanda

1. INTRODUCTION

According to African Rights, the genocide in Kibuye Prefecture was the most thorough
genocide of all the prefectures in Rwanda (African Rights, 1995, p. 394). Locked in between
the Prefectures of Gisenyi in the north, Gitarama in the east, Cyangugu in the south and Lake
Kivu in the west, the Tutsi from Kibuye had nowhere to flee. They were left at the mercy of
their killers. The latter wanted to kill every man, woman and child known as Tutsi in their
communes. Knowing that they could only count upon themselves, a large number of Tutsi,
estimated in this paper at almost one quarter of all Tutsi who were killed in Kibuye Prefecture,
mounted a strong resistance against the forces of genocide. In Bisesero, a mountainous area in
Gishyita commune, these Tutsi succeeded in defending themselves for more than a month
after the start of the genocide. One of the findings of this paper is that the fate of the Bisesero
Tutsi was unlike the fate of all other Tutsi in the prefecture. As I shall document in the paper,
seventy-five percent of the Tutsi from Kibuye were killed before the end of April 1994, the
Tutsi of Bisesero forming the main exception.
The analysis presented in this paper is a statistical study of the genocide in Kibuye Prefecture.
The main purpose of the study is to use diverse statistical tools to document the genocide, to
show its mechanisms, its brutality and above all, its speed. Section two gives a brief profile and
a map of the prefecture. Section three presents the main data file used for this study and a
general overview of population figures. I discuss the shortcomings of the data file. In section
four, I present analysis of survival chances for one commune. Section five presents an
estimation of killing over time, which may be the main contribution of the paper. In order to
keep the text readable, numerous computations are put in the appendix. I estimate mortality
figures throughout the three months of the genocide in Kibuye in general and Bisesero in
particular. Section six extrapolates figures from Kibuye together with other prefectures to
Rwanda as a whole. In part seven, I do a detailed study of the use of firearms in the genocide
in Kibuye. Part eight compares the results of the statistical analysis with documents written by
the perpetrators of genocide and section nine concludes.

3

2. PROFILE OF THE PREFECTURE

2.1. Neglect and Interest
The Prefecture of Kibuye is situated in the west of Rwanda bordering the Kivu Lake in the
west, the Prefecture of Gitarama in the east and the Prefecture of Gisenyi in the north. Before
the genocide, there was no paved road linking the prefecture with other prefectures or with
the capital Kigali. Kibuye was and is one of the poorest prefectures of Rwanda. In terms of
per capita income, Butare and Gikongoro are poorer, but Kibuye was the least integrated in
the Rwandan economy, as a result of the lack of transport infrastructure and the geographical
location of the prefecture. Jobs giving access to off-farm income were in short supply. This
means that the population lived by virtue of its soil, its cattle and its small intra-prefecture
trade, even more than in other prefectures. There was trade with Zaire, especially in cattle
products, but it remained small-scale trade. The main reason for the neglect of the prefecture
by the Kayibanda and Habyarimana regimes was the absence of political representation from
the prefecture at the highest level of government. Kayibanda favored Gitarama and
Habyarimana favored Gisenyi. After Butare Prefecture, Kibuye counted the largest Tutsi
population of Rwanda.
The tea-plantations and tea-factory in Gisovu Commune were the only object of interest for
the Habyarimana regime. The plantation and the factory were managed by Ocir-thé and
directed by Alfred Musema, member of the Akazu. Since Rwanda only had six tea-plantations,
the Gisovu plantation was of considerable importance for export earnings. With the decline in
the price of coffee at the end of the eighties, an increase in tea production and tea export
became an important objective for the Habyarimana government. The local peasant
population was very hostile to the establishment of the tea-plantation since their land was
expropriated. The peasant families had to move to other, less fertile land or even migrate1. All
but one of the tea-plantations was state-owned. The value of tea production to the regime
became clear in the event of war. In January 1994, Human Rights Watch writes that Rwanda
bought a huge quantity of arms in exchange for 1 million US$ in cash together with the future
tea harvest from the Mulindi tea plantation2. As in the case of Mulindi, the Gisovu plantation
and factory were state-owned. Most of the tea producing facilities were financed by donor
agencies, making the tea industry, and more specifically its high operating costs, a good
1
2

Bart., F., Montagnes d ‘Afrique, terres Paysannes : Le cas du Rwanda, 1993, p. 456.
Human Rights Watch Arms Project, 1994, p. 15.

4

example of rent-seeking by the Akazu members. Only the Akazu really benefited from tea
production3.

2.2. Map of the prefecture
Rwanda

Kibuye

Communes of Kibuye
(Ru) Rutsiro; (Ma) Mabanza; (Ki) Kivumu; (Git) Gitesi; (Mw) Mwendo; (Go) Gisovu ;
(Git) Gitesi ; (Gy) Gishyita ; (Rw) Rwamatamu
* Bisesero sector in Gishyita commune

3

I refer to P. Uvin’s book on development aid and its relation with the genocide for a general discussion of
rent-seeking in Rwanda and many examples.

5

3. DESCRIPTIVE STATISTICS

3.1. The method used by IBUKA
The organization of the survivors of the genocide, named IBUKA, has undertaken a large
research project with the main objective of finding all the names of the victims of the
genocide in Kibuye Prefecture. They proceeded alongside the administrative organisation of
Rwandan society. Kibuye Prefecture is divided into nine communes. Each commune, having
on average 50.000 inhabitants, is subdivided into several sectors. These sectors on their turn
consist of several cells. Commune by commune, sector by sector and cell by cell, IBUKA
collaborators went into all families of Tutsi survivors and of Hutu who did not participate in
the genocide to find the names of the murdered Tutsi. The project was financed by the Dutch
Embassy in Rwanda and employed about two hundred enumerators. The enumerators came
from the commune where they were doing the interviews or were familiar with it. The
majority of the enumerators had high school training behind them. It was not easy for IBUKA
to find experienced enumerators, since almost all educated Tutsi were killed. This lack of
experience, together with the choice of IBUKA to work only with survivors as enumerators,
had a negative effect on the quality of the data collection process in some communes. There
was one supervisor for each commune who was monitoring the work of about 20
enumerators, at least one enumerator per sector. The enumerators and supervisors did not
receive any statistical or interview training. The result is a nominative dictionary with the
names of almost 60.000 victims of genocide of the prefecture published in December 1999.
The project also registered whenever possible, the age and the profession of the victim, the
place where the person was killed and the weapon that was used. The present paper uses the
IBUKA data file to analyze the statistics of genocide in Kibuye Prefecture. It is worth
emphasizing that the data file was not constructed for statistical purposes, but for the nominal
documentation of genocide victims in Kibuye. The latter objective remains the prime value of
the documentation project. Given the death toll among Tutsi, the very difficult living
conditions after the genocide, the lack of training and adequate research facilities, the result of
the project is all the more remarkable.
It follows that the use of the data file for statistical purposes is problematic. Firstly, the author
did not take part in the data collection process and secondly, the quality of the data differs

6

substantially between communes and sectors. These problems should be clear from the outset
of the paper. I tried to tackle them on several fronts: I did personal interviews of the
enumerators and the supervisors of the data collection process and I applied statistical
methods to correct for missing data. Since the enumerators were all survivors of the genocide
in Kibuye Prefecture, it is possible that the number of victims was inflated or that respondents
made up stories to satisfy the enumerators. From my interviews, I retain that most of the
respondents were Hutu, who have, I believe, no incentive to inflate victim figures (on the
contrary one would say). One also has the impression that Hutu who did not participate in the
genocide feel a need to come forward with accurate information to clear themselves of guilt.
This is important if one wants to avoid a whole ethnic group becoming identified with the
perpetrator image. Some of the respondents indeed identified more with the survivors of the
genocide then with the perpetrators. This is especially the case for Hutu widows of Tutsi
husbands. In the event where data were doubtful, I decided not to use them. This is the case
for the many victims whose date of death is not registered, but for whom the weapon used to
kill the victim is nevertheless indicated. This seems strange and for these victims I did not use
the data on the weapon in my analysis of weapons used to kill (I refer to section 3.4 for this).4
The organizers of the data collection process also intended to register the 'cause' of the death
of a person. The different categories that were mentioned for the 'cause ' of death in the
registration books are: being Tutsi, being a Tutsi-friend, having a Tutsi physical look, political
opposition, having a Tutsi mother. However, 98% of the entries indicate the first possibility,
meaning that this part of the data collection failed :5 IBUKA has only registered the Tutsi
victims of genocide and did not, or only sporadically, register persons who were killed for
other reasons than being Tutsi.6
The published Nominal Dictionary gives the names of the victims of the genocide in Kibuye
Prefecture. Looking at the original books that were used to collect the data however, I learned
that information on the surviving members of each household was also collected (but not

4

I owe this remark to Alison Desforges. The distribution of all weapons used in the genocide (% for each
weapon) however does not change very much when one includes these cases.

5

This is problematic, because it means that the IBUKA data file only contains Tutsi among the registered
victims and not for example the Hutu widows of Tutsi husbands. The author could not verify the
magnitude of interethnic marriage in Kibuye, which maybe sizeable, but should not be overestimated either.

6

It is interesting to see that the Government of Rwanda used the IBUKA scheme to make a national census
of genocide victims in August 2000. Results of this national data collection process are not yet known at the
time of writing.

7

published). This additional information can give empirical insight into the execution of
genocide. The information on the survivors also deals with age, sex, profession and area of
residence, comparable to those of the victims.
As with the victims, the quality of the data differs substantially between communes and
sectors. Since the data on the survivors was not computerized, I decided to make a new
database with victims AND survivors. In that respect I could also code the responses in a way
that make the data more suitable for statistical analysis. Budget limitations allowed me only to
organize the re-coding of one entire commune (Mabanza) and of half of the sectors of
another commune (Gitesi). Needless to say, it would be interesting to re-code the remaining
communes too. In some communes however, the benefit of re-coding would be small,
because the data are bad, whereas in other communes the benefit would be great. In general,
one cannot overcome the missing data problem. For some sectors of Gitesi for example,
enumerators did not enlist ages or dates or professions of the victims nor of survivors. In that
case, it makes no sense to re-code data. I decided nevertheless to re-code some sectors of
Gitesi commune in order to learn as much as possible on the genocide in this commune. We
come back to this in section four of the paper (see also appendix 2).
Table 1: Quality of the data collection process
Commune

ages

dates

places

survivors

re-coding worthwhile

Bwakira
Gishyita
Gisovu
Gitesi
Kivumu
Mabanza
Mwendo
Rutsiro
Rwamatamu

+
+
+
+
+
+
+
+
+

+
+
+
+/+/+

+/+
+/+
+/+

+
+
+/+
+
+

yes
yes
no
partly
partly
yes
yes
no
yes

data is good (+), data is average (+/-) or data is bad (-) where ‘good’ means that for most victims (or
survivors) the data on that item were indeed collected. ‘bad’ meaning that data were not collected (missing) and
‘average’ meaning that data for some victims were collected and not for others.

8

As one can see, the quality of the data for Gitesi, Gisovu and Rutsiro communes is particularly
poor. Gitesi commune in particular is an example of poor fieldwork.
I did not have access to the raw data of the 1991 census in Rwanda; tool demographers
normally use it to analyze excess mortality. The present data file only contains data on the
Tutsis who were killed during the genocide. The IBUKA project succeeded in finding 59.050
victims of genocide in Kibuye Prefecture, meaning 12 % (59.050 of 500.000) of the
population of that prefecture. Table 2 presents the general figures. The figure of 59.050 is an
underestimation since not all victims were registered and the Hutu victims were also not
registered.

3.2. General Figures of Genocide in Kibuye Prefecture
Table 2: General figures of genocide in Kibuye Prefecture

Total population of the Prefecture in 1991
Population registered as Hutu
Population registered as Tutsi
Population registered as Twa
Foreign, other or undetermined
Murdered population found by IBUKA

Number*
473.920
399.470
71.225
1.490
1.735
59.050

Tutsi population registered as murdered
Tutsi population not registered as murdered

Number
59.050
12.175

*

Percent
100
84,3
15,0
0,3
0,3
12,4
% of Tutsi**
82,9
17,1

I have no access to exact figures of the population size in March 1994. Total population in the prefecture
probably reached 500.000 (≅473.920*(1.03)2) by March 1994.

** Accounting for population growth, the figures become 78% registered as murdered and 22% not registered.

According to the 1991 census and the figures found by IBUKA, 12,4% of the population of
Kibuye Prefecture was killed in the genocide, meaning approx. 80% of the Tutsi population.
This means that 15 to 20% of Tutsi, (between 10.000 and 15.000 persons) survived the
genocide in Kibuye Prefecture. Table 3 provides information on the genocide in each of the
communes of the prefecture.

9

Since I did not have access to the 1991 population figures according to ethnic affiliation by
commune, I cannot determine exactly how many Tutsi survived the genocide in each of the
communes. Apart from this, the IBUKA data file gives a lot of other information that one
would normally not find in census data. Since the file has, among other information, the dates
and places of the massacres and the weapons used in the massacres, it is an unique source of
information for the study of the genocide in Kibuye Prefecture.
Table 3: Victims of genocide by commune *
Commune

Number of inhabitants
1991*

Bwakira
Gishyita
Gisovu
Gitesi
Kivumu
Mabanza
Mwendo
Rutsiro
Rwamatamu

53.555
43.090
39.365
61.341
55.361
63.460
43.632
56.768
54.494

4.674
11.273
3.003
11.118
3.934
8.782
4472
941
10.853

8,7
26,1
7,6
18,1
7,1
13,8
10,2
1,6
20,0

471.066

59.050

12,5

Total***
*

Number of victims
data file

% of population
killed**

I did not have access to 1991 census data on ethnic affiliation per commune.

** To be lowered by 0.5% when population growth between 1991 and 1994 is taken into account.

3.3. Distribution of killing through time with untreated data
IBUKA found 59.050 victims of genocide in Kibuye Prefecture. The dates on which these
victims were killed are known for 25.716 persons, or in 43% of the cases. Table 4 lists the
number of victims by commune as well as the number of victims whose dates of death are
known. One can see that Gishyita, Rwamatamu and Gitesi count the largest number of
victims; this is directly related to the high number of Tutsi living in these communes before
the genocide. The number of victims with known dates of death is unequally distributed over
the communes. We know the dates of almost all victims in Rwamatamu and Kivumu
communes, but we know very few dates in Rutsiro commune and Gitesi commune. In the
case of Gitesi commune, dates of death are lacking for almost all victims.

10

Table 4: Victims per commune and date of death known
Commune

Number of victims
in the data file

Number of victims of whom
date of death is registered

in %

Bwakira
Gishyita
Gisovu
Gitesi
Kivumu
Mabanza
Mwendo
Rutsiro
Rwamatamu

4.674
11.273
3.003
11.118
3.934
8.782
4.472
941
10.853

1.381
4.221
1.640
135
3.265
3.566
2.028
199
9.284

29,5
37,4
54,6
0,01
82,9
40,6
45,3
21,1
85,5

Total

59.050

25.716

43,5

Figure 1: Victims of Genocide in Kibuye Prefecture
Number of victims per day, untreated data, n=25.716
4000

Number of Victims

3000

2000

1000

0
4
L-9
-JU
01 -94
N
-JU
25 N-94
-JU
19 -94
N
-JU
13 -94
N
-JU
07 -94
N
-JU
01 Y-94
A
-M 4
25 Y-9
A
-M 4
19 Y-9
A
-M
13 -94
AY
-M
07 Y-94
A
-M
01 R-94
P
-A
25 -94
PR
-A
19 R-94
-AP
13 -94
PR
-A
07 -94
PR
-A
01

Date

Figure 1 shows that the majority of Tutsi from Kibuye were killed in the first two weeks of
the genocide. Later on (section five) we will go into more detail about the possible bias in this
figure, but it is clear that the genocide in Kibuye was very intense in the first two weeks.

11

3.4. The weapons used
Table 5: Weapons used to kill Tutsi (all ages) with number of victims for each weapon

Weapon

Entire file
Number
Percentage

Dates of death known
Number
percentage

Machete
Club
Gun, rifle
Grenade
Drowned
Hoe, hack
Buried alive
Latrines
Spear
Burnt alive
Pick-axe
Stone1
Hanged
Sword
Starvation
Tractor
Other
Unknown or missing

31.117
9.779
8.706
1.058
847
444
442
437
421
401
337
31
100
79
23
12
636
4.020

52,8
16,6
14,7
1,8
1,4
0,8
0,7
0,7
0,7
0,7
0,6
0,2
0,2
0,1
0,0
0,0
1,1
6,8

13.272
4238
4575
609
486
328
340
150
209
226
192
84
35
50
15
7
197
659

51.6
16.5
17.8
2.4
1.9
1.3
1.3
0.6
0.8
0.9
0.7
0.3
0.1
0.2
0.1
0.0
0.8
2.6

Total

59.050

100

25.719

100

Table 5 is a good example of one of the data problems. The weapon that was used to kill the
victim is 'known' in 92% of the cases. The date of death however is only known for 43% of
the victims. It seems highly unlikely that a witness remembers the weapon that was used to kill
a person but cannot recall the date the event occurred. Of all the tables in this paper, table 5
best documents the brutality of the genocide. Most victims were hacked to death with
traditional weapons such as a machete or a club. We do notice however the importance of
firearms, being the gun or rifle and the grenade. From the 25.719 victims whose date of death
is registered in the data file, 20,2% were killed by a firearm (gun, rifle or grenade). In section 7,
we will analyze the weapons data in further detail.

12

4. ANALYSIS OF SURVIVAL CHANCES

During the fieldwork by IBUKA data on survivors in some communes were registered, but
not computerized. Because of budgetary limitations, I could only computerize data for two
communes. I decided to focus on the genocide in the center of Kibuye. Therefore, the data on
the survivors from Mabanza and Gitesi communes was computerized. The fact that
computer work for survivors had to be done, was used to re-code data of victims. The quality
of the data for Mabanza commune made this a worthwhile investment. Data on places of
death in Mabanza commune were computerized, but were not suitable for statistical analysis.
Names of places e.g. were spelled out instead of numbered. This section will deal with
Mabanza commune. Data and analysis for Gitesi are presented in appendix 2. Other
communes where re-entry of the data would be worthwhile (but has not yet been performed)
are Bwakira, Gishyita and Rwamatamu. Entry of survivor data and re-coding could specifically
teach us something about survival chances in these communes. I will now present an analysis
of survival chances for Mabanza commune. The underlying database is named ‘the new
database’ to distinguish it from the IBUKA nominal database.
Table 6: Death and survival in Mabanza commune
Table 6.1: Descriptive statistics
Total number of inhabitants registered in the 1991 census
Total number in 1994 (1991 figure * (1.03)2)
Total number of Tutsi registered in new database
Total number of victims
Total number of survivors
Number of people without entry
Percentage of 1994 inhabitants killed
Percentage of Tutsi in new database killed
Percentage of Tutsi in new database survived

63.460
67.325
10.785
9.257
1.477
51
13,7%
86,2%
13,8%

After re-entry of the data for Mabanza commune, I found 9.257 Tutsi killed and 1.477 Tutsi
who survived the genocide. This means that we have about 86% of the registered Tutsi in
Mabanza commune killed in the genocide. This is 13,7% of the resident population. The fact

13

that I have 500 victims more in the re-coded data file compared to the IBUKA publication is
because after that publication, more names of victims were found7.
Table 6.2: Places where the Tutsi residents of Mabanza were killed
Name or Location
In the cell of residence
In another cell within the sector
In another sector within the commune
In the Gatwaro Stadium
In Nyamagumba
In Bisesero8
In the Kivu Lake
At any another place
At an unknown place

Number of victims
1.905
836
329
3.359
677
300
18
645
1.188

Total
Total-unknown place

9.257
8.069

% of victims
20,5
9
3,5
36,2
7,3
3,2
0,2
10,0
12,8
100
87,1

Table 6.2 gives a geographic insight in the genocide in Mabanza commune. We can see that
about 20% of the victims (and 25% of those for which that place is known) were killed in
their own cell of residence. Since we have no figures for other communes, it is difficult to
compare. We know however that more than 1 out of 3 victims (36.2%) in Mabanza commune
were killed in the Gatwaro Stadium. Since this percentage is very high, we expect that in other
communes, more than 20% of the victims may have been killed in their own cell. In Figure 2,
we also observe between-sector variation in Mabanza commune: two sectors in Mabanza
commune, Kibirira and Nyagatovu, have more than 50% of all the victims residing in that
sector (respectively 287 and 499) killed in their own cell. Not surprisingly, these sectors have
very few residents who were killed in the Gatwaro Stadium. Kibingo sector has more then half
of its victims killed in Nyamagumba. The latter place is the name given by the perpetrators of
genocide to a hill in Kibingo sector. The name is the same name as that of a hill in Ruhengeri
Prefecture where Tutsi were killed in 1963-19649. Tutsi from the northern sectors (Kibingo
and neighboring sectors) of Mabanza commune did not gather at the communal office, took

7

Author’s communication with IBUKA, Kigali, August 2000.

8

'Bisesero' means the hills of Bisesero (plural).

9

Interview Mabanza commune, November 11, 2000.

14

refuge at ' Nyamagumba ' and resisted the Interahamwe and the soldiers from April 9 to April
12.
Figure 2 : Residence specific survival chances in Mabanza commune
40

percentage survived

30

20

10

0
1

2

3

4

5

6

7

9

11

12

13

14

SECTOR

Legend: Buhinga 1,Gacaca 2, Gihara 3, Gitwa 4, Kibirira 5, Kigeyo 6, Kibingo 7, Mushubati 9, Nyagatovu 11, Nyarugenge
12, Rubengera 13, Rukaragara 14.

Figure 2 shows Gihara sector to be the most deadly sector, only 5% of the resident Tutsi
population survived. Gihara sector has two cells where all Tutsi were killed. This is also the
case for three cells in Gitwa sector and one cell in Rukaragara sector. Sectors Mukura (8) and
Ngoma (10) are left out of the graph because very few Tutsi lived in these sectors. In two cells
in Mukura sector, all Tutsi were killed. There are cells with a lot of Tutsi where more then
90% of them were killed. On average, as we have seen, 86% of all Tutsi in Mabanza commune
were killed.10 A Tutsi had the highest survival chance, on average, when she or he resided in
Gacaca sector or Kibirira sector, respectively 24.5% and 30% of the Tutsi population in these
sectors survived. One cell in Kibingo sector, with a relatively high number of Tutsi, had the
highest survival percentage in Mabanza, namely 58%.

10

Among the 1.477 survivors, we counted 123 soldiers. The latter have been left out of the subsequent
analysis since they were most probably not present in Mabanza during the months of April, May and June
1994

15

One important question one can try to answer with the geographical data is whether or not
one’s survival chances depended on the place of refuge. For this purpose we constructed a
specific table (Table 6.3) with the overa ll survival percentage for each sector together with the
percentage of people who died in the Gatwaro Stadium.

11

Table 6.3: General survival chances and death in the Gatwaro Stadium
Sector

% survived
genocide

Gatwaro Stadium
% killed
number killed

Buhinga
Gacaca
Gihara
Gitwa
Kibirira
Kigeyo
Kibingo
Mushubati
Nyaragatovu
Nyarugenge
Rubengera
Rukaragara

8.6
24.5
5.0
15.7
30.0
7.4
22.3
14.9
12.8
7.6
10.4
6.5

62.8
54.5
66.9
55.4
26.2
67.1
2.3
47.9
3.6
48.8
59.1
8.5

615
192
477
149
124
753
22
444
32
185
331
35

Total

13.8

36.2

3.359

We then make two groups from the 12 sectors (Mukura and Ngoma have few victims to do
the analysis with and are left out). The first group includes those sectors of which less then
50% of its victims died in the Gatwaro Stadium, the second group includes the sectors of
which more than 50% of its victims died in the Gatwaro Stadium.
Group 1 : Buhinga, Gacaca, Gihara, Gitwa, Kigeyo and Rubengera
Group 2 : Kibirira, Kibingo, Mushubati, Nyaragatovu, Nyarugenge and Rukaragara
On average, 20.9% of the victims in the first group died in the Gatwaro Stadium, compared to
63.3% for the second group. We then compute survival chances for each group and test
whether they are significantly different for both groups in Table 6.4

11

I decided to pursue more detailed work on (differences in) survival chances thanks to an insightful
discussion with Susan Cook.

16

Table 6.4: Survival chances in two different groups
Killed in or survived
the genocide ?

Killed

% victims killed in Gatwaro St.
<50%
>50%
Group 1
Group 2

Total

number
percent

4.609
83.1

4.590
89.6

9.199
86.2

Survived number
percent

939
16.9

530
10.4

1.469
13.8

5.548
100

5.120
100

10.668

Total

number
percent

Chi-Square Tests of the equality of survival chances

Pearson Chi-Square
Continuity Correction
Likelihood Ratio
Fisher’s exact test
Linear-by-Linear Association
Number of valid cases

Value

df

96.898
96.345
98.256

1
1
1

96.889
10.668

1

Asymptotic
Sign.
(2-sided)
.000
.000
.000
.000

Exact
Sign.
(2-sided)

.000

Exact
Sign.
(1-sided)

.000

We find that Tutsi from sectors with few victims (in %) killed in the Gatwaro Stadium had a
higher overall survival chance. From the first group, 16.9% survived the genocide, whereas
only 10.4% from the second group survived. This difference is statistically highly significant,
as shown by the test statistics. In general, we do not find a higher survival percentage in the
sectors whose population sought refuge in ‘Nyamagumba‘ and fought the Interahamwe there.
We found 677 Tutsi who went to Nyamagumba, from two sectors in particular, Kibingo and
Rukaragara. Both belong to our second group, but the survival percentage was much higher
for Kibingo sector (22.3%) compared to Rukaragara (6.5%). Although few Tutsi from these
two sectors went to the Gatwaro Stadium, we cannot draw general conclusions on the
question whether fighting the Interhamwe increased one’s survival chances. It seems that only
escaping to Zaire or hiding really improved one’s chances. From this, one can derive a
behavioral guideline for persons targetted for extermination in the future : do not go where
the crowd goes.

17

Table 6.5: Occupations of the victims of Mabanza
Name of occupation
Pupil
Farmer
Teacher
Administrator
Businessman
Driver
Construction worker
Soldier
Doctor/nurse
Pastor/nun
Technician
Lawyer
All other occupations
Unknown occupation
Total

Number of victims
2.486
4.774
97
59
41
20
21
22
16
9
3
3
35
1.671

% of victims
26,8
51,5
1
0,5
0,4
0,2
0,2
0,2
0,17
0,1
0,03
0,03
0,3
18

9.257

100

Table 6.5 gives an overview of the occupations of the victims of Mabanza commune. Because
of the presence of a secondary school, this commune had more professionals compared to
other communes. However, more than 90% of the adults whose occupation was registered,
were farmers.

18

Figure 3: Distribution of killing through time, cases with known dates =3427*
800

Number of victims

600

400

200

0
7

9

11

13

15

17

19

21

23

25

27

29

Days of April 1994

About 150 Tutsi from Mabanza whose date of death is known, died after April 30th , 1994; they are not
included in the graph. In total, we have 3566 people in Mabanza whose date of death is known. Almost half
of all the Tutsi from Mabanza commune of which we have the date, died either on April 13th or on April 17th .
The former marks the attack on Nyamagumba and the latter the killing in the Parish of Kibuye and the
Gatwaro stadium. It is more likely that the dates of the victims who died in these major massacres are known
compared to victims who were killed in their houses, in the woods or on the hills : we have the dates of 50%
of the victims killed in either Parish, Stadium or Nyamagumba, compared to 30% of the victims killed outside
these major massacres. The major massacres are thus overrepresented in Figure 3, but the bias is not very
large. It is not the case that off-farm occupations are over represented in the dates.

Figure 4 : Age specific survival chances in Mabanza commune
25

20

percentage survived

*

15

10

5

0
0-10

10-20

20-30

30-40

AGE GROUP

40-50

>50

19

Since death and survival are discrete events in statistical and demographic analysis, we can
perform a logistical regression analysis with a survival dummy variable ‘0’ for death and ‘1’ for
survived. We use age, sex and sector of residence as explanatory variables. The binary choice
model, which is underlying the logistical regression, is derived in appendix 1. In the present
estimation , ‘the event occurring ‘ is the survival of a person during the genocide in Mabanza
commune. The following variables were included in the regression : Age, Age squared, Sex,
Sex*Age interaction, occupational dummy and sector dummies except for the sectors Mukura
and Ngoma (very few cases) and sector Buhinga which was used as the base-line. The number
of people included in the regression is 8.289 (number for which we have complete
information).
Table 7 : Regression results, survival dummy is dependent variable
Explanatory Variables
AGE
AGE 2
SEX
AGE*SEX
OFF-FARM
SECTOR dummies
GACACA
GIHARA
GITWA
KIBIRIRA
KIGEYO
KIBINGO
MUSHUBATI
NYAGATOVU
NYARUGENGE
RUBENGERA
RUKARAGATA
CONSTANT

B

S.E.

Sig

.0442 **
-.0011 **
.2312
.0110 **
.6589 **

.0096
.0001
.1447
.0051
.1391

.0000
.0000
.1101
.0311
.0000

1.2155 **
-.3640
.8141 **
1.6166 **
-.1128
1.2329 **
.7198 **
.5897 **
-.0324
.4016 **
-.0456
-2.9649 **

.1675
.2211
.1847
.1517
.1743
.1462
.1559
.1590
.2207
.1956
.2334
.1924

.0000
.0997
.0000
.0000
.5177
.0000
.0000
.0000
.8833
.0401
.8452
.0000

n=8.298
the effects are robust for other specifications of the logistical regression
** significant at the 5% level

20

Since the interpretation of the results requires some knowledge of the logistical procedure, we
shall concentrate on the essential findings. All the variables, except the sex variable and four
sector dummies, are significant (Sig < 0,05) meaning that these variables have a separate and
statistically relevant link with our dependent variable, survival during the genocide.

The effects of the age variable is not linear, but quadratic. This means that a victim’s
probability to survive the genocide increases with age at low levels of age, reaches a maximum
at the age of 20 (0.0442 / (2*0.0011)) and decreases again at high levels of age. Women in general
did not have a higher survival chance, but older women did. Tutsi having a non-farming occupation
had a higher survival chance than farmers and pupils. The latter result is puzzling and
interesting at the same time. Were they better informed then poorer people and chose the flee
earlier ? Where they physically more able to flee ? Did they have more cash to pay off
Interahamwe ? This regression exercise cannot give answers to these questions. All it can do is
show that one’s occupation did influence one’s survival chance in Mabanza. Compared to
Buhinga sector, Tutsi living in the sectors Gacaca, Kibirira, Kibingo had a relatively high
survival chance.

5. THE DISTRIBUTION OF KILLING THROUGH TIME AND SPACE:
ESTIMATION PROCEDURE AND RESULTS

5.1. Dealing with missing data
Since the number of dates of death known by commune is highly unequal, it is possible that
Figure 1 offers a biased perspective of the distribution of killing through time. The bias occurs
if the killings in the communes where we have little data differs substantially from the
communes where we have a lot of data. The shape of the curve in Figure 1 suggests that most
of the Tutsi of Kibuye were killed in the first weeks of April 1994. Further investigation has to
show whether missing data change this shape.
In order to present a graph for the entire sample, we have to devise an estimation procedure
that corrects for the lack of data in some of the communes. For the communes of Bwakira,
Kivumu, Mabanza, Mwendo, Rutsiro and Rwamatamu, a simple and straightforward

21

estimation procedure was used. We consider the victims of which the dates are known as a
representative sample of all the victims of that commune. This means we apply a weighing
procedure where the weight by commune is adjusted according to the number of dates known
for that commune. For example, for Mwendo commune, where we have almost half of the
dates, each date is weighted with 2,2. This procedure thus assumes that the distribution of the
dates of killing of the victims of which the date is known is representative for all the victims
residing in that commune.
For the communes Gishyita and Gisovu a slightly different procedure was used for three
reasons: firstly, we noticed that the date of death of the victims living in these communes
varied a lot according to the place where they died. In these communes, it made a difference
whether one was able to take refuge in the hills of Bisesero or not. The victims taking refuge
in Bisesero died later than the victims who were killed in other places. Secondly, especially in
Gishyita commune, where many Tutsi lived, the number of cases in which the date is known
is larger for the victims who did NOT die in Bisesero. This means that if we apply the same
weight for the whole commune of Gishyita, the estimation for the victims from Bisesero is
biased. We would then proceed as if the distribution of the dates of death of the victims at
Bisesero is the same as the distribution of the victims who died at other places12. Thirdly, since
it is important to know how many people died in Bisesero and when they died, we undertake a
separate estima tion for Bisesero. For these three reasons, the dates of victims from Gishyita
and Gisovu communes are weighted according to whether they died in Bisesero or not13.
For the commune of Gitesi, not only the date of death is not registered in the IBUKA
database, but also the place is not registered for more than 50% of the victims. At this point,
the IBUKA database is unreliable. For Gitesi commune therefore, another weighing
procedure was used. Table 8 gives an overview of the estimation procedure per commune.

12

The Tutsi living in Gishyita commune were not only killed in the hills of Bisesero, but also in the Mugonera
Hospital, in the Catholic Church of Mubuga and in other places throughout the commune.

13

When I use Bisesero, I mean the hills (plural) in Bisesero sector where the Tutsi of Gishyita, Gisovu, Gitesi
and Rwamatamu took refuge and defend themselves against the forces of genocide.

22

Table 8: Different weighing Factors
One weighing factor
Bwakira
Kivumu
Mabanza
Mwendo
Rutsiro
Rwamatamu

two weighing factors
Gishyita
Gisovu
depending on
whether they died
in Bisesero or not

three weighing factors
Gitesi
depending on when and
where they died
- Somewhere in the commune
between April 15 and 19
- in Bisesero
- in all other places

To repeat, in order to estimate the distribution of killing through time, we need to weigh the victims with known
dates of death. Depending on the commune the weighing factors differ. For most communes we only need one
weighing factor, namely one that multiplies each victim with known date to arrive at the total number of victims
in that commune.
For Gishyita and Gisovu, we use two weighing factors, depending on the place of death, namely in Bisesero or
not in Bisesero. We therefor needed to determine first how many people died in Bisesero, otherwise we cannot use
reliable weighing factors. The same is true for Gitesi commune. Since dates are lacking for Gitesi, we needed to
determine first the number of victims at several places in the commune before applying weighing factors in the
time distribution. This was done with the help of the re-coded data in appendix 2.
In order to keep the text readable, the estimation procedure for Bisesero is presented in the appendix 3. In order
to do our estimation for Bisesero, we also had to look at the number of Tutsi killed in Kibuye town. The latter
helps us to calculate the number for Bisesero. Without the figure for Bisesero (which is of course also an
estimation) one cannot determine the weights needed for the distribution through time. In order to calculate the
number of Tutsi who died in Bisesero, a number of assumptions had to be made.
On the basis of the IBUKA data, my own re-coding of the Mabanza and Gitesi data and a
number of assumptions specified in the appendix 3, I estimate the number of Tutsi who died
in Bisesero at 13.000, of which 6.800 from Gishyita, 1.333 from Gisovu, 3.700 from Gitesi,
700 from Rwamatamu and 400 from Mabanza commune.

23

5.2. Estimation results by day for Kibuye Prefecture

With the 13.000 estimate for Bisesero and the weighing procedure for each commune and for
Bisesero, we can present a table and a graph of the distribution of killing through time in
Kibuye prefecture. The method namely considers the victims in Bisesero of which we have
the dates of death as a representative sample of all the victims of Bisesero. To calculate these
victims is important because the Tutsi at Bisesero, on average, died later than the other Tutsi.
This accordingly gives you another distribution through time. Since we only have the dates of
2.500 victims who died in the hills of Bisesero, these people are giving a weighing factor 5,2 to
arrive at 13.000 (my best estimate).
According to the IBUKA data, 340 people died before April 6. From other sources, it is
known that several people were killed in the first months of 1994 in different attacks, but 340
in five days seems to be a high number. We were not able to verify this information, but it
may very well be that enumerators registered 'April 1' instead of 'May 1', meaning that we may
be dealing with registration errors. We find support for this reasoning when we notice that the
number of people registered as killed in the first week of May is very low. When one adds the
figures of April 1 to the figures for May 1 (and April 2-5 to May 2-5) we may get more correct
figures for killing in the first week of May. We realize the importance of knowing exactly how
many people were killed before April 6 for the study of the genocidal process, but from a
statistical point of view, these numbers make no big difference. That is why we decided to
include them in the presentation
Measured in numbers of people killed per day, the genocide in Kibuye reached its peak in the
middle of April. 75% of Kibuye’s Tutsi killed during the genocide were killed in the first few
weeks. After 50 days (by the end of May), the genocide was almost completed in Kibuye,
leaving 59.050 dead. This makes a daily average of 1.200 Tutsi killed. The first few weeks
however show a much larger number of people killed per day. Between April 7th and April
21 st , about 3.000 Tutsi (75% of 59.050 divided by 15 days) were killed on average every day
with April 13 th as peak showing an estimated 6.408 (10.8% of 59.050) people killed that day. A
day later an estimated 6.206 people were killed. The most deadly weekend during the genocide
in Kibuye Prefecture was probably April 16 and 17 with 4.800 and 5.300 people killed each
day respectively.
We stress that the figures presented in table 9 are estimations, not exact numbers.

24

Table 9 : Victims of genocide in Kibuye Prefecture (including Bisesero) per day
Date

April 1
April 2
April 3
April 4
April 5
April 6
April 7
April 8
April 9
April 10
April 11
April 12
April 13
April 14
April 15
April 16
April 17
April 18
April 19
April 20
April 21
April 22
April 23
April 24
April 25
April 26
April 27
April 28
April 29
April 30

Number of Tutsi
killed that day of
which date is known
134
7
7
175
17
11
63
193
379
655
952
2.398
3.683
3.492
2.468
2.525
1.629
1.194
312
681
146
100
84
203
207
76
43
870
59
81

Weighted number
of Tutsi killed
that day (applying
‘best estimate’)
134
7
7
175
17
20
160
445
1.133
1.503
2.135
4.238
6.408
6.206
4.416
4.839
5.296
3.488
1.392
1.506
300
208
163
772
501
142
93
1.099
101
179

Percentage
of Tutsi
killed that day

Cumulative
percentage

0,2
0,0
0,0
0,3
0,0
0,0
0,3
0,8
1,9
2,5
3,5
7,1
10,8
10,5
7,4
8,2
9,0
5,9
2,3
2,5
0,5
0,4
0,3
1,2
0,9
0,2
0,2
1,9
0,2
0,3

0,6
0,7
0,7
1,1
1,2
1,2
1,5
2,2
4,2
6,7
10,1
17,2
28,1
38,6
46,0
54,2
63,2
69,1
71,4
73,9
74,4
74,8
75,1
76,3
77,2
77,4
77,6
79,5
79,7
80,0

25

Table 9 : Continued
Date

Number of Tutsi
killed that day of
which date is known

May 1
May 2
May 3
May 4
May 5
May 6
May 7
May 8
May 9
May 10
May 11
May 12
May 13
May 14
May 15
May 16
May 17
May 18
May 19
May 20
May 21
May 22
May 23
May 24
May 25
May 26
May 27
May 28
May 29
May 30
May 31
All of June
and later
Total

67
136
39
32
51
22
9
36
28
103
37
57
782
227
203
69
24
53
20
97
26
14
14
13
71
21
0
33
15
34
13
771
25.716

Weighted number
of Tutsi killed
that day (applying
‘best estimate’)
212
547
146
86
173
62
37
157
66
342
100
205
3.654
1.029
902
231
111
209
54
361
93
40
52
44
291
69
0
93
46
131
65
2.120
59.050

Percentage
of Tutsi
killed that day
0,4
0,9
0,3
0,1
0,3
0,1
0,1
0,3
0,1
0,6
0,2
0,4
6,2
1,7
1,6
0,4
0,2
0,4
0,1
0,6
0,2
0,1
0,1
0,1
0,5
0,1
0
0,2
0,1
0,2
0,1
3,5

Cumulative
percentage
80,4
81,3
81,6
81,7
82,0
82,1
82,1
82,4
82,5
83,1
83,3
83,7
89,9
91,6
93,2
93,6
93,8
94,2
94,2
94,8
95,0
95,1
95,2
95,3
95,8
95,9
95,9
96,1
96,2
96,4
96,5
100,0
100

26

Figure 5 : Victims of Genocide in Kibuye Prefecture
number murdered per day , treated data, n=59.050

7000
6000

Number of victims

5000

4000

3000
2000

1000
0
4
L-9
-JU
01 N-94
-JU
25 -94
N
-JU
19 N-94
-JU
13 N-94
-JU
07 -94
N
-JU
01 -94
AY
-M 4
25 Y-9
A
-M
19 -94
AY
-M
13 Y-94
A
-M
07 Y-94
A
-M
01 R-94
-AP
25 R-94
P
-A
19 R-94
P
-A
13 R-94
-AP
07 R-94
P
-A
01

Date of death

From the data, we learn that killing started immediately, especially in the communes of
Rutsiro, Mabanza, Rwamatamu and Gishyita.. The graphs that were published in the
Nominative Dictionary of IBUKA are approximately correct, but do not do apply the
weighing factors. The main difference between the graphs in the dictionary and the present
graphs is that the former underrepresent the number of Tutsi killed on May 13th. While figure
5 clearly shows the first two weeks of the genocide in Kibuye as the period in which most
Tutsi were killed, it also details two other devastating days, namely April 28 th and May 13 th.
The latter date is known in the survivor community as the date on which Interahamwe from
all over Rwanda assembled in Bisesero to kill the Tutsi there who succeed in resisting the
genocide. For several weeks, the Tutsi at Bisesero mounted a formidable resistance against the
forces of genocide. Using the steep hills, gathering and throwing stones and implementing a
method of strategic fighting (Kwivunga sha), they succeed in staying alive. Against attackers
with overwhelming firepower, they had no chance anymore. Figure 6 in particular shows the
extra-ordinary death path of the Bisesero Tutsi. The second peak, April 28 and 29 are the days
of the massacre at Kiziga Hill in Rwamatamu commune.

27

Figure 6 : Victims of Genocide at Bisesero
Number of Tutsi murdered per day, treated data, n=13.000
4000

Number of victims

3000

2000

1000

0
94
N-JU
27 -94
N
-JU
21 -94
N
-JU 4
15 N-9
-JU
08 N-94
-JU
02 Y-94
A
-M
26 Y-94
A
-M
20 -94
AY
-M
14 Y-94
A
-M
08 -94
AY
-M
02 -94
R
-AP
26 -94
PR
-A
20 -94
R
-AP
14 -94
R
-AP
08 -94
R
-AP
01

Date of Death

6. EXTRAPOLATION FOR RWANDA

Extrapolating the estimation results for Kibuye and combining this with numbers of killed
Tutsi put forward in the literature, we can make an approximation of the killing in the entire
country. I want to stress that the following table is only a tentative approximation and should
not be regarded as a careful count. I used my own calculations for Kibuye and calculated all
the other figures from the Human Rights Watch publication "Leave none to tell the story ". I
added the numbers in that publication for the major massacres that occurred during the
indicated days in Gikongoro and in Butare. The figures are only rough estimations. I also took
the 500.000 number from HRW as the estimated number of victims. From census data we
know that Kibuye, Gikongoro and Butare taken together have about half of the Tutsi
population before the genocide. Since it was difficult to get estimates for massacres in the
other prefectures, I extrapolated the number killed in these three prefectures for the whole of
Rwanda.

28

Table 10 : Estimation (by extrapolation) of killing over time*
Kibuye

Butare

April 6-10
April 11-14
April 15-18
April 19-22
April 23-26
April 27-30

3.000
20.000
20.000
2.000
2.000
2.000

20.000
35.000
35.000
10.000

Total April
May-June
Total

49.000
11.000
60.000

100.000
20.000
120.000

Gikongoro

3 pref.

Rwanda **

10.000
25.000
20.000
15.000

13.000
45.000
60.000
52.000
37.000
12.000

25.000
87.000
115.000
100.000
70.000
23.000

70.000
10.000
80.000

219.000
41.000
260.000

420.000
80.000
500.000

*

I repeat that this table should not be read as if I do not think that there were no killings in Butare Prefecture
before April 15. I am only interested in gross figures, not in exact calculations here.
** Extrapolation is done starting from the sum of the three prefectures and using the HRW estimate of 500.000
as a baseline. I rounded off to thousands, avoiding the use of hundreds.

Table 10 suggests that between April 15th and April 18th, more than 25.000 were killed every day
in Rwanda. This can be traced back to the major massacres in that period : For Kibuye in the
Gatwaro Stadium and the Parish of Kibuye, for Butare in Cyahinda church complex in the
commune of Nyakizu and for Gikongoro in Kibeho church. The figure of 25.000 is my
approximation of the average number of people killed between April 11 th and April 22nd :
87.000 + 115.000 + 100.000 ) / 12 ≅ 25.000

29

7. ANALYSIS OF THE WEAPONS USED IN THE KILLING PROCESS

7.1 Focus on age
Figure 7 : Murdered with a firearm
Realized probability by age

Percentage murdered by fire-arm

.4

.3

.2

.1

0.0
0

10

20

30

40

50

60

70

80

AGE

The dots in figure 7 present the percentage of victims of each age that were killed by a firearm.
Statistical software allows to draw a line (best fitted quadratic expression) through the cloud of
these dots. A clear pattern emerges: few children and elderly people were killed by firearms
and (relatively) many young adults. 20 to 25% of all victims in their early twenties were killed
by firearms whereas for victims in their late fifties the figure was 10%. In a regression analysis
(table 12) we will see that this age-effect is statistically significant.

7.2 A word on occupation and gender
From the IBUKA data file, we learn that 30.843 of the registered victims were farmers, 15.494
were pupils, 2.003 had an occupation outside farming and 10.710 victims were registered with
unknown occupation. These figures confirm the basic characteristic of Rwandan society in
general and in Kibuye in specific, namely the fact that it is a very rural society. Only a small
minority of the working population did not farm. This is basically true for Hutu as well as for
Tutsi. The former may have been more present in public administration and the latter more in
commerce, but more than 90% of the people of both groups were farmers. Whereas in the

30

past, Tutsi used to be more involved in cattle breeding than Hutu, in the early nineties, there
was no outspoken ethnic specialization in either agriculture or cattle breeding anymore 14.
Depending on wealth, Hutu as well as Tutsi held cattle and grew crops. The IBUKA data file
does not mention the landholdings or the number of cattle of the victims and makes only a
distinction between farming and non-farming. 15 In table 11, I show the occupations of the
victims, together with the weapons, that were used to kill the victims. I only use data on
victims for whom the date of death, the occupation and the weapon is known. The pupils are
also taken apart because this category is determined by age, which we already treated in the
previous section. From a total of 21.293 victims for whom we have information on
occupations, 772 (5% of the adult victim population) had a non-farming occupation. We
notice that 30% of them were killed with a firearm. This percentage is significantly higher than
that for farmers and pupils.
Table 11 : Occupations and weapons used
Occupation
Farmers
Non-farmers
Pupils

number
13.589
772
6.932

killed by firearm

%

1902
231
1039

14
30
15

More men than women were killed during the genocide in Kibuye, 30.528 men and 28.471
women (with 51 unknown) according to the IBUKA file. As far as the weapons are
concerned, the difference between men and women is less outspoken compared to the
differences by the age and occupations. Slightly more men than women (in absolute figures as
16
well as in percentages) were killed by bullets.

14

Historically, it may be more correct to say that people involved in cattle breeding were, over time,
considered Tutsi. I refer to the literature on this important and complex topic, e.g. Newbury (1988), De
Lame (1996).

15

More research is needed to investigate in what respect real or perceived ethnic labor specialization and
inequalities in landholdings and cattle played a role in the genocide. Sources are f.e. A. Desforges (1999),
P.Uvin (1998), C.André and J.-Ph. Platteau (1998).

16

In the regression (table 12) gender appears to be significant in interaction with other variables.

31

7.3. Inequality until death ? A regression analysis of the weapons used to kill
The scholarly community involved in genocide research agrees on the use of traditional
weapons as the main killing instruments in the Rwandan genocide. Less agreement is reached
on the importance of firearms during the genocide. In this section, I shall test statistically
whether or not firearms were used indiscriminately of the victim’s person. In the previous
sections, we discussed five variables that each may have had an effect on the weapon used to
kill an individual during the genocide in Kibuye Prefecture. These variables were the date of
death, the commune of residence, the gender of a person, the occupation and the age of this
person. A statistical procedure, logistical regression analysis, allows one to trace the specific
effect of each of these variables on the weapon that was used to kill the person in question.
The logistic analysis is performed for those cases in which the dates of death were known,
namely 43% of entire sample. We assign the value ‘0’ in the event the victim was killed with a
traditional weapon and the value ‘1’in the event the victim was killed with a firearm. The latter
variable is either a bullet or a grenade. Our fifth variable is now a binary variable, (0 or 1)
which is necessary to perform a logistical regression. The question we want to answer here is
the following: are the independent variables able to explain the way in which the victim was
killed, namely with a firearm or with a traditional weapon? The binary choice model is derived
in appendix 1. In the database, we have 20.419 victims between April 6 and June 30 for whom
we have complete information, meaning no missing data for any of the variables to be used in
the regression analysis. It first needs to be said that I did not have a lot of choice on the
question which variables to include in the regression. Apart from the mentioned variables, I
also have the name of the place of death of the victim. I could not include the place of death
since the data contain hundreds of places where people were killed and these places, as we
have seen earlier in the paper, were not coded. We did include dummies for communes.
Remember that we left out Rutsiro and Gitesi communes because of lack of data. Bwakira
commune is used as base-line commune and the effects of other communes are thus fixed
effects relative to Bwakira commune. 17 The following variables were included in the analysis:
age, age squared, sex (female=1), sex*age interaction, occupational dummy (off-farm work is
1), number of days after April 6, number of days after April 6 squared, sex*number of days
after April 6 interaction, sex*number of days after April 6 squared interaction and dummies
for communes.

17

One can use any other commune as base line.

32

Table 12 : Regression results, dependent variable weapon used 18
Variable
AGE
AGE 2
SEX
SEX*AGE
OFF-FARM
DAYS after April 6
DAYS after April 6 sq.
SEX*DAYS after April 6
SEX*DAYS after April 6 sq.
COMMUNE dummies
GISOVU
GISHYATA
KIVUMU
MABANZA
MWENDO
RWAMATAMU
CONSTANT

B

S.E.

Sig

.0345 **
-.0005 **
-.1342
-.0067 **
.5200 **
.1147 **
-.0014 **
.0491 **
-.0015 **

.0041
.0000
.1285
.0021
.0801
.0069
.0001
.0135
.0003

.0000
.0000
.2963
.0016
.0000
.0000
.0000
.0003
.0000

1.2435 **
1.2841 **
1.5782 **
3.2234 **
.6327 **
2.4434 **
-4.9339 **

.1753
.1642
.1660
.1604
.1884
.1576
.1838

.0000
.0000
.0000
.0000
.0008
.0000
.0000

n= 20.419
the effects are robust for other specifications of the logistical regression
** significant at the 5% level

The probability of the victim in Kibuye Prefecture to be killed by a traditional weapon or a
firearm depended on the person’s age, his occupation, the place of residence before the
genocide, on the number of days the after April 6 the person was killed and on interaction
effects of these variables with the sex. All the variables determine the ‘choice’ the killer made
between a traditional arm or a firearm.
The effects of the age and days-after variables are not linear, but quadratic. This means that a
victim’s probability to be killed with a firearm increases with age at low levels of age; it reaches
a maximum at age 34,5 (0.0345 /(2*0.0005)) and decreases again at high levels of age. This is
also true for the days-after-April-6 variable : the probability to be killed with a firearm

18

Gitesi and Rutsiro communes are left out because of lack of data.

33

increases as the days move away from April 6, but decreases again towards the end of the
genocide. The significance of the results on age and number of days after April 6 in fact
confirms that middle-aged people (20 to 40 years of age) were more likely to be killed with
firearms and that the genocide reached its most intensive period (meaning here the period in
which the probability to be killed by a firearm was highest) after a few weeks.
Since the sex variable by itself is not significant, women in general did not have a smaller chance to be
killed by a firearm. The significance of sex (or gender for that matter) reveals itself in
combination with other variables (see below). The only variable capturing (part of) the socioeconomic situation of the victim is the occupation variable. In the regression, victims with a
non-farming occupation (a small minority in Rwanda) have the value ‘ 1’ on the occupation
dummy. The regression shows that non-farmers had a higher probability to be killed with a
firearm compared to farmers and pupils.
Both the age and days-after-April-6 effects interact with the sex variable. A woman's
probability to be killed by a firearm, compared to a man, decreases with her age (sex-age
interaction negative and significant). For the days-after-April-6 variable, there is an additional
positive effect for women at the beginning of the genocide and a negative effect towards the
end of the genocide. This means that the probability of women to be killed by a firearm was
higher, compared to men, a few weeks into the genocide (the period that I label 'the most
intensive period'), and lower, compared to men, towards the end of the genocide. Although all
these effects are significant (that is the reason why I present them), one should not
overestimate their magnitude. The effects, especially the magnitude of the squared effects, are
small. The effects of the communes of residence are also significant and on top of that they
are very strong. This means that the probability of being killed by a firearm was high for
residents of Rwamatamu and Mabanza communes compared to Bwakira. The regression does
not tell the reasons why these communal dummies are significant and strong. Apart from
Mabanza (see below), this is a question open for research. The effects are not due to the
enumerators, since 15 to 20 enumerators were doing the data collection in one commune and
the commune where one can speak of commune-wide bad data collection is left out of the
regression.

34

In section four, I researched Mabanza commune in more detail. The reason why the
probability of a Tutsi from Mabanza commune to be killed by a firearm was the highest in the
entire prefecture was that many victims from Mabanza were killed in the Gatwaro Stadium. In
that stadium, the perpetrators of genocide installed machineguns in the tribunes and fired at
the crowd who were locked in the Stadium. Reasons for high prevalence of killing with
firearms in a commune might be the size of the local stock of firearms, the presence of armyunits in a commune, the participation of militia armed with firearms from another prefecture,
to name but a few. A reason for the high prevalence of killings with traditional weapons in a
commune might be the distribution of machetes to the local population. Detailed on-site
investigations are necessary to confirm or refute this hypothesis. From the statistics however,
it is clear that significant differences between the use of firearms existed between the
communes.

19

On a more general level, the results of the logistical regression show the organized character
of the genocide. If age, occupation, commune of residence, the number of days after April 6th
and the sex (interaction) variables all prove significant to explain the weapon used, then
genocide was all but a random process. Indeed, one can read this regression analysis as
statistical evidence of the organized character of the genocide in Kibuye prefecture.
As reasons why young Tutsi men who were working in the modern sector of the economy
had a higher probability of being killed with bullets then other victims, we can only find one:
in a number of cases, killers had to economize on bullets and thus used firearms only against
the people in the village who could mount resistance. These people in turn were most likely
young men aged 20 to 40 with a respected status in the commune or sector. That may have
been one reason why they were killed by bullets. But again, the reasons behind this statistically
significant effect cannot be derived from the analysis, but have to be found by other means.
The present result only tells us that there is a significant effect and gives us some insight in the
genocide, but cannot give waterproof explanations.
19

There are not many sources with which to compare the results from this logistical regression. An
observation from the previously mentioned French officer, Lt. Colonel Stabenreth, does give some
indications “…From his investigations, he established that the Tutsi refugees who had sought shelter at the stadium had been
attacked by soldiers and militia who had shot until they had run out of ammunition…”. As I said before, most of the
refugees at Gatwaro Stadium came from Mabanza commune and the officer concludes from his
investigation that the Tutsi in the stadium were killed with firearms. My statistical result also corresponds to
the observations of eyewitness Doctor Blam who heard the intervention of firearms and grenades during
the massacre in the Stadium.

35

8. REPORTS AND
PREFECTURE

DOCUMENTATION

OF

GENOCIDE

IN

KIBUYE

8.1. The content of reports and letters written by the prefect
In this section of the paper, I use the words of the perpetrators of genocide without giving
much comment. I quote from the reports in order to show what they themselves wrote about
the crime. In the next section, I discuss the use of the words in italics. The following reports
and letters were obtained from Human Rights Watch : (own numbering of the document in
brackets)
-

A detailed report by the Prefect of Kibuye of the day to day events in all the communes
between April 6 and April 10 dated on April 11. (k1)
The report of the meeting of the security committee of Kibuye Prefecture on April 11,
1994 (k2)
A letter of the Préfet to all the Burgomasters concerning the self-defense program of the
population, on April 30th, 1994 (k3)
A letter from the Préfet of Kibuye Prefecture to the Minister of the Interior and
Communal Development, on May 5, 1994 (k4)
A letter from the Minister of Interior and Communal Development to the military
commander in Gisenyi asking to send troops to Kibuye Prefecture to support an
operation in the sector of Bisesero, Commune of Gishyita, dated June 18, 1994 (k5)
A undated letter addressed to the Minster of Interior and Communal Development
concerning the management of Gisovu Commune. (k6)

In his first report during the genocide (k1) the Préfet of Kibuye , Dr. Clément Kayishema,
gives a detailed overview of the situation in the communes and of his measures taken during
the first days of the genocide. I quote from this report 20: On April 7, Kibuye was scheduled to
receive a visit by the Special representative of the UN, Dr Jacques Roger Booh-Booh, but he
did not come because of the events in the country. The same day the Préfet reports rumors of
interethnic clashes, especially in the communes of Gishyita, Rwamatamu and Mabanza. On the
night of April 7 to April 8, the Préfet reports trouble in all the communes of Kibuye Prefecture,
but he adds that it are still rumors and that the local officials are trying to calm down the
population. One rumor for example is that a group of people is going to cut the telephone lines
of the Mugonero hospital in Gishyita commune. In Kivumu commune, the Préfet reports that
two children of a Tutsi family are killed and the women wounded. On April 8, the Préfet
20

Since the report contains several pages, I am only taking out the information that seems most important to
me.

36

reports that Hutu in Gitesi commune want to organize security walks at night. The Préfet also
reports the decisions of the security committee of the prefecture on the morning of April 8. The
names of the members of this security committee meeting are mentioned in the document.
They decide that
- the fuel of the Petrorwanda fuel station is reserved for security,
-

public meetings are forbidden until further notice,
the permission to circulate for hotel personnel and visitors should be analyzed case by case
and
a number of vehicles are allocated to certain communes and communal police.

The report continues the overview of the events in the communes in the afternoon of April 8.
Several refugees from Kayove arrive in Rutsiro and Mabanza communes. There is a rumor that
the Tutsi of Bisesero are going to attack the Hutu of Bisesero. Tutsi would hide arms at the
Hospital. In the night of April 8 to 9, about 300 refugees are signaled in the communal Bureau
of Rwamatamu together with their cows and goats. On April 9, the security committee of the
prefecture meets again and decides that
-

no shooting between 6 p.m. and 6 a.m.

-

authorization needed from the Burgomaster to travel to another commune
closing of the markets near the Kivu Lake
Allocate 500.000 Rwandan Francs from account 90.002 at the Commercial Bank of
Rwanda to buy urgently needed fuel.

The report continues with the overview of the situation in the communes with a great deal of
attention to Gishyita commune. The Burgomaster of that commune asks for urgent
intervention of the gendarmes. The Tutsi of Mabanza commune have gathered at the
communal bureau. In Mwendo commune, 11 criminals are arrested but taken to Kibuye
center because the prosecutor of Birambo does not dare to do investigations. The
burgomaster of Gisovu reports he has found two motorbikes and that some people are
carrying guns. Refugees from Muko commune in Gikongoro arrive in Mwendo commune.
On April 10, the Préfet reports that the Hutu of Gishyita are discouraged because the Tutsi seem
to have firearms. In Mabanza commune, the houses of the Tutsi are burning in several sectors
and almost all Tutsi have gathered in at the communal bureau. A schoolteacher was killed by a
grenade. In Mwendo commune, several houses of Tutsi are burning in Gashari sector but the
gendarmes have intervened. Tension is rising and the refugees from Muko are relocated to the

37

primary school. In Rutsiro commune one has started to kill the cows and two people have
died. The UNAMIR soldiers have left Kibuye on April 10 and relocated at Butare. The only
expatriates who remain in Kibuye are religious people and some health workers. A list with
their names is attached to the report.
Document (k2) is a short listing of all events on April 11. It relates the presence of 1000
refugees in Mwendo commune, mainly originating from Muko commune and 3000 refugees in
Mabanza commune who are threatened by people from Rutsiro commune. There is a rumor that
gendarmes are going to rob the Commercial Bank of Rwanda in Kibuye (Gitesi commune).
The council tries to find solutions for the short supply of fuel and vehicles. They buy fuel, the
Préfet decided to empty the MRND youth fund. The communes have requested additional
policemen, firearms and ammunition, but the Préfet reports that it is up to the communes
themselves to provide this. The measures of the April 9 meeting are repeated adding that a
mobilization campaign in the framework of pacification will be undertaken. After this meeting, which
lasted from 10 a.m. to 1 p.m., the members of the committee formed two groups to visit
several communes. This activity was labeled a ‘visit to the field’ in the report (k2).
Document (k3) informs the burgomasters of the decision by the government of Rwanda to
install a civilian self-defense program. The government wants the population to organize controls,
set up roadblocks and look for infiltration by the Inkotanyi. In that respect, the burgomasters
should recruit persons who will receive training from military reserve personnel. These
persons should be physically and morally apt, of good conduct and with a certain credibility in
the eyes of the population.
Letter (k4) addresses the security report for the April 11 to April 30 period to the Interior
Minister21. The Préfet, author of this letter, states that that period was characterized by interethnic killings in almost the whole prefecture, a massive displacement of people and livestock,
looting and destruction of houses and of public buildings. He continues that calm is gradually
returning from April 25 onwards and that people take up their normal activities again. There
is, the Préfet writes, a small area of insecurity in the Bisesero area bordering Gishyita and Gisovu
communes. In the remainder of the letter, the Préfet reports seven disquieting facts among
which, the absence of funds, the climate of vengeance between people and the arguments over
the distribution of looted goods, abandoned fields and houses by the population.

21

The author is not in the possession of that report.

38

Document (k5) is a letter addressed to the military commander in Gisenyi by the minister of
the interior. By governmental decision, the minister orders the commander to supply troops to
Kibuye for a search operation in Bisesero (in French “operation de ratissage”). This operation
should be terminated at the latest by June 20 .
Document (k6) is a complaint against the burgomaster of Gisovu commune, Ndimbati who is
accused of mismanagement of his commune. The letter is not dated but features a receive stamp
dated July 8, 1994. The burgomaster is accused of looting furniture, raiding cattle, stealing
money from the local bank, stealing 73 metal sheets from a medical center and killing 7
persons including 2 soldiers. These seven persons were residents of Gishyita commune and as
a result of the killing, there is increasing tension between the people of Gisovu and Gishyita
communes. That is why the author of the letter proposes to replace the burgomaster.

8.2. The (un)reported facts and the use of population statistics
As in Nazi-Germany, the word ‘genocide’ was never used to describe the killings. The Préfet
used ‘the war‘, ‘inter-ethnic rivalries’, ‘security measures’ to describe the acts of genocide.
When the Préfet gives an overview of the ‘events’ in Kibuye in the first week of the genocide
(April 6- 10) (k1) and in his report of April 11 (k2), he mentions the burning of houses, the
displacement of people, the presence of refugees and sometimes he also reports the killing of
a few people. Compared to the statistics presented in the first part of this paper, the figures of
murdered people in the reports of the Préfet are ridiculous. One person killed in Kivumu, 3
persons killed in Mabanza, he writes…whereas in fact more than 5000 Tutsi were killed at the
end of the first week.
The Préfets’ statement that on April 7 especially the communes of Mabanza, Rwamatamu and
Gishyata were ‘hit by interethnic clashes ‘ seem to correspond, at least as far as the date and
the communes are concerned (the description of the ongoing events is something else), with
the early killings in these communes derived from the IBUKA data file. This means that from
two independent sources, we can state that the genocide started in all communes on April 7,
but was especially fast in the three communes mentioned. These reports from the Préfet of
Kibuye, Dr. Kayishema Clément, are official documents of a leading official in Rwanda. The
Préfet was the highest authority in his prefecture but did not have close contact with the
population when compared to the burgomasters. Notwithstanding the language used by the
Préfet, we can derive from his reports that the genocide in Kibuye Prefecture started
immediately after the death of president Habyarimana. The population of Kibuye did not wait

39

one day but started killing on April 7. The speed of the genocide in Kibuye Prefecture is
incredibly high, as documented by the IBUKA figures, which is an indication that many Hutu
joined the killing campaign. Thus the reports of the Préfet only give a glimpse of the real
facts. It was virtually impossible for the Préfet to make a day to day head count of all the
victims in his prefecture, but as Alison Desforges writes
“Administrative officials recorded changes in the population extremely carefully before the genocide, nothing births,
deaths, and movement into and out of the commune on a monthly as well as a quarterly basis. With this data,
officials knew how many Tutsi, whether male or female, adult or child, lived in each administrative unit,
information useful in any attempt to eliminate them. Prefect Kayishema was so concerned about the accuracy of
this data that he took time in early May to review census data submitted by burgomasters for the last quarter of
1993. He found errors in at least two of the reports, that of Mabanza, which recorded the increase in female
Tutsi as fifty-two instead of fifty-three, and that of Rwamatamu where an error of seven was made in accounting
for the male Tutsi population and an error of six was made in recording that of female Tutsi.“22

This makes it clear that population statistics in Kibuye Prefecture and especially the accurate
reporting system of demographic changes that existed in Rwanda before the genocide became
a deadly tool in the hands of the prefect. Desforges writes that during the genocide,
administrators gave orders to register all displaced persons immediately. (p. 240). She also
refers to documents where the burgomaster of Bwakira commune asked councilors to submit
a list of household heads who had died, the number of people in the household killed and the
number who had fled (p. 240-241). The use of statistics in the pursuit of genocide is not
unique to Rwanda. In a recent paper, W. Seltzer shows the intricate involvement of statistics
and statistical systems in the planning and advancement of the genocide against the Jews of
Europe in Nazi-Germany, Poland, France, the Netherlands and Norway23. What is highly
disturbing in the Rwandan case is the fact that the international community did not take action
against the registration of ethnicity on the identity cards of Rwandans when this could have
saved thousands of lives. According to Desforges, influential donors overlooked the
systematic discrimination against Tutsi before the genocide and did not insist on the
elimination of ethnic affiliation on the cards that served as death warrants for many Tutsi in
1994 (p. 17) Indeed, as the statistics helped the Préfet to pursue genocide as accurately as
possible, the identity cards (one of the visible outputs of a statistical system) helped the
Interahamwe to sort out Tutsi from a crowd of people.

22

Desforges, A., Leave none to tell the story, Human Rights Watch, 1999, p. 239.

23

Seltzer, W., Population Statistics, the Holocaust and the Nuremberg Trials, Population and Development
Review 24 (3), 1998.

40

The Préfet did not only trace Tutsi, he organized the massacres himself and chose a
euphemistic language to talk about them. I did not find a word about the massacre in Gatwaro
Stadium for example. However, this does not mean that we should totally discredit his reports.
When, in document (k4) the Préfet writes that “calm” gradually returned from April 25
onwards, he means that most Tutsi from Kibuye had been killed by then. From the end of
April, the Préfet tried to restore ‘normality’ in the prefecture. As if nothing had happened,
children had to go back to school and adults back to work.
The group of Tutsi that defended themselves in Bisesero where among the few Tutsi that
were still alive. The Préfet describes this is his letter (k4) as a “small area of insecurity”.
Survivors of the massacres at Bisesero told African Rights researchers they have seen the
Préfet several times at Bisesero. He was considered one of the leading organizers of the
genocide24. The Préfet, together with Obed Ruzindana (rich businessmen), Alfred Musema
(director of the tea factory in Gisovu) and the burgomasters of Gisovu and Gishyata drove
around in their cars and in the trucks of the tea factory to deliver interahamwe and soldiers to
the massacre sites25.
The answer of the Préfet that the communes themselves have to provide firearms and
26
ammunition can be understood in the context of the national policy of self-sufficiency. This
policy did not only mean that each household should produce their own food, but also that
each commune should take care of itself, making that each prefecture and in the end Rwanda
can care for itself. As it was national policy before the genocide that Rwanda should solve its
own problems, or find Rwandan solutions for its problems, the genocide itself was interpreted
as a Rwandan solution to its own problem. Every cell, sector, commune or prefecture should
solve its own problem; that is how political leaders tried to transmit their messages to local
officials. Only when resistance was too great and local officials could not manage the situation,
the higher level would intervene. This was the case in Bisesero where the burgomaster of
Gisovu sent someone with his car to Cyangugu to call John Yusufu Munyakazi to Bisesero
because the interahamwe from Bisesero could not ‘do the job‘ by themselves. It was with the
help of this relentless mass murderer that the Tutsi of Bisesero lost the unequal battle.
(African Rights pp. 29-30).

24

Resisting Genocide, p.18 and 28. I do not have access to testimonies by survivors in the course of judicial
procedures at the International Court in Arusha, where the Préfet Clément Kayishema is currently
imprisoned. Such testimonies could shed light on the exact role of the Préfet during the genocide in
Kibuye.

25

Ibidem, and p. 51.

26

I refer to another paper of mine where I discuss the development and peasant ideology of the Habyarimana
regime, paper published in the November 2000 issue of the Journal of Genocide Research.

41

9. CONCLUSIONS

In this paper, I have presented a statistical analysis of the genocide in Kibuye Prefecture,
Rwanda. I used the data file collected by IBUKA and assumed that the data in this file are
(fairly) reliable. I discussed the numerous data problems of this file. For two communes,
Mabanza and Gitesi, I re-coded the data on victims to use them for statistical purposes and
also computerized data on survivors. From the regression analysis, I derive tha t middle aged
Tutsi had a better survival chance then the very young and the very old, especially when they
did not join the crowd at the Gatwaro Stadium. Fleeing gave the highest probability to survive
the onslaught. The effects of several sector dummies in the regression are very significant,
indicating that one’s survival chance indeed depended on the sector of residence. The reason
for this is that the Tutsi population of half of the sectors of Mabanza went to the Stadium
whereas the Tutsi of the other sectors did not (or not so much) go to the Stadium.
In order to calculate the distribution of killing through time, I first had to estimate the number
of people killed in the hills of Bisesero. This was only possible with the help of some
assumptions to overcome gaps in the data file, especially for the commune of Gitesi. My best
estimate for the number of Tutsi killed in the hills of Bisesero is 13.000. Using this number, I
then calculated the number of Tutsi killed every day in the entire prefecture. The results from
this calculation confirm what the scholarly community already knows, namely that the speed
of the genocide was incredibly high. 75% of the Tutsi of Kibuye were already killed by the 22nd
of April, meaning a daily average of 3000 for the first two weeks of the genocide. And this is
only for one prefecture.
The paper also features a statistical analysis of the weapons used in the genocide. Half of
Kibuye’s murdered Tutsi population was murdered with a machete, one in six with a club and
one in six with a firearm. The place of residence before the genocide, the age and gender of
the victim and the victim’s occupation all proved statistically significant in a regression
determining the weapon (a binary variable) used to kill a particular victim. Combined with the
speed of the killing, these results could only have been obtained, we believe, with broad
participation of Hutu peasants in the genocide. As Minna Schrag, a former prosecutor with
the International War Crimes Tribunal for Yugoslavia, observed at a 1997 conference on the
use of quantitative data and analysis “data can help us tell the story of the crime “. (from W.
Seltzer, p. 543). From the data, I conclude that speed, accuracy and resistance are the
keywords in the story of the genocide in Kibuye.

42

Appendix 1 : the binary choice model
The binary choice model is derived as follows27 :
The probability of having y = 1 instead of zero can be written as
Prob { y = 1} = G( x k , β)

(1.1)

where G is a functional form containing the vectors x and β.
Usually, the functional form is restricted to
G( x k , β) = F( x′kβ)

(1.2)

where F is a cumulative distribution function
It is possible to derive a binary choice model using a latent variable presentation of the model
y ∗ = ∑ k β k xk + ε

(1.3)

where y* is an unobserved latent variable and ε symmetrically distributed with zero mean and
cumulative distribution function F(ε). What we observe is a dummy variable y, a realization of
a binomial process, defined by
y=1

if

y ∗ > 0 and

y = 0 otherwise

therefore

(
)
= Pr ob(ε > −∑ β x )
= 1 − F (− ∑ β x )

Pr ob( y = 1) = Pr ob ∑k β k x k + ε > 0
k

k

k

k

k

(1.4)

k

The specific functional form of F depends on the assumptions that one makes concerning the
distribution of ε. In case of the binary logit model, we assume that ε follows a logistic
distribution. This distribution is similar to the standard normal distribution but instead of a
variance of 1 it has a variance of π 2/3.

27

We follow the approach taken in Liao, T.F, Interpreting probability models, Quantitative applications in the
social sciences series, Sage publications, 1994.

43

In that case

(

)

Pr ob( y = 1) = L ∑k β k x k =

e ∑k
1+ e

βk xk
∑ k β k xk

(1.5)

The model in (1.5) is the binary logit model, it represents the probability of the event
occurring (y=1). In the two estimations presented in this paper, 'the event occurring' is the
survival of the genocide in the first regression and the use of a firearm in the second
regression.

Appendix 2 : Death and survival in Gitesi commune
Gitesi commune, which is Kibuye’s ‘urban’ commune, had a numerous Tutsi population
before the genocide. Unfortunately, the quality of the data collection on the genocide was
weak. It is this lack of good data that made me visit Gitesi commune in October 2000. I
wanted to find out more about the genocide in that commune. Together with a re-coding of
the data, the visit to Gitesi commune yielded the following information:
-

From April 6th till April 11th, the large majority of the Gitesi Tutsi stayed in their homes28.
They did not gather by the thousands in front of the communal office as did the Tutsi
from Mabanza. This explains why, after April 11, the Tutsi of Gitesi were killed at a large
number of places throughout the entire commune, more dispersed then the Tutsi of
Mabanza commune.

-

The dispersion also explains why fewer dates on which the killings happened were
recorded. Members of the same household fled in different directions and did not know
about the fate of their household members.

-

The burgomaster apparently did not actively pursue the genocide, neither did he oppose it.
He behaved more as a bystander. This may give an additional explanation of the
dispersion of the Tutsi from Gitesi, at least in the first few days of the genocide.

-

A number of Tutsi of Gitesi came to Kibuye town center from April 12 onwards; many of
them however were killed in other places throughout the commune.

As I indicated at the beginning of the paper, the quality of the data for Gitesi commune is
poor. Re-coding is only a worthwhile undertaking if it gives the researcher additional and new
information. Looking at the registration books for Gitesi commune, I decided that re-coding
would be worthwhile for half of the sectors, namely the sectors Burunga, Buye, Bwishyura,

28

Interview, Gitesi commune, October 31, 2000.

44

Gitarama, Kayenzi and Rurangwe. The results are presented hereafter. The procedure is thus
the same as for Mabanza commune, except that for the latter, we re-coded all the sectors.

Table 13: Descriptive statistics for Gitesi commune

Total number of inhabitants registered in the 1991 census
Total number in 1994 (1991 figure* (1.03)2)
Total number of Tutsi registered in new database
Total number of victims
Total number of survivors
Number of Tutsi without entry
Percentage of 1994 Tutsi killed
Percentage of Tutsi in new database killed
Percentage of Tutsi in new database survived

6 sectors

full*

30.950
32.835
6.269
5.425
832
12
16.5%
86,5%
13,2%

61.900
65.670
12.538
10.850
1664
24
16.5%
86.5%
13,2%

* This extrapolates the findings of the six re-coded sectors to the entire commune (12 sectors) (* with 2). The
latter column does not give exact figures thus.

The re-coded sectors make up half of the population of Gitesi commune. From the data, we
learn that about 16,5% of the population in these sectors was killed, totaling 5.425 victims
from a population of 33.000 just before the genocide. These victims represent 86,5% of the
Tutsi population in the six re-coded sectors, leaving 832 survivors. Using the information for
the re-coded sectors to give approximate figures for the entire commune, we estimate that
10.850 Tutsi residing in Gitesi commune were killed and 1.664 survived.

45

Table 14 : Places where the Tutsi residents of Gitesi commune were killed
Name or location

Number of victims
Entire
(6 re-coded communes) commune

% of victims

In the cell of residence
In another cell within the sector
In another sector within the commune
In the Gatwaro Stadium
In Bisesero29
In Parish of Kibuye
In Karongi (sector Gitwa)
In the mountains
In the Kivu Lake
At any another place
At an unknown place

1.149
175
21
359
180
233
443
1.591
13
45
1.216

2.298
350
42
718
360
466
886
3.182
26
90
2.432

21,1
3,2
0,3
6,6
3,3
4,3
8,1
29,3
0,2
0,8
22,4

Total
Valid total

5.425
4.209

10.850
8.418

100
77,5

Table 15 : Places of refuge*
Tutsi from the re-coded sectors
Burunga
Buye
Bwishyura
Gitarama
Kayenzi
Rurangwe
From the other sectors
Bubazi
Gasura
Gitesi
Kagabiro
Mbogo
Rubazo

sought refuge
in Kibuye ville (Parish et Stadium)
at Karongi hill, in the mountains
in Kibuye ville, Lac Kivu, Karongi
in the mountains
in the mountains
in the mountains
in the mountains
Karongi, Nyamishaba secondary school
in the mountains
Paroisse Mubuga, in the mountains
in the mountains
Paroisse Mubuga, in the mountains

* source : author’s interviews in Gitesi commune

29

As mentioned earlier, 'Bisesero' means the hills of Bisesero.

46

Since in one out of four and a half cases the place of the killing was unknown and one out of
three places is coded as "in the mountains", it means that we have no details on the place of
death of half of the victims in our database (see table 14). Even after re-coding, we cannot fill
the gap left by the enumerators from Gitesi commune. Only interviews can bring additional
information. The persons interviewed said that a large part of the Tutsi population of Gitesi
commune was dispersed throughout the commune. This was their explanation why they used
the broad category "in the mountains", just as in the IBUKA file. They would add that a
considerable number (they did not know how much) first sought refuge at Karongi hill and
afterwards continued to Bisesero.
Figure 8 : Distribution of killing through time, n=948 cases
500

Number of victims

400

300

200

100

0
6

11

14

17

20

23

27

Days of April 1994

Since all Tutsi killed on April 17th of whom this date is known, come from sector Bwishyura
and since all Tutsi killed on April 26th of whom that date is known, come from sector
Gitarama, figure 8 may overestimate the number of people that died on these two dates when
we take all sectors into account. Since April 17th is the day of the massacre in the Parish of
Kibuye, the same remark about over representation applies to figure 8 as it did for figure 3.

47

Figure 9 : Residence-specific survival chances in Gitesi commune

percentage survived

30

20

10

0
2

3

4

6

9

12

Sector of residence

Legend : Burunga 2, Buye 3, Bwishyura 4, Gitarama 6, Kayenzi 9, Rurangwe 12.

Figure 9 shows that sector Buye was the most deadly sector among the re-coded sectors and
that considerable between-sector variation existed in Gitesi commune. This observation
corresponds to the statement of a person I interviewed in Gitesi. The interviewee said that in
Buye practically all Tutsi were killed. However the person failed to give reasons for the
between-sector variation in survival chances.
Figure 10 : Age-specific survival chances, only sectors Burunga, Bwishyura and
Gitarama

percentage survived

15

10

5

0
0-10

10-20

20-30

30-40

40-50

>50

Age group

For figure 10, I could only use data for sectors 2, 4 and 6 since enumerators in the other
sectors did not register the age of the survivors. That is also the reason why I did not perform
a regression analysis as I did for Mabanza commune.

48

Appendix 3 : estimation procedure of killing through time
Gishyita and Gisovu communes
My assumptions are not random, but based on the available information in the data file. For
Gishyita commune, I can trace 5.800 people (out of 11.273) who died in the hills of Bisesero,
because the place of death is mentioned in the data file. This figure is the ‘sure’ estimate for
Gishyita. I also trace 4.000 people who did not die in the hills of Bisesero. This leaves about
1.473 Tutsi from Gishyita who may or may not have died at Bisesero. Their place of death is
unknown or unclear in the database. Since a lot of people died in the hills of Bisesero, it is
more likely that the exact place is not known, in any case more likely than when they were
killed in Mugonera hospital or in another place in the commune. I therefore assume that 1.000
out of these 1.473 (two thirds) were also killed in the hills of Bisesero. This gives an estimated
6.800 Tutsi from Gishyita, meaning three out of five Tutsi from Gishyita, were killed at
Bisesero. For Gisovu commune, I find at least 1.000 people (out of 3003) who died in
Bisesero (the place is indicated in the data file) and 1.500 who died in other places30. As in the
case of Gishyita, I assume that 333 out of the remaining 503 (two thirds) also died in the hills
of Bisesero. This makes 1.333 Tutsi of Gisovu, meaning 44% of its killed Tutsi population,
who died in Bisesero.
Table 16
commune

total number
of Tutsi killed

Gishyita
Gisovu

11.272
3.003

30

not killed in
Bisesero
sure
assumed

killed in
Bisesero
sure
assumed

in %

4.000
1.500

5.800
1.000

60
44

4.473
1.668

6.800
1.335

About 500 of Gisovu’s Tutsi population was killed at Kiziga Hill, commune of Rwamatamu.

49

Gitesi and Mabanza communes
From interviews in the commune, we know that a considerable number of Tutsi from Gitesi
commune managed to take refuge in Bisesero. African Rights writes that Tutsi who survived
several massacres in Kibuye and Gitesi arrived in Bisesero31. The Tutsi from Gitesi commune
more specifically survived or escaped from massacres at the Gatwaro Stadium, the Parish of
Kibuye and the Home of St.-Jean. From the IBUKA file however, one can derive that the
majority of Tutsi who were killed at these places had their residence in Mabanza commune. I
refer to table 17 for the figures of Mabanza and Gitesi commune.
Table 17 : Commune of residence *
Place of death

Mabanza

Gitesi

Gatwaro Stadium
In Bisesero
In the mountains
Other specified places
Unknown

3.359
300
177
4.233
1.188

718
360
3.182
4.158
2.432

Total number of victims

9.257

10.850

*

Source IBUKA data file. In the text, I discuss the data problems for the commune of Gitesi. We note that for
Gitesi commune, the unknown places were added to "in the mountains" in the IBUKA file, a procedure that
I did not use in the re-coding of the data. The figures for Gitesi commune are extrapolations from a re-coded
file of 6 sectors.

The observation from the data file that the majority of Tutsi killed in Kibuye center came
from Mabanza commune is corroborated by interviews with survivors and the eye-witness
account of a German expatriate doctor who stayed in Kibuye center until his evacuation in
mid-May. He recalls that survivors told him that thousands of Tutsi from Mabanza commune
gathered at the communal office. During two days at the communal office the Burgomaster
told them not to leave the compound. Tutsi from other communes of Kibuye arrived and
then the Burgomaster told them that he had received the order to send everybody to Kibuye

31

Resisting Genocide, African Rights, p.11.

50

town. On Wednesday 13 th , they were taken to Gatwaro Football Stadium in Kibuye town
center. 32
Doctor W. Blam, working for the German Development Cooperation writes that (from April
11 and April 12 onwards):
“the next few days, waves of refugees, most of them from Mabanza, arrived and by Friday
(15th), more than 10.000 were concentrated in the town of Kibuye. More than 5000 in the
Gatwaro Football Stadium next to the hospital (where he was residing), several thousand at the
Catholic parish and an undetermined number with friends or parents and already at the isles
close to the Kivu Lac.”33
Doctor Blam also writes that these refugees were telling horrible stories of terrible massacres
against groups of refugees in Rutsiro commune (p.108). The figures of (at least) 10.000
refugees on the whole and 5.000 in Gatwaro Stadium corresponds with other figures, gathered
by military officers. African Rights quotes two French officers who did on-site investigations
some time after their arrival
“Colonel Patrick Sartre told Reuters that at least 4.500 Tutsis, including women and children,
were slaughtered in the Kibuye Stadium on April 16 and 17. He calculated that about twelve
thousand Tutsi had been murdered in those two days, at the church, in the stadium and in the
surrounding countryside. Lt. Colonel Eric de Stabenrath told Keith Richburg of The
Washington Post that he found 4.300 bodies piled on top of each other in Kibuye’s church and
seven to nine thousand more bodies in a sports stadium. From his investigations, he established
that the Tutsi refugees who had sought shelter at the stadium had been attacked by soldiers and
militia who had shot until they had run out of ammunition. He concluded that between eighty
and ninety-five percent of the Tutsi population had been destroyed in this area.”34
Exactly how many Tutsi were killed in the Gatwaro stadium is difficult to say. Doctor Blam
says he saw more than 5.000 refugees before the massacre, Colonel Patrick told Reuters 4.500
people were slaughtered in the stadium, Lt. Colonel Stabenrath counted seven to nine
thousand bodies and the IBUKA dictionary project traced 4.179 names of victims who died in
the stadium. The latter figure is an underestimation since we do not know the exact place of
death of a considerable number of victims.

32

Rwanda, death, despair and defiance, African Rights, 1995, pp.395-396

33

Blam, W. Genocide as ‘modern’ political instrument, original text published in German in H. Schürings
(ed.) Ein Volk verlässt sein Land. Krieg und Völkermord in Ruanda, Köln, 1994. Author’s translation from
the French version published in Jean-Pierre Chrétien, Le défi de l’ethnisme, Karthala, 1997, p.108. The
French text was checked and authorized by the author, Doctor W. Blam.

34

African Rights, Rwanda : Death, Despair and Defiance, 1994, p. 289.repeated in the 1995 edition p.416 and
p. 424.

51

In contrast to the Tutsi from Mabanza commune, a considerable number of Tutsi from Gitesi
commune were not trapped in the town center. The exact number is difficult to determine
given the unspecified place ‘in the mountains’ in the data file. This does not mean that ALL
the Gitesi refugees reached Bisesero. Gitesi indeed is an adjacent commune to Gishyita, but
Bisesero sector in Gishyita commune borders Gisovu commune and not Gitesi commune.
Moreover, with all the killing throughout the communes, a large number of Tutsi were killed
on the roads, in their houses and while hiding with friends. From my re-coding of half of the
Gitesi sectors, we learn that (when we extrapolate for the whole commune) one out of three
Tutsi from Gitesi were killed in either Bisesero (3.3%) or "in the mountains" (29.3%), while
8.1% were killed in Karongi and for 22.4%, the place is unknown. Because of the lack of
reliable data, we have to make two arbitrary assumptions: we assume that two out of three
people who died "in the mountains" reached Bisesero and that one out of two people whose
place of death is unknown also reached Bisesero. This gives a total of 3,3 + 19,5 + 11,2 =
34% of the Gitesi victims, meaning about one third or 3.700 (out of 10.850) Tutsi from Gitesi.
The assumptions of course are the weakest part of this estimation. 35
The distribution over time of the remaining 7.150 (not killed in Bisesero) victims of Gitesi
commune is also based on my re-coded data file and observations from eyewitnesses. From
the 948 people killed from Gitesi of which we have the date, 642 (67%) were killed between
April 15 th and April 19th, with 400 (62% of 642) on the 17 th, the day of the massacre in the
parish of Kibuye. This, I agree, very limited source of information, combined with
information from interviews (African Rights, Doctor Blam and author’s interviews in Gitesi),
allow us to make the following (rudimentary) assumptions: Since the major massacres in that
commune took place from April 15 to April 19, I assume that 4.300 (60% of 7.150) died
during these days, with a concentration on April 16, 17 and 18. Because of lack of accurate
data, I distributed these 4.500 victims as follows: Friday 15th (250), Saturday 16th (250), Sunday
17 th (2.000)36, Monday 18th (1.000) and Tuesday 19th (800). As a consequence, I assume also
that the remaining 2.850 (=10.850-3.700-4.300) Tutsi from Gitesi commune followed the
distribution of the entire prefecture. This is plausible given the evidence that Tutsis who were
hiding in the hills were hunted down throughout the territory of the entire prefecture and
throughout the three months of the genocide. All weighing factors for each of the dates are
thus augmented, proportionally to the number of people who died on that date, to account for
35

These assumptions are indeed arbitrary, but table 11 provides at least some support for it.

36

2000 is only 46% of 4300, meaning less than the 62% represented by the 400 victims out of 642 which died
between April 15 and 19. These 400 however almost all come from one of the re-coded sectors. Taking the
whole commune into account, this sector represents probably less then 62% of the people killed on April
17th . We also refer to our discussion of figures 6 and 4. In order to avoid overestimation for April 17th , I
assumed 2000 and not 2666, the latter corresponding to 62% of 4300.

52

these remaining 2.850 victims. The data file also shows a small number of Tutsi from
Mabanza who died in Bisesero. Because Mabanza commune is not bordering Bisesero sector
in Gishyita commune, I estimate that only a very small percentage of the ‘unknown’ in the
Mabanza file reached Bisesero.
As for Rwamatamu commune is concerned, most Tutsi were killed in the commune itself and
early in the genocide. Few people managed to escape to Bisesero. In the database, I found
about 9.000 Tutsi from Rwamatamu who were not killed at Bisesero. This leaves about 1.000
refugees who could have reached Bisesero. The presence of Tutsi from Rwamatamu at
Bisesero is corroborated by interviews with survivors37. The result is presented in table 18.
Since only the commune of Mabanza and half of the commune of Gitesi were re-coded, I
decided to stay with the overall figure of 59.050 victims found by IBUKA as a baseline for the
estimation over time, thus not accounting for the (small) differences in number of victims I
found for both coummunes after the re-coding.
Table 18 : Number of Tutsi killed in Bisesero
Commune of residence

Minimum or
‘sure’ estimate

Probable or
‘best’ estimate

Maximum
estimate

Gishyita
Gisovu
Gitesi38
Rwamatamu
Mabanza

5.800
1.000
600
400
300

6.800
1.333
3.700
700
400

7.300
1.500
4.000
1.000
800

Total number

8.100

12.933

14.600

37

Resisting Genocide, African Rights, p. 11.

38

Again, I want to stress that the difference between the minimum estimate and the best estimate for Gitesi
commune is due to the lack of data in the IBUKA file. It should be clear by now that I want to stay as close
to the data as possible. When the data do not tell me where the victim died, or only died ‘in the mountains’,
I cannot be sure that they died in Bisesero. That is where assumptions play a role. I do not think that 8100
people were killed at Bisesero, the 8100 is the bare minimum which we can be sure of. On the same token,
we can almost be sure that more Tutsi from Gitesi died in Bisesero then Tutsi registered as such in the data
file. Statistics give you tools that allow you to go beyond 100% certainty. My best estimate on the basis of
what we know of the genocide in Kibuye prefecture and Bisesero is 13.000. This is the figure that I believe
is the closest I can get to reality. 13.000 is the figure that I can reasonably defend, given the information in
the database plus some necessary assumptions to fill the gaps of the database. Some mistakes in the data or
some adjustments in the number of people from Gitesi commune who died in Bisesero (for example
because one is not sure whether they died on Karongi hill or on Bisesero hill) would not make a big change
to the 13.000 figure. Taking a 10 percent error into account, one can argue that between 11.700 and 14.300
Tutsi were killed in Bisesero between April 6, 1994 and July 1, 1994.

53

The 13.000 figure is my best estimation, based on the IBUKA file together with my re-coding
and some assumptions for Gishyita, Gisovu and Rwamatamu communes that are pretty close
to reality and two assumptions for Gitesi commune that may be less realistic. Changing the
assumptions however would not make a big difference. My minimum estimate for the number
of Tutsi at Bisesero is therefore 8.100 (which is surely lower than the real number) and my
maximum estimate is 14.600. The minimum is based on the number of victims in the IBUKA
file where the place is mentioned in the data file39.
The 13.000 estimate is lower than figures in other publications, but the author believes it is a
good estimate because it is based on large-scale data collection40. African Rights, for example,
writes that 50.000 people were killed in Bisesero. This is an overestimation. Given the total
number of victims in Kibuye Prefecture (59.050) found by IBUKA, the figure of Tutsi killed
in Bisesero given in the African Rights publication (50.000) is too high. Granted that the
IBUKA figure may be an underestimation and that the IBUKA figure only contains Tutsi
living in Kibuye before the genocide, the African Rights figure of 50.000 for Bisesero alone
remains very high nonetheless. This is because it would mean that either 10% of the
population of the entire prefecture (or two thirds of all the Tutsi from Kibuye) had gathered at
Bisesero or either that a large number of Tutsi from Gitarama, Gisenyi, Gikongoro or
Cyangugu Prefectures had found refuge in Bisesero. Both hypothesis seem highly unlikely
given the sheer impossibility to travel long distances for a number of people large enough to
reach 50.000 at Bisesero41.

39

If future research shows that most Tutsi from Gitesi did not die ‘in the mountains’ as is mentioned in the
IBUKA file, but were killed in other major massacres in Gitesi commune, the figure of 13.000 Tutsi killed
in Bisesero has to be revised downward (with one or two thousand) and the figure for the Parish has to be
revised upward. On the other hand, if future research shows that more than 3.700 Tutsi from Gitesi
commune reached Bisesero, my estimate for Bisesero has to be revised upward (with one or two thousand).

40

Provided of course that the IBUKA dictionary research project did not fail to find a lot of victims, which I
assume not to be the case here. If future research shows that the dictionary is incomplete, the estimate has
to be revised upward.

41

I do not think my estimation of the number of victims in the Hills of Bisesero diminishes the great value of
the interviews published in ‘Resisting Genocide’. Indeed, both the horrible suffering and human power of
survivors is very well documented in that publication and it continues to be the standard work on Bisesero.

54

LITERATURE AND DOCUMENTS CONSULTED

-

-

African Rights, Rwanda : Death, Despair and Defiance, 1995.
African Rights, Resisting Genocide, Bisesero April-June 1994.
Bart, F., Montanges d’Afrique, Terres Paysannes : le cas du Rwanda, Bordeaux, 1993.
Blam, W. Genocide as ‘modern’ political instrument, original text published in German in
H. Schürings (ed.) Ein Volk verlässt sein Land. Krieg und Völkermord in Ruanda, Köln,
1994. Authors translation from the French version published in Jean-Pierre Chrétien, Le
défi de l’ethnisme, Karthala, 1997.
Desforges, A., Leave none to tell the story, Human Rights Watch, 1999.
Human Rights Watch Arms Project, 1994.
IBUKA, data file Kibuye Dictionary Project, Kigali, Rwanda, 1999
Liao, T.F, Interpreting probability models, Quantitative applications in the social sciences
series, Sage publications, 1994.
Letter of the Préfet to all the Burgomasters concerning the self-defense program of the
population, on April 30th, 1994
Letter of the Préfet of Kibuye Prefecture to the Minister of the Interior and Communal
Development, on May 5, 1994.
Letter by the Minister of Interior and Communal Development to the military commander
in Gisenyi asking to send troops to Kibuye Prefecture to support an operation in the
sector of Bisesero, Commune of Gishyita, dated June 18, 1994.
Report by the Préfet of Kibuye of the day to day events in all the communes between
April 6 and April 10 dated on April 11.
Report of the meeting of the security committee of Kibuye Prefecture on April 11, 1994.
Seltzer, W., Population Statistics, the Holocaust and the Nuremberg Trials, Population and
Development Review 24 (3), 1998.
Undated letter addressed to the Minster of Interior and Communal Development
concerning the management of Gisovu Commune.
Uvin, P. Aiding Violence, the development enterprise in Rwanda, 1998.
Verwimp, Ph., Development Ideology, the Peasantry and Genocide, Journal of Genocide
Studies, November 2000 and Working Paper, Genocide Studies Program, Yale University,
1999.
World Bank, Rwanda agricultural strategy review, 1991.

Copyright © 2001 @ the author(s). Discussion papers are in draft form. This discussion paper
is distributed for purposes of comment and discussion only. It may not be reproduced without
permission of the copyright holder. Copies of working papers are available from the author.

Haut

fgtquery v.1.9, 9 février 2024