1 Tuesday, 20 February 2007
2 [Open session]
3 [The accused entered court]
4 --- Upon commencing at 9.00 a.m.
5 [The witness entered court].
6 JUDGE BONOMY: Good morning, Dr. Baccard. The cross-examination
7 by Mr. Ivetic will now continue.
8 Mr. Ivetic.
9 MR. IVETIC: Thank you, Your Honour.
10 WITNESS: ERIC BACCARD [Resumed]
11 [Witness answered through interpreter]
12 Cross-examination by Mr. Ivetic: [Continued]
13 Q. Good morning again, Doctor. Thank you again for your time and I
14 will try to finish this as briefly as possible since I only have two more
15 topics to finish with you.
16 Yesterday we finished up with the Izbica site and now I would like
17 to move to another local quickly in Kacanik. And just to confirm a couple
18 of things to make sure I'm reading the reports right. With respect to the
19 autopsies performed at Kotlina, am I correct that they were conducted --
20 that these autopsies were rather brief and were undertaken in less than
21 desirable circumstances at the Kacanik cemetery which is some distance
22 away from the site where the bodies were excavated?
23 THE INTERPRETER: Kindly slow down, counsel, please.
24 THE WITNESS: [Interpretation] I wouldn't use those words. The
25 post mortem examination were restricted because of the condition in which
1 we found the bodies. These were largely degraded because of the wounds
2 and the putrefaction processes.
3 MR. IVETIC:
4 Q. All right. You've answered my second question as well. I'm going
5 to have to slow down though because the interpreters have asked me to slow
6 down so I will have to comply.
7 Now, Doctor, with respect to the artefacts that were recovered
8 with these remains that were excavated from the well in Kotlina, am I
9 correct that several of the sets of remains had leather all-terrain boots
10 on their feet?
11 A. I did not take part in the post mortem examination conducted at
12 Kotlina, so I personally did recover no boots. And as far as this detail
13 is concerned, it is not in my -- contained in my summary report. You need
14 to refer to the report by Drs. Markwalder and Daniel Wyler.
15 Q. Do you know since we talked a little bit about --
16 JUDGE BONOMY: Mr. Ivetic, do you know if the report by these two
17 doctors contains information about the boots?
18 MR. IVETIC: It does contain information about boots. I was
19 trying to figure out if the -- what -- if he had any more information
20 about these boots. P361 at page 24, for instance, describes "all-terrain"
21 which I'm not sure what that means in terms of describing the leather
22 boots that were on one of the bodies and there were I believe a total of
23 six bodies that had some sort of leather boots. So I was wondering, in
24 his discussions with the team, if he had any additional information on
25 that. If he doesn't, I'm prepared to move on.
1 JUDGE BONOMY: Thank you.
2 MR. IVETIC:
3 Q. Now --
4 A. I did not meet the members of the team because I was only working
5 on the reports.
6 Q. All right. Fair enough. Then one more thing, if you know. We
7 talked a bit about ballistics and shells yesterday. Do you have any
8 information to add on to the -- what's already specified in the report --
9 in the Austrian report at page 75 regarding the discovery of
10 Chinese-manufactured bullet casings at the scene in Kotlina in addition to
11 7.62-calibre bullets of Yugoslav origin?
12 JUDGE BONOMY: Mr. Stamp.
13 MR. STAMP: Could we have a reference to that part of the report.
14 MR. IVETIC: P361, page 75. I believe it identifies Chinese
15 munitions from a particular factory manufactured in 1964 is my
17 Q. And again, Doctor, if you know. If you don't know, we can move
19 A. Mr. Ivetic, let me remind you that I'm a forensic pathologist.
20 I'm not an ammunitions and weapons expert, and therefore I have not
21 included this detail in my summary report. I just focussed on the
22 percentage figures, causes of death. Those reports stated that 84 per
23 cent of the victims had died due to the effects of an explosion, whereas
24 12 per cent had died through bullet wounds.
25 Q. Fair enough, Doctor --
1 JUDGE BONOMY: Mr. Stamp.
2 MR. STAMP: Just for clarification, the reference I have here and
3 which is what I saw yesterday is that it is not cartridge casings but a
4 singular cartridge case out of many that were found at the spot.
5 MR. IVETIC: Yes, I'm being refreshed by the ... [French on
6 English translation]
7 I'm getting some kind of translation of my own words. I'm seeing
8 now that the last sentence that talks about cartridges may refer in
9 general to all of the 12 that were found including the one that was the
10 Chinese make.
11 JUDGE BONOMY: All right. Thank you.
12 MR. IVETIC:
13 Q. Now, with respect to the last area on Kotlina, before we move on
14 to my last area, there were -- am I correct that there were I believe two
15 or three corpses -- I believe it was three corpses recovered from the
16 cemetery near the mosque in Kotlina as opposed to the well?
17 A. Three bodies according to the identification contained in the
18 report drafted by my colleagues.
19 Q. Okay. Now, when talking about these corpses recovered from the
20 mosque, I'm assuming that no conclusions could be reached about the
21 location where these individuals had expired. Is that accurate?
22 A. This did not appear in the report.
23 Q. Fair enough. Now, I'd like to move to Krusha e Vogel or Mala
24 Krusa which is another area you covered in your report, but before I do, I
25 would like to cover some points about the extent and limitations of
1 forensics investigations since I've been reading up on some literature
2 offered by some of your colleagues including John Clark and an
3 anthropologist Dr. Susan Black who actually did the Mala Krusa
4 investigations and I want to make sure that I'm on the same page and that
5 I'm understanding the basic principles correctly. Now, my understanding -
6 and I'm asking you if you would agree with this - is that in normal
7 pathology practice, deciding whether an injury has occurred in life as
8 opposed to post mortem is determined by looking at things such as
9 bruising, swelling, and bleeding and that while this is very difficult in
10 decomposed bodies, in skeletonised remains it is nearly impossible.
11 A. That's not quite correct. On a microscopic examination, on the
12 edge of a fracture, you can distinguish blood residues and blood
13 infiltration, but as a rule, it is on soft tissue that this is easier to
14 highlight this.
15 Q. Fair enough. Now, if we can move to Mala Krusa and the findings
16 of the team from Glasgow that you reviewed. According to I believe both
17 Dr. Black's report and yours there was only one set of remains, that is to
18 say, remains from one individual recovered from this site, from this
19 village, that showed signs of cremation and that the others are said to
20 have no indications of cremation. Now what I want to ask you is, based
21 upon that, can we conclusively exclude to a degree of reasonable
22 scientific certainty the possibility that any of these other remains were
23 subjected to fire; that is to say, if they do not show signs of cremation,
24 can we exclude the possibility that the remains were exposed to fire?
25 A. In the report provided by Professor Vanezis and Professor Black,
1 as far as I'm concerned, I did indicate that these human remains included
2 four different individuals and that signs of burning related to the four
3 bodies but these reports would be better displayed on the screen and
4 therefore I could check this.
5 Q. If we can look at P110, the first page should suffice, since they
6 were nice enough to do a summary of all the remains recovered from the
7 three locations on the first page.
8 MR. IVETIC: And we can perhaps assist the doctor and finish the
9 examination since this is my last area.
10 Q. Doctor, I don't believe that we have -- I don't believe that we
11 have a French version of this exhibit. We have the English there on the
12 left. Is that something that we can use to assist to refresh your
13 recollection? And I think the part you're probably going to be interested
14 in is the part under the heading remains.
15 JUDGE BONOMY: Mr. Stamp.
16 MR. STAMP: I think -- my learned friend might have been misled by
17 the notification and I don't think I used this particular document in the
18 examination-in-chief. This document, as can be seen, is dated the 29th of
19 October, 1999, whereas at page 32, Dr. Baccard's report in respect to Mala
20 Krusa is listed a report by Dr. Black of the 27th of October, 1999. The
21 document placed on notification was placed in error and that is why in the
22 examination-in-chief we referred to another document which is P2748.
23 JUDGE BONOMY: Page 32 indicates that the examinations were
24 carried out on the 23rd of June. Now, you tell us the date the report
1 MR. STAMP: In the supporting documents immediately above that,
2 Dr. Baccard says: "This report is based on the following documents, the
3 report of Dr. Peter Vanezis of Glasgow, regius professor of forensic
4 medicine," et cetera. And the report of Dr. Susan Black, consultant on
5 forensic pathology et cetera, and it's dated the 27th of October.
6 JUDGE BONOMY: And what does the document we are looking at now
7 relate to?
8 MR. STAMP: The document we are looking at relates to another
9 examination that both of them did but it is not in the indictment, another
10 examination in Kosovo --
11 JUDGE BONOMY: And was it also at Krusha e Vogel?
12 MR. STAMP: Yes. The document in relation to the indictment in
13 relation to Dr. Baccard's report is as I indicated Exhibit --
14 JUDGE BONOMY: Well, there must be two. There's one by Vanezis
15 and one by Black.
16 MR. STAMP: 2748 --
17 MR. IVETIC: That's what I'm looking for right now, Your Honours,
18 since I don't think I had that on the notification. Does counsel have a
19 number for that exhibit?
20 MR. STAMP: The forensic pathologist, Dr. Vanezis, and the
21 forensic pathologist, Dr. Black, brought together Exhibit 2748 is a
22 combined report. The part signed by Dr. Vanezis is of the 10th of
23 November, 1999, as indicated by Dr. Baccard and that signed by Dr. Black
24 is the 27th of October, 1999, also as indicated.
25 JUDGE BONOMY: And that's P2748?
1 MR. STAMP: 2748.
2 MR. IVETIC: Herein lies my confusion, Your Honour. That report
3 relates to Krusha e Vogel, Velika Krusa, Krusha e Madhe, Velika Krusa, not
4 Krusha e Vogel. So therein is my confusion. I thought we were talking
5 about Mala Krusa.
6 JUDGE BONOMY: Well, your point -- your point may be equally valid
7 because if what you're after is evidence of the absence of --
8 MR. IVETIC: I agree --
9 JUDGE BONOMY: -- burning injuries, then you're getting it from
10 the document you're relying on. And so far it would appear that the one
11 the Prosecution are relying on doesn't relate to the locus you're
12 concerned about.
13 MR. IVETIC: I hope so. I hope there's not another document that
14 I don't know about that I should have had to prepare. I think I can still
15 ask the two questions I had in mind --
16 JUDGE BONOMY: Before you do that --
17 MR. IVETIC: Sure.
18 JUDGE BONOMY: Mr. Stamp, there does seem to be an issue here over
19 the locality that we are concerned about.
20 MR. STAMP: These are sister villages. My understanding
21 effectively it's the same place Mala Krusa in -- Krusha e Vogel --
22 JUDGE BONOMY: Yes, but if you put on the screen 2748, the one
23 you've just identified. Can we see it, please. Velika Krusa is a
24 different village from Krusha e Vogel. I think that means old Krusa
1 MR. IVETIC: It means greater or large, big Krusa and small Krusa.
2 JUDGE BONOMY: Small Krusa.
3 Can you remind me, Mr. Ivetic, of the paragraph in --
4 MR. IVETIC: The doctor's report --
5 JUDGE BONOMY: Paragraph 75 of the indictment that deals with
6 these villages?
7 MR. IVETIC: Well, there are two separate incidents alleged, one
8 in Velika Krusa one in Mala Krusa. It's my understanding these villages
9 are apart.
10 JUDGE BONOMY: You can --
11 MR. IVETIC: Which --
12 JUDGE BONOMY: -- identify the paragraph in -- it must be in 75
13 somewhere --
14 MR. STAMP: 75 (C) --
15 MR. IVETIC: Since they are both in the same municipality but they
16 are by no means the same village.
17 JUDGE BONOMY: 75 --
18 MR. IVETIC: (C), I think, Your Honour. But I don't have a copy
19 in front of me. I'm working off memory.
20 JUDGE BONOMY: Just a second.
21 So it would appear that the bodies with burning injuries may not
22 be connected to the principal incident that's referred to in that
23 paragraph, which is the shooting and burning event in the barn, I think,
24 that we've heard quite an amount of evidence about.
25 MR. IVETIC: That's correct, Your Honour, and the confusion that I
1 think -- due to the fact that at page 32 of the witness's report the
2 heading is Krusha e Vogel.
3 JUDGE BONOMY: Yes.
4 MR. IVETIC: And that would be Mala Krusa. And it looks now, as
5 I'm reading through the 2748, it looks like the description of the remains
6 is actually from Velika Krusa, so that the --
7 JUDGE BONOMY: Now, let's see.
8 Dr. Baccard, can you clarify this at all for us?
9 THE WITNESS: [Interpretation] I think it is important to
10 understand that when I prepared the report, I worked on the basis of
11 documents that had been made available to me and the information that was
12 provided to me. And I was told that these documents concerned Krusha e
13 Vogel. I don't know the geography of Kosovo at all, so this is why I have
14 included this name with the corresponding reference, which you have on
15 your screen VK0002. So I see here that it says Velika Krusa.
16 JUDGE BONOMY: Well, it sounds to me as though this is something
17 that is going to have to be resolved in the wash in due course. It may be
18 that you've no further questions on this. I don't know.
19 MR. IVETIC: I think I would still like to ask questions if he can
20 review the first page of P110 and tell me if what I'm proposing is
21 accurate -- if my reading of this document is accurate, since it is a
22 forensic report, he has experience in it, and he's based his findings on
23 reviews of reports conducted by these same individuals, I think he should
24 be able to give us at least an indication on that one area, i.e. --
25 JUDGE BONOMY: All right.
1 Mr. Stamp, you want to say something else.
2 MR. STAMP: I was just going to indicate that we will provide the
3 nexus between these bodies -- we will bring evidence to provide the nexus
4 between these bodies and --
5 JUDGE BONOMY: Very well. So you want P110 on the screen do you?
6 MR. IVETIC: Sure. I think that was -- yes, P110, it was on
8 Q. Now, Doctor, if you could review again the summary which is on the
9 first page. It talks about sites A, B, and C, and there they have
10 descriptions of the remains that were recovered and I'll give you a moment
11 or -- as long as you need to read that section. And then I just want to
12 ask you one question relative to that.
13 A. May I ask you to confirm that this document which is on the screen
14 was not part of the documents that were given to me. Is that right? A
15 priori, this is not part of my report.
16 Q. That is the understanding we've come to now in discussions that
17 we've had. It looks as if this is not in your report, although it is
18 Krusha e Vogel, the location that is listed in your -- that is listed in
19 the heading of your report, although I think we've now agreed that your
20 report actually deals with Krusha e Madhe, or Velika Krusa.
21 A. Could you possibly scroll down, please.
22 I've finished reading this part of the document, Your Honour.
23 JUDGE BONOMY: [Previous translation continues]...
24 MR. IVETIC:
25 Q. The question, Doctor, is based on these findings by Dr. Sue
1 Black. Can we exclude the possibility -- strike that. Let me start
3 With respect to the bulk of the bodies here, where it says that
4 there were no evidence of cremation, with respect to the bodies that say
5 that there was no evidence of cremation, can we exclude to a reasonable
6 degree of scientific certainty the possibility that these remains were
7 exposed to fire?
8 A. Could you focus the document or could we see the middle part here.
9 According to what I see here - this is the first time I see this document
10 today. Three sites are mentioned, three sites where human remains were
11 recovered. And these human remains were commingled. They belonged to
12 several people. As far as site A is concerned and the PGZ sample 1 SR,
13 there were no signs of cremation and seemingly, only one person was
14 mentioned in Sue Black's report as far as site A and PG 10 SR is
15 concerned, I can read that this sample contained a minimum of three
16 individuals but, more likely, six people. There were no signs of
17 cremation on the second sample.
18 As far as the third sample is concerned, still on site A, PGZ 15
19 SR, there were signs of cremation, and this -- these were signs on one
20 person only. As far as site B is concerned Sue Black mentions two samples
21 here. It is difficult to understand whether a skeletal inventory was
22 possible. This is not mentioned in the report, and there are no signs of
23 cremation here. As far as site C is concerned, only one sample was
24 recovered and, here again, no mention -- it is not stated how many people
25 were included in the sample and whether there was -- and no -- nothing is
1 said about signs of cremation in this case.
2 Q. And when there are no signs of cremation, can we then exclude
3 that the remains were exposed to fire? That is to say, if remains are
4 exposed to fire, will there always be signs of cremation?
5 A. Cremation can be stronger or weaker. It depends on how close to
6 the fire the victim has stood or -- but to give you a general answer,
7 we -- from what I understand, here we have samples of skeletonised bones
8 where the soft tissues exposed to fire, where soft tissue burnt. I have
9 no -- nothing here that allows me to say it, but as far as the bones
10 themselves that are protected by soft tissues on the fresh body, the
11 anthropologist has noticed nothing --
12 Q. Okay. Thank you.
13 A. -- for PGZ 15 SR.
14 Q. Thank you. And just to clear up, then. With respect to all the
15 reports you looked at, am I correct that you have not looked at any
16 reports related to Krusha e Vogel apart from the one that we have now
17 determined deals with Velika Krusa?
18 A. The list of the reports I have reviewed is to be found in my own
19 summary report, and apparently, there was some confusion with respect to
20 Krusha e Vogel, but this is something, information that was given me
21 orally when I was drafting the report.
22 Q. And I guess my last question: This information was given orally
23 to you by the doctors that did the report or by the Office of the
25 A. By the investigators who gave me the reports. I was not in
1 contact with the physicians, with the doctors, who authored these reports.
2 Q. All right. Well, thank you, Doctor, you've cleared it up for me.
3 I apologise for the confusion and for the length of my examination. I
4 meant it to be much shorter than it was. Thank you for your time.
5 MR. IVETIC: Thank you, Your Honours.
6 JUDGE BONOMY: Well, perhaps you can share with us the light that
7 you've seen at the end of this, when you say that it's been cleared up for
9 MR. IVETIC: That Krusha e Vogel that is referenced in this report
10 is not Krusha e Vogel, that it's Krusha e Madhe.
11 JUDGE BONOMY: When were you given that information, Dr. Baccard?
12 THE WITNESS: [Interpretation] I believe that it was at the end of
13 2001 at the time when I was preparing my report. I was told that it
14 was -- they were the same -- one and the same locality.
15 JUDGE BONOMY: And you say, Mr. Stamp, that you can produce
16 something which will clarify this, make the proper link?
17 MR. STAMP: Yes, Your Honour.
18 JUDGE BONOMY: Thank you. Do you have re-examination?
19 MR. STAMP: Yes, Your Honour.
20 JUDGE BONOMY: Okay.
21 Re-examination by Mr. Stamp:
22 Q. Doctor, you were asked a series of questions about the post mortem
23 examinations for the victims that were found in the well at Cirez, eight
24 victims, and it was suggested perhaps that there was no evidence that they
25 died from drowning, that the drowning was ante mortem. And you were asked
1 about the basis for the finding that the drowning was ante mortem. I
2 would like you to look at the report of the forensic pathologist who
3 examined those bodies. I have a copy of it in French. This is Exhibit
4 P383, which I would like to be brought up. But I would like to hand the
5 doctor a copy of the French version.
6 What I'm going to ask you first, Doctor, is to look to see if
7 there is evidence on which the forensic pathologist could conclude that
8 there was ante mortem drowning, and then I'll ask you a specific
10 MR. STAMP: While the doctor is looking at the original French
11 version, I could indicate to the Court that the English version does not
12 have the photographs which are in the French version. So -- but there's a
13 description of what's in the photographs.
14 THE WITNESS: [Interpretation] I'll try to give a very clear answer
15 to your question and I believe that my colleagues based their diagnosis of
16 drowning by immersion on one element, and that's the aspect of the lungs.
17 Because you can -- there was an edema in the lungs that was -- that was
18 caused by the drowning.
19 MR. STAMP: Thank you. Before you finish your explanation, could
20 we just have a look at the relevant page. It's page 4 in the English
21 translation. Could we scroll down a little bit. Could we scroll down to
22 under the heading "Autopsy." Is that page 4? Is this Exhibit 383? I'm
23 sorry. I have the page number incorrect. I think we are looking for page
24 number with ERN 5 -- ending 5686. And that is page 56, actually, in
25 the -- in e-court, and we have it here. Yes.
1 MR. STAMP:
2 Q. This, Doctor, we can see from the report, is to some degree
3 typical of all the victims that there are no signs of damage to vital
4 organs likely to fit apart from what is described as signs of pneumothorax
5 in the lungs, and that is typical of all the victims. Is that correct or
6 can you explain what you see here?
7 A. Could you -- we please scroll up the page because I would like to
8 see what body this particular report or part of the report refers to. I
9 see, we have body S1P1C1. Can we scroll down again, please. I would like
10 to see what is said about the lungs. There is a mistake in translation
11 because here we -- because the pulmonary edema has been translated in
12 English instead of a pneumothorax. It's not the same thing at all. A
13 pulmonary edema is when serum, serosity goes into the lungs, into the
14 alveoles because of drowning, except here we have signs of pneumothorax.
15 And pneumothorax means that you have air that goes between the chest
16 cavity and the lungs. It is something completely different. It has
17 nothing to do with what you find in the original French reports.
18 Q. So the translation to pneumothorax is incorrect. What is the
19 correct translation?
20 JUDGE BONOMY: Well, you've been given it, pulmonary edema.
21 MR. STAMP: Yes.
22 JUDGE BONOMY: Can I ask you one question on this. Is this not a
23 particularly unfortunate mistranslation? Because I take it that
24 pneumothorax could be caused by a fracture of a rib.
25 THE WITNESS: [Interpretation] Yes, you -- apparently you have good
1 knowledge of medicine apparently, but here we see that there is a mention
2 of pneumothorax. It can be a complication you find after a rib is
3 fractured, but this does not appear in the original French reports. There
4 is no mention of pneumothorax in the French report. You're talking about
5 two things that are completely different.
6 JUDGE BONOMY: But it might explain why counsel, who are relying
7 on an English version, could think that there is a possible challenge to
8 the view that -- or the opinion that the body was -- the deceased died
9 through drowning; however, you've clarified the point for us because the
10 report was originally written in French.
11 Mr. Stamp.
12 MR. STAMP: Thank you, Your Honour.
13 Q. And I think you said, just to be sure, that pulmonary edema, as
14 indicated in the original report, is a symptom or a sign of drowning?
15 A. Yes, that's correct, among other things.
16 Q. Thank you. From the photographs that accompany the report, in
17 respect to the two victims that were allegedly raped, were there attempts
18 to indicate the signs of blood on the vulvar opening and the abrasions to
19 the inside of the thigh in those reports or in that report?
20 A. Yes, that's correct, for S1P1C1 and S1P1C2.
21 Q. Thank you. If we could move to Izbica, I just have one question.
22 From the medical reports that you have, the forensic pathologist reports
23 that you have, would you or would you not be in a position to comment on
24 how the bodies were moved and when -- if when -- if so, when?
25 MR. IVETIC: Well, Your Honour, I thought Izbica was a site where
1 there were no bodies that were actually recovered so I'm a little -- I
2 just want to make sure we're talking about the same site before I can know
3 whether --
4 JUDGE BONOMY: Izbica is where the video relates to, where the 127
5 or 130 -- there was 129 or 137 bodies that were apparently exhumed.
6 MR. IVETIC: 101. 127 were on the video, Your Honour. The only
7 exhumation we know about was 101.
8 JUDGE BONOMY: But the video is what I think Mr. Stamp is dealing
10 Mr. Stamp.
11 MR. STAMP:
12 Q. From the video you would not be in a position to comment on how,
13 when, and where, and did -- the distance the bodies were moved, if they
14 were moved?
15 A. No, sir. As I said previously, there are limits to the
16 conclusions you can draw from these photographs and this video. The only
17 obvious thing that you could note, that was some of the victims had been
18 turned over because some of the victims were face down and their bodies
19 had been turned over and they were lying on their back. And there were
20 some dermabrasions on limited surfaces of some of the bodies.
21 Dermabrasion, it means that there is excoriation, superficial excoriation
22 on the skin and that meant that the body -- the face had been bruised by
23 the soil, by the ground, but it was impossible to know exactly along what
24 distance these bodies had been moved. And that was the only thing that
25 could be said about these bodies and their movements. The only thing you
1 could say for sure was that some of these bodies had been turned over.
2 MR. STAMP: Thank you very much, Your Honours.
3 Q. Nothing further, Dr. Baccard. Thank you very much.
4 JUDGE BONOMY: Mr. Stamp, there are one or two matters perhaps
5 relating to exhibits here. I've a note and I've now forgotten why I have
6 this note about an appropriate application being submitted to identify six
7 bodies from Izbica. Is that something you mentioned yesterday or is it
8 something that --
9 MR. STAMP: No, no --
10 JUDGE BONOMY: -- that didn't arise?
11 [Trial Chamber and legal officer confer]
12 JUDGE BONOMY: Mr. Stamp, one of the things that arose in the
13 cross-examination by Mr. Milosevic was whether there were documents of the
14 Yugoslav authorities relating to Izbica. And I think that it was
15 following that in the transcript I picked up some undertaking or
16 indication that an attempt would be made to identify six additional
17 victims. However, that's not something that you are familiar with, is it?
18 [Prosecution counsel confer]
19 MR. STAMP: I don't think the situation has somewhat advanced
20 since the Milosevic case because during this case we proffered documents
21 to Tomasevic who put in a number of reports in respect to those bodies.
22 JUDGE BONOMY: Yeah.
23 MR. STAMP: Some, I think, with identifications.
24 JUDGE BONOMY: Okay. Now, the other thing is that in your list of
25 exhibits used in this case we saw a document by a policeman from the
1 metropolitan police in London relating to what he had discovered about the
2 identity of victims. The one we saw may have related to Bela Crkva. Is
3 that the only one, the only document of that nature?
4 MR. STAMP: No, Your Honour, in respect to other sites.
5 JUDGE BONOMY: Yes.
6 MR. STAMP: My understanding is that similar documents were
7 tendered from some of the crime-base witnesses are documents identifying
8 the people who were --
9 JUDGE BONOMY: Are there any documents of that nature in this --
10 MR. STAMP: Yes.
11 JUDGE BONOMY: There are.
12 MR. STAMP: In most of the autopsies, the persons are identified
13 at the autopsy and the doctors who performed the autopsy indicate the base
14 of the identification.
15 JUDGE BONOMY: And that particular one, which may be an example of
16 others, is referred to in this list as an exhumation report issued by the
17 metropolitan police, but when we saw it, what it was was a statement by a
18 policeman of things that had been told to him.
19 MR. STAMP: Your Honour, it was -- the exhumation -- what is
20 described as an exhumation report is composed of statements of
21 photographers, hundreds of photographs, the statement of those police
22 officers -- two police officers involved in the identification of the
23 bodies, and the reports of the doctors who performed the autopsies on the
24 body or body fragments that were exhumed. So the expression "exhumation
25 report" covers a composite set of materials involved in the whole
1 exhumation process itself, from photographs of the terrain to the actual
2 disinterment of the bodies to the identification of the bodies, the
3 photograph of the bodies and the post mortem examinations.
4 JUDGE BONOMY: Can we take it that the counsel for the accused are
5 satisfied that this supporting material relied upon by the Prosecution as
6 being the basis for Dr. Baccard's summary is presented in a satisfactory
7 form? It's obviously hearsay as he presents it, but the written material,
8 no doubt, would not change in its content if we insisted on other
9 formalities being applied to it. But, for example, where there's a
10 document that's a statement rather than a report, there's a process, a
11 procedure, for admitting that if it's contentious. But I wonder if we've
12 sensed the situation correctly, that everyone's content that this material
13 is presented in this form and that the real issue in the end of the day
14 will be the weight to be given to it, bearing in mind that Dr. Baccard has
15 a limited personal knowledge of the nature of the inquiries carried out by
16 other professionals in a variety of disciplines.
17 Now, I'm regarding the silence in response to that as a
18 confirmation that we're proceeding along the correct lines and that we do
19 not need to take an unduly formalistic approach to this.
20 Mr. Ivetic.
21 MR. IVETIC: I have another concern tied to that. Now I really
22 don't know what exhibits the Prosecution is tendering and what exhibits I
23 need to work with Mr. Stamp on to come with a redacted form since the one
24 that I thought we had was 110 and I was advised that that one is not
25 coming in. So I think I need to know from Mr. Stamp which ones on the
1 witness notification are being sought to be tendered with this witness
2 before -- it won't change my answer to Your Honour's question but it will
3 be interesting for me to know --
4 JUDGE BONOMY: You've got until tomorrow to sort that out as I
5 understand it.
6 MR. IVETIC: True.
7 JUDGE BONOMY: You and Mr. Stamp are to meet to discuss that.
8 MR. IVETIC: That's correct.
9 JUDGE BONOMY: We had assumed from that that you, having
10 challenged a limited number of the documents, that you were accepting for
11 the purposes of, I think, a very reasonable approach by the Defence to
12 this issue, that the way in which this was being done was an order. In
13 the Milosevic trial, I don't think any attention was paid to the
14 formalities of presenting evidence of this nature. On the other hand,
15 that may have been because there was a measure of realism or it may have
16 been - and possibly more likely - because he wasn't legally represented
17 and there was a measure of -- well, there was a limit to which the amici
18 would take up issues. It's not clear that they would necessarily have
19 considered an issue like this. Maybe they did and I'm being unfair to
20 them in suggesting that it was simply a question of realism. But we don't
21 want to leave this unsaid because we really feel that we should not have
22 to be exploring technical difficulties about evidence of this nature and
23 we should be able to get straight to the meat of it and the weight of it,
24 which we see as a very debatable subject.
25 Now, can I take it, Mr. Ivetic, that between today and tomorrow
1 you and Mr. Stamp will sort out which of these documents have not to be
2 considered and insofar as there is no issue between you, that we should
3 proceed simply to give them the appropriate weight in our deliberations?
4 MR. IVETIC: I definitely will meet with Mr. Stamp and accomplish
5 that. I had planned to do that as soon as we finish this witness, on my
6 side, going through and presenting highlighted versions of the exhibits
7 with the material that I thought needed to be excluded and we were going
8 to start that process and hopefully complete it today.
9 JUDGE BONOMY: Insofar as you can't reach agreement, we will
10 obviously have to be involved in adjudicating.
11 MR. IVETIC: I appreciate that, Your Honour. The indication I've
12 gotten so far is that we should be able to reach some agreement I think on
13 most if not all the matters.
14 JUDGE BONOMY: Thank you.
15 [Trial Chamber confers]
16 JUDGE BONOMY: Dr. Baccard, we thank you for coming to give
17 evidence again to the Tribunal. That completes your evidence. We
18 appreciate the limitations upon your understanding and knowledge of all
19 the work that was done underlying these various sites and the rather
20 unsatisfactory state in which you may feel from time to time the evidence
21 is left orally, but you can rest assured that we will be giving close
22 attention to all the underlying material. And in this case, where there
23 are six accused, all separately represented, that we will have
24 submissions, no doubt, about the weight to be attached to the various
25 conclusions reached.
1 So I think we have a long way to go in considering the material
2 that you have presented to us, and we are grateful to you for the
3 assistance you've been able to give in dealing with counsels' questions.
4 You're now free to leave. Thank you.
5 THE WITNESS: [Interpretation] Thank you, Your Honour.
6 [The witness withdrew]
7 [Trial Chamber and legal officer confer]
8 JUDGE BONOMY: The next witness, Mr. Stamp?
9 MR. STAMP: That is Dr. Patrick Ball.
10 [The witness entered court]
11 JUDGE BONOMY: Good morning, Dr. Ball.
12 THE WITNESS: How do you do, Your Honour?
13 JUDGE BONOMY: Would you please make the solemn declaration to
14 speak the truth by reading aloud the document now before you.
15 THE WITNESS: I solemnly declare that I will speak the truth, the
16 whole truth, and nothing but the truth.
17 JUDGE BONOMY: Thank you. Please be seated.
18 Mr. Stamp.
19 MR. STAMP: Thank you, Your Honour.
20 WITNESS: PATRICK BALL
21 Examination by Mr. Stamp:
22 Q. Good morning, Dr. Ball.
23 A. Good morning.
24 Q. Could we start by you telling us your full name, your occupation,
25 and your present job.
1 A. Yes. My name is Patrick Ball. I'm a sociologist and a
2 statistician. I'm currently the director of the human rights programme
3 and the chief technical officer at the Benetech Initiative in Palo Alto,
5 Q. How long have you been a sociologist and a statistician?
6 A. I've been working in these tasks since my graduation from my
7 undergraduate institution in 1988. I got my Ph.D. in 1998. I've been
8 doing data and statistical work in human rights since 1991.
9 Q. And what's the Benetech Institute?
10 A. The Benetech Initiative is a non-profit that has a series of
11 projects which work to use technology for the benefit of humanity,
12 including projects that develop advanced technology to detect land-mines,
13 projects that monitor biodiversity, and projects that assist blind people
14 to access print material.
15 JUDGE BONOMY: The transcript has missed the location of the
16 Benetech Initiative. Could you give us that again, please.
17 THE WITNESS: Palo Alto, California, United States.
18 Q. It's Benetech?
19 A. Yes, it's a contraction of beneficial technology.
20 Q. In 2001/2002 you were a deputy director of the American
21 Association for the Advancement of Science?
22 A. The science and human rights programme of the American Association
23 of the Advancement of Science, yes.
24 Q. And on the 3rd of January, 2002, did you prepare a report and
25 submit it to the Office of the Prosecutor in respect to your studies on
1 killings and migration in Kosovo in 1999?
2 A. I did, together with four co-authors.
3 Q. And those co-authors are mentioned on the cover of the report?
4 A. Yes, and their biographies to that point are included in the
6 MR. STAMP: That, Your Honour, is P1506.
7 Q. And on the 19th of February, 2002, did you prepare an addendum to
8 that report -- did you along with your colleagues that you mentioned
9 before --
10 A. Yes.
11 Q. -- prepare an addendum?
12 A. We did.
13 Q. And subsequently on the -- --
14 MR. STAMP: That addendum, Your Honours, is P1391.
15 Q. And thereafter, on the 15th of November, 2002, you prepared a
16 corrigendum to the original report along with your colleagues?
17 A. Yes, I did.
18 MR. STAMP: And that, Your Honours, is P1394.
19 Q. In respect to the reports that you prepared on killings and
20 population movements in Kosovo in 1999, which I just mentioned, did you
21 testify in the trial of Slobodan Milosevic on the 13th to the 14th of
22 March, 2002, and again on the 2nd of May, 2003?
23 A. Yes, I did.
24 Q. And was your testimony truthful? In other words, did you answer
25 accurately to the best of your ability and would your answers today be the
1 same if you were asked the same questions?
2 A. Yes.
3 MR. STAMP: The testimony in sequence, the three days in sequence,
4 is Exhibit P269 -- P2690.
5 JUDGE BONOMY: Thank you.
6 MR. STAMP:
7 Q. And finally with respect to the documents, did you and your
8 colleagues -- or at least some of your colleagues submit another addendum
9 to your report on the 28th of January this year?
10 A. My former colleagues have reviewed the report, but this report was
11 written by a different group. A long time has passed and people's
12 interests have moved on.
13 Q. Okay. But did you lead and supervise the group that wrote this?
14 A. Yes, I did.
15 Q. Thank you.
16 MR. STAMP: That last addendum, Your Honours, is P2678.
17 JUDGE BONOMY: Thank you.
18 MR. STAMP:
19 Q. You told us a little bit about your past and your experience in
20 the field of statistics, sociology, and human rights. Did you also
21 prepare or request an updated curriculum vitae -- updated to this year?
22 A. Yes, I did.
23 MR. STAMP: That document had been uploaded into e-court, Your
24 Honours, as P2791.
25 I do not propose to go through what is a very lengthy document.
1 Q. But I see here, Doctor, that you have published many books and
2 technical publications in your field of studies over the last 15 or so
3 years. And you -- in your recent employment that we are going over, you
4 are the director of the human rights programme in 2003 to now at the
5 Benetech Institute?
6 A. Benetech Initiative, yes, I am.
7 Q. Thank you. We can see from the -- from your testimony in the
8 Milosevic case that the crime report -- I should call it the major report
9 of the 3rd of July was peer reviewed. It was subject to a scholarly
11 A. Yes, it was the central piece of that project, and so it was
12 subjected to two rounds of review, actually, including an extensive
13 international team.
14 Q. And that review -- that report, can you tell us how since then it
15 has been received in academic circles.
16 A. Yes. First I was, along with one of my co-authors, given an award
17 by the American Statistical Association for the use of science, of
18 statistical techniques in this way. That award was given to us in August
19 of 2002 by the social statistics section of the American Statistical
20 Association --
21 Q. Before you go on.
22 A. Yes.
23 Q. Thank you. I see at page 11 of your curriculum vitae a reference
24 to fellowships and awards and I see for 2002 there is an award, a special
25 achievement award of the Social Statistics Section of the American
1 Statistical Association. Is that the award that you just referred to?
2 A. Yes.
3 Q. Before you go on speaking about how your report has been received
4 by your peers, I see that there are two -- there are quite a few awards
5 listed here, but subsequent to that award in 2002 you have received two
6 other awards for your work. In 2004 the Eugene Lawler Award for
7 Humanitarian Contributions in Computer Science Informatics of the
8 Association for Computing Machinery. Can you tell us about that, what
9 that was in respect of?
10 A. Yes. The Association for Computing Machinery, the ACM, is the
11 professional association for computer scientists in the United States. I
12 was given that award for my design of several pieces of software, which --
13 the ideas of which underlie much of this work but have become mature
14 pieces of software and released in their own right since then. It was
15 both a technical award for the technology that we used in the -- in
16 these -- in these pieces of software as well as a humanitarian award for
17 the uses to which the software has been put, which is the promotion of
18 human rights.
19 The second award to which counsel referred is -- excuse me -- was
20 the Pioneer Award, awarded by the Electronic Frontier Foundation, which is
21 given annually to one or a small number of people who have promoted what
22 in the US are called civil liberties, more broadly, human rights, through
23 the use of electronic techniques. So those are the two awards I believe
24 you're referring to.
25 Q. Yes, thank you. So -- yes. You were answering my question when I
1 interrupted you in respect to how your original report "Killings and
2 Refugee Flows in Kosovo, 1999," how it was received in academic circles.
3 Can you continue that answer, please.
4 A. Yes. The methods were considered very interesting in this
5 application. As I mentioned at some length in my earlier testimony, these
6 methods are not especially advanced in their field. They are widely used
7 throughout mathematical demography and mathematical statistics more
8 generally. However, the application of these techniques to human rights
9 problems is interesting to the field. So we were asked to -- my colleague
10 Jana Asher and I were asked to write an article for the Journal of the
11 American Statistical Association called "Chance," which we published in
12 the summer, the US summer of 2002.
13 We also, as part of the process of disclosure to the court in the
14 Milosevic case disclosed the statistical data which underlie the
15 calculations found in all the material in the 2002 reports and most of the
16 material in the recent 2007 report, although the 2007 report also contains
17 new material made available to me by the Office of the Prosecutor.
18 Q. Thank you.
19 MR. STAMP: May I just indicate to the Court that the new material
20 which the witness just referred to has been disclosed to the Defence.
21 THE WITNESS: Should I continue?
22 MR. STAMP:
23 Q. Please.
24 A. So I have heard anecdotally through appearances at conferences and
25 other speaking engagements at various universities that this material is
1 being widely used by graduate students who study political violence. This
2 was our intent, that we should spur a new generation of students to think
3 about political violence in a more rigorous and statistical way,
4 particularly social scientists but also statistical students. So our
5 broad objective in the disclosure of this material has been, I think, to
6 some measure achieved. Does that --
7 Q. Yes. May I just follow up. The method you used in the
8 application of standard statistical methods to human rights issues in
9 Kosovo, how well-established or no well-established at all - can you
10 comment on that - are these methods in the scientific community?
11 A. There are two different methods that were used in the analysis of
12 data in Kosovo. One of the methods was used to analyse killings, patterns
13 of killings; the second method was used to analyse the pattern of
15 The analysis of killings used a technique called multiple systems
16 estimation, and the purpose of multiple systems estimation is to
17 determine, given a number of independent sources of information what might
18 be missing from them. The reason one would do this is that the
19 information that is missing could be in some way systematically different
20 from the missing - excuse me - from the data which is known. This
21 technique is very widely used and has been used for over a hundred years
22 in the correction of censuses in the analysis of industrial processes, in
23 the study of wildlife populations, and in many other statistical areas.
24 To my knowledge, the use applied to human rights data has only
25 been done in, I believe, now five countries, and my group has been
1 involved in all of them. Principally because most human rights projects
2 do not have the resources to do this work; it's hard.
3 Q. And is it -- when you say "hard," what do you mean?
4 A. You have to collect a lot of data and then you have to put it all
6 Q. Okay.
7 A. And putting it all together means, quite literally, taking every
8 one of many, many thousands of records and comparing it to tens of
9 thousands of other records by hand. So, for example, in the work that we
10 did in Kosovo, we had to make over 17.800 independent decisions about
11 whether a particular record in one database was the same as a record in
12 another database. However, because our results depended very sensitively
13 on whether that decision was made properly, in no case did we allow that
14 decision to be made only once. We made the decisions multiple times. We
15 made every one of those match comparisons independently at least twice,
16 and many of the comparisons were made three, four, and even five times
17 independently by separate coders. This took several months, very intense
18 effort, and there are few human rights groups with the statistical
19 capacity to manage a project of this kind.
20 Q. Thank you, Doctor.
21 A. I'm sorry, may I add to that?
22 Q. Of course, please.
23 A. When we made those comparisons, of course we then compared the
24 comparisons to each other to determine if they made the same decisions.
25 And I'm sorry that I don't recall the precise statistic. It's in appendix
1 1 of the January 2002 report, but the agreement across all raters, all the
2 people making the decisions, was in excess of 90 per cent. We had a very,
3 very high level of agreement across the people making those decisions, and
4 when they disagreed, we had a supervisor review all those decisions and
5 make the final determination.
6 Q. I'd like to focus a little bit briefly on the latest addendum,
7 because your previous report and your previous addendum were discussed
8 quite extensively in your testimony. Well, maybe I should start by
9 looking at the first addendum of the 19th of February, 2002. And just
10 briefly, you were asked by the OTP to re-run certain analysis of some of
11 the figures and also to answer the question as to whether or not the
12 refugee flows in Kosovo resulted from the killings, if that could be
13 scientifically proven. And your answer basically was: No, it could not
14 be scientifically established on the basis of the data, but that addendum
15 did not affect any other findings that you made in your original report,
16 in the substantive killing and refugee flows in Kosovo. Have I summarised
18 A. Yes, you have.
19 Q. And the corrigendum of 15th of November, 2002, there you
20 re-plotted some of your calculations and graphs as the result of three
21 items of information that had changed, and the result of these
22 re-plottings caused very minor and insignificant changes in your
23 calculations, and therefore did not in any significant way affect the
24 conclusions and analyses in the substantive report. Have I summarised
25 that correctly as well?
1 A. Yes, although in the November 2002 report the changes that we
2 detected seemed to be potentially stronger than the results we had
3 presented in the earlier report.
4 Q. What do you mean by "potentially stronger"?
5 A. I believe that we noted in the second page that there were --
6 rather than there being no relationship between NATO air-strikes and
7 killings and migration, we found that there -- that NATO air-strikes could
8 be interpreted to have caused a reduction in killings of Kosovar
9 Albanians. We interpreted that result as coincidental, as the result of
10 being asked to include the entire period from early March until late May
11 in the analysis. Earlier we had -- in our previous work we had cut off
12 the analysis in early May. By extending the analysis through late May, we
13 found that there was some -- what we thought unlikely statistical findings
14 that the -- that NATO air-strikes could have negatively impacted killings
15 simply because there were many more NATO air-strikes in May than there had
16 been earlier. Finding many more in May would then cause this negative
17 correlation. However, the change that we found would have tended toward
18 strengthening the original rejection of NATO as a possible cause of the
20 Q. Thank you.
21 JUDGE BONOMY: Was -- can I ask you a question about that which
22 puzzled me?
23 THE WITNESS: Sure.
24 JUDGE BONOMY: Was NATO ever a plausible cause of the killings? I
25 can understand that NATO bombing might explain people being driven from
1 their homes, and I can understand that KLA killings might explain that.
2 But was NATO ever a plausible cause of people being killed? That's the
3 type of -- the people you identify as victims here who are Albanians.
4 THE WITNESS: Your Honour, I don't really know. We chose to
5 explore all the alternatives that seemed plausible at the time in order to
6 rule them out. We didn't want to exclude NATO, since that was part, at
7 the time, of the political debate.
8 JUDGE BONOMY: This -- your ultimate finding -- your later finding
9 which you say is coincidental was that the NATO air-strikes could have
10 been interpreted as causing a reduction in killings. That might make some
11 plausible sense without trying to make a judgement of it, it may be an
12 action that because it engages the Serbs, distracts them from killing
13 Albanians, that I can see. I'm trying to think about the factual
14 situation. But I can't imagine a factual situation where NATO bombing
15 could explain a surge of killings of Albanians.
16 MR. STAMP: May I refer Mr. Ball to --
17 JUDGE BONOMY: Certainly.
18 MR. STAMP:
19 Q. You referred to the hypothesis in respect of air attacks by NATO
20 at page -- I think it's page 2 of the substantive report.
21 A. Yes, that's the January 2002 report.
22 Q. Can you read hypothesis 2.2 and explain to the Court on what basis
23 did you use that hypothesis?
24 A. Yes. The hypothesis was written at air attacks by the North
25 Atlantic Treaty Organisation created local conditions that lead to
1 Kosovars being killed and leaving their homes. The NATO influence could
2 have either been direct because people were killed in air-strikes and
3 others fled or indirect because local Yugoslav authorities responded to
4 the air-strikes by killings Kosovars and forcing them from their homes.
5 So in a sense, there's a double possibility here. Either that the
6 air-strikes themselves were the cause of the deaths and then the survivors
7 fled, or that local authorities were motivated by the air-strikes to
8 either commit the killings and/or drive people from their homes. I'm not
9 in a position to say. I'm not an expert on the specific history of the
10 region. What we took as hypotheses were the three major armed groups that
11 were in conflict in the region and asked ourselves: Could any -- could
12 the actions of any of these three armed groups have caused the enormous
13 migration and substantial killings that were observed, and that was the
14 task we set ourselves.
15 Q. And could you also look at footnote --
16 JUDGE BONOMY: Can I just --
17 MR. STAMP: Sorry.
18 JUDGE BONOMY: -- complete this?
19 In that hypothesis, you take retaliation into account -- well, the
20 possibility of retaliation, but you don't do that in the earlier -- the
21 hypothesis in relation to the activity of the Kosovo Liberation Army.
22 Have I understood that correctly?
23 THE WITNESS: You have understood that correctly. From an
24 interpretive point of view, however, there would be no distinction between
25 hypothesis 2.1 and 2.2.
1 JUDGE BONOMY: Thank you.
2 MR. STAMP:
3 Q. You discussed in your -- can I call it the principal report. When
4 I say the principal report, I'm referring to the report of killings of
5 refugees in Kosovo from March to June 1999. At footnote 14 on page 12 of
6 that principal report -- perhaps I can read the text.
7 "The analysis considered the number of NATO air-strikes, as
8 reported by Yugoslav government sources." And the footnote is "Yugoslav
9 government was the primary proponent of the claim that NATO air-strikes
10 were responsible for the killings and refugee flow in Kosovo."
11 A. I'm sorry, where are you?
12 Q. I'm at page 12 of the principal report.
13 A. Page 12, I'm sorry.
14 Q. And it's footnote 14.
15 A. Yes. At the time -- well, sometime before this report was written
16 but during 1999, my team and I were actively following the Yugoslav
17 government websites in order to capture information about NATO air-strikes
18 in particular and then later KLA activity. And we found that there were
19 many claims to that effect, that the air-strikes were responsible for the
20 killings and refugee flow. It seemed to us at the time that this was a
21 relevant hypothesis. In retrospect, that may seem less obvious.
22 JUDGE BONOMY: No, no, that's -- yes, killings and refugee flow.
23 Refugee flow, yes; killings, I still find difficulty with. The -- and you
24 are relying on published material, a result of assertions by the Yugoslav
25 government. Does the inaccuracy, you know, the fundamental inaccuracy of
1 even the suggestion that NATO air-strikes could be responsible for the
2 large number of deaths that we're concerned about, does that affect this
3 at all?
4 THE WITNESS: I do not believe so because we disregarded those
5 claims, as we disregarded the claims by the various deponents or -- who
6 gave interview data about the alleged causes of their -- of the suffering
7 they witnessed. Instead, our interest in the on-line materials that we
8 gathered was simply in the dates and locations of the alleged air-strikes
9 and KLA activities. Similarly, with interview material, we were concerned
10 only with the dates and locations of the -- of the killings and the
11 migration process. It was our intent to make our substantive inferences
12 purely from the statistical patterns, rather than from any allegations by
13 witnesses or by other of the primary materials.
14 JUDGE BONOMY: Thank you. Mr. Stamp.
15 MR. STAMP: Thank you.
16 Q. Just to be clear about what you were saying earlier, in your
17 corrigendum you found that there was a negative correlation between NATO
18 bombing and population movements. But you did not draw any conclusions
19 from that. In other words, you did not draw any conclusions that
20 population movements or lack of population movement was the result of more
21 NATO bombings and vice versa?
22 A. No.
23 Q. In respect to the last addendum. As you indicated, you were given
24 some additional material by the OTP and you factored those into your
25 calculations and analyses. Can you elaborate a little bit upon the
1 results that you got when you factored in this new material. Firstly,
2 tell us in respect of killings what the new material was.
3 A. Yes. The material I believe that you're referring to starts on
4 page 6 and is subtitled "Additional Data from the ICMP, ICRC, and OMPF."
5 These sources were provided --
6 Q. ICMP being the International Community -- ICRC being International
7 Committee of the Red Cross?
8 A. Mm-hmm, yes. The ICMP is the International Committee of Missing
9 Persons, I believe.
10 Q. International Committee of Missing Persons?
11 A. And the OMPF.
12 Q. Office of Missing Persons in Kosovo.
13 A. Each of those three sources provided a list of deaths, of missing
14 people identified by the date on which the person went missing and the
15 place where the person was last known to have been seen. The people were
16 identified by name for almost all records. We did not have time or
17 resources to match all of these data sets to the existing four data sets.
18 There were only three months available to us to produce this material. So
19 we, instead, examined a descriptive analysis shown in figure 2, page 7 of
20 this report. There are three things I think that are worth highlighting
21 in figure 2.
22 First I would observe that the grey-shaded area or black-shaded
23 area apparently in the photocopy, but originally it was a grey-shaded
24 area, shows the high and low counts summed over two-day periods for the
25 collection of the ICMP, ICRC, and OMPF data sources. That is, we did not
1 sum them together, since they quite likely include many of the same
2 deaths. Instead, we simply took the high and low values from each of them
3 for each of those periods and plotted the high and low values in a cloud
4 that you can see is the grey area. Again, the reason we did this is that
5 we did not have the resources to deduplicate the three data sets. We
6 could not identify the repeated the records across them. That's, as I
7 mentioned earlier, a time-consuming task which requires a lot of
9 The scale of the ICMP, ICRC, and OMPF data is shown in the left
10 vertical axis. That is the top value, the maximum value reached in any
11 two-day period by the three sources is 439 deaths. The grey area is
12 overlaid by a black line, which is the original estimate we presented in
13 the 2002 report. It's labelled at its peak as the MSE estimate. Its
14 scale is shown on the right vertical axis. It has a maximum of 1.226
15 deaths in a single two-day period, which happens to be the first point of
16 the graph.
17 The core conclusion that I take from this graph reaches back to
18 the analysis that we drew in the 2002 report, which is that the pattern of
19 deaths and the pattern of migrations, but deaths specifically in this
20 graph, follow a three-phase rising and falling structure, rising in late
21 March to an early-April -- late-March/early-April peak, and then falling
22 again -- falling - excuse me - for the first time toward 0 on the night of
23 the 6th/7th April; remaining at a low period between the 7th of April to
24 the 10th of April, when they begin to rise again to a mid-April peak.
25 After the mid-April peak, the number of killings and migration that is
1 shown in another graph decline to a low period in late April. They rise
2 again to a series of peaks in late -- at the very end of April and early
3 May. After early May -- after the 11th of May, they continue at a very
4 low level until the conflict ends.
5 The principal conclusion from this graph is that the new data are
6 nearly identical to the old data. They are extremely similar in pattern,
7 and there is no reason to believe that adding these data would change the
8 results. In fact, it's quite likely that adding these data would increase
9 the precision of the results substantially, since they show such a very
10 close relationship.
11 Q. So how many data sources now have you -- are you now able to rely
12 on in respect to these patterns that you mapped in this graph?
13 A. This graph includes seven independent data sources, including the
14 interviews from Human Rights Watch, which are the smallest of the seven,
15 going on to the interviews conducted by the American Bar Association,
16 Central East European Law Initiative, and their Kosovar partners, the
17 Organisation for Security and Cooperation in Europe, and the exhumations
18 conducted by the first-round of exhumation teams provided to me by the OTP
19 in 2001. To those four sources, we've added these three, the ICMP, ICRC,
20 and OMPF.
21 This graph may be easily re-drawn, and I would be happy to provide
22 it to the Court, if it were interesting, showing all seven lines. It is a
23 bit of a mess, seeing seven lines on a single graph, but to my eye, they
24 show precisely the same thing with seven independent lines. So I found
25 this presentation much clearer.
1 Q. Thank you.
2 JUDGE BONOMY: It may actually be of interest to see the seven
3 independent lines. Can we take it that it's not possible to combine all
4 seven into one graph?
5 THE WITNESS: No, Your Honour, it's not possible to -- I mean
6 it -- I'm sorry. Let me ask -- let me clarify to make sure I understand
7 the question. I can put seven lines on one graph --
8 JUDGE BONOMY: I understand that. But what we're looking at here
9 is two separate figures, one based on four sources and one based on three.
10 THE WITNESS: Right.
11 JUDGE BONOMY: Can you do one based on all seven or is that not
12 something that would assist us?
13 THE WITNESS: I could certainly do so, but to do that would
14 require that I deduplicate the three sources that I've just received, the
15 ICMP, ICRC, and OMPF. In order to include those three sources with the
16 others, we would have to determine which of the deaths in each of the
17 three new sources are in common with each other and in common with the
18 four existing sources.
19 JUDGE BONOMY: Is that a weakness of the three new sources that
20 you haven't been able to distinguish the deaths that may be replicated
21 among them?
22 THE WITNESS: I think it's less a weakness in the data sources
23 than a weakness in the resources available to us to do this.
24 JUDGE BONOMY: Yeah, but that's a deficiency when compared with
25 the work that you've done on the four original data sources.
1 THE WITNESS: I was thinking about that this morning as I woke up
2 horribly jet lagged in my hotel room. There's nine hours' difference from
3 California. And what I realised is that the first graph in appendix 2 of
4 the substantive report addresses specifically that point. And so if I
5 could ask you to turn to page 36 in the January --
6 JUDGE BONOMY: I think what we'll do, if you don't mind is --
7 we've overrun already to the point where I think we're probably exhausting
8 the others who are serving us. So what we'll do is have our break at this
9 stage and return to this point after the break. The usher will show you
10 where you should wait while this break occurs.
11 THE WITNESS: Okay.
12 JUDGE BONOMY: And if you could please leave the courtroom with
13 the usher. Thank you.
14 [The witness stands down]
15 JUDGE BONOMY: And we shall resume at 20 past 11.00.
16 --- Recess taken at 10.52 a.m.
17 --- On resuming at 11.21 a.m.
18 [The witness takes the stand]
19 JUDGE BONOMY: Mr. Stamp.
20 MR. STAMP:
21 Q. When we broke you were speaking about a weakness or a deficiency
22 in the amount of time you could work with the last set of data sources
23 that you got in respect to the last addendum, and you were about to
24 explain the nature of this by reference to page 36 of the principal
1 A. Yes. As I mentioned earlier, the purpose of the estimation of the
2 total deaths was to consider whether the deaths we did not know about
3 would be in some way systematically different than the deaths we did know
4 about. So in figure 2 that we observed before, figure 2 of the most
5 recent report, the one submitted in January 2007, we saw this dark black
6 line marked "Multiple Systems Estimates." It's the one that I believe is
7 still on the screen. I'm not sure what screen I'm looking at, but it says
8 "Courtbook Sources." And if you compare that graph to figure 1 on page
9 36, it's figure 1 of appendix 2, page 36 of the January 2002 report, again
10 we see this very consistent pattern of the three-phase structure where
11 the -- there is a bulk of killings in the early period, late-March/early
12 April, that fall dramatically to a period between the 6th and 7th of April
13 until about the 10th, and then there's a rise again around the 10th to a
14 peak approximately the 15th to the 18th, and then it drops down in late
15 April and so forth.
16 What figure 1 shows is the deduplicated total for the four systems
17 we talked about earlier, that is, this is not an estimate. These are the
18 counts, the simple number of deaths reported to Human Rights Watch, the
19 OSCE in the exhumation data and in the American Bar Association and their
20 partners' testimonies. In some sense, this addresses the Court's
21 questions to me about the impact -- yes, that is the correct graph that's
22 now on the screen. Thank you.
23 This addresses the Court's question to me about what the change
24 introduced by the estimation might be, and the answer is very little.
25 There is very little change introduced by the estimation process. In
1 fact, the -- while we were greatly concerned about potential statistical
2 biases inside the four systems, once we had done the estimates, we found
3 relatively little bias, as evidenced by this comparison. There's a
4 variety of other tests that we can discuss if there's interest.
5 I think that most germane for this discussion is the striking
6 similarity between the line in figure 1 that's on the screen now and the
7 lines that are encapsulated by the grey area describing the ICMP, ICRC,
8 OMPF patterns. Again, what that means is that the new data, the data
9 that's just become available to my team now, is substantively identical to
10 the data that was available to us in 2001 when we did this work. I feel,
11 therefore, very confident that if we were to be given somehow additional
12 resources to do this to include these three sources into the four we have
13 now, the results would not change, except that the errors would decrease.
14 Our precision would improve, which is the way of all -- of all science.
15 If you have additional resources, you can repeat the experiment, you can
16 get more data. If you got it right the first time, your results won't
17 change but your errors will decrease.
18 Q. Thank you. I think you touched on, but could you briefly in
19 layman's terms, briefly, just explain the reasons for the difference in
20 the count for the two axes in figure 2 of the 2007 addendum. The vertical
22 A. Yes. I believe this is now on the right side of the screen, is
23 this the --
24 Q. Yes.
25 A. Yes. Okay. The axis on the left is showing the total numbers
1 represented in the grey area, that is the high and low values of the ICMP,
2 ICRC, and OMPF series. So, for example, on approximately the 28th of
3 March or perhaps the 1st of April - I'm not sure of the exact date - there
4 were 439 deaths reported in one of those three sources. That's the peak
5 shown in that grey area. Another of the sources showed considerably
6 fewer, but there's a range in there, and I think that it's important to
7 observe the range, because this is -- that's natural with data -- with
8 data that's collected ad hoc, as these were. It was not a probability
10 The axis on the right that has its maximum at 1.226 deaths shows
11 the range in the black line labelled MSE estimate at the mid-April peak.
12 This line is the same line as that presented in the earlier work, in the
13 2002 work, and is the estimated total number of deaths by two-day period
14 throughout the period 24 March to 11 May. Again, we did the estimate to
15 control for bias. Reference to figure 1, which we've just covered,
16 suggests that that concern may have been excessive, but statisticians are
17 very sensitive to the possibility of bias. So we -- but we expend a lot
18 of energy worrying about it.
19 Q. Could we move on quickly to figure 3 of the latest addendum. That
20 is dealing with the refugee flows, and briefly, if you could explain what
21 does figure 3 indicate in respect to the use of the new data from the
22 UNHCR in respect to refugee flows.
23 A. Yes. There are three series shown here -- or four series - excuse
24 me - four -- three series in solid lines, though they have slightly
25 different weights, and one series in a dashed line. The three series in
1 solid lines are the three series that were used in our original reporting
2 which first appeared in a report from AAAS, from my employer at the time
3 the American Association for the Advancement of Science entitled "Policy
4 or Panic" in which we analysed -- we presented the first analysis of
5 migration. The fourth line, the dashed line, comes from what purported to
6 be daily refugee counts compiled by the United Nations High Commission for
7 Refugees in Geneva, and these were provided to me by the Office of the
8 Prosecutor several months ago.
9 Now, there's a couple things that I think are worth observing
10 here. First, in many of the particulars, the series are very similar.
11 In -- there are some differences, but what's -- the crucial pieces of
12 agreement for the argument are the decline on 6/7 April, which is of
13 substantial importance, the relative low period between the 7th of April
14 and the 10th of April, the peak in mid-April, the decline in late April,
15 and then the series of small peaks later, in late April and early May.
16 That phase structure is intact in all three structure series.
17 However, I'm skeptical about the quality of the data from the
18 UNHCR Geneva because it has a few anomalies that suggests to me that
19 perhaps the data they have received may have been averaged over a series
20 of days. They may have received a block count for a week and then they
21 projected it backward by dividing the number by 4 or something. Seems,
22 for example, when we look at the flat line between about the 5th of April
23 and the 10th of April, well, that's implausible to me that exactly the
24 same number of refugees would cross every day. That's not really how
25 refugee movement works. Refugee movements ebb and flow, even a little
1 bit, but to have an identical number across a series of days is very
3 And so what's much more likely is that someone at some point said,
4 well, we've just gotten a new report, we know it covers this period, so in
5 order to create daily counts, we'll just divide by 5 or something and
6 present those as daily counts. When we compare in particular the UNHCR's
7 reports from Geneva to those reported both in Kukes, which was very near
8 the flow, or in Tirana where their press office was where I got the data,
9 it's clear that the data from Albania was much more detailed and
10 up-to-date or more current as time passed. It may be that later in the
11 process, the two series came together a bit more as UNHCR improved their
12 communication process between Albania and Geneva, but in the early period
13 I'm pretty skeptical about the Geneva data. On the whole, I do not
14 believe the new data has any substantive impact.
15 Q. Yes.
16 JUDGE BONOMY: Can you remind me what EMG is?
17 THE WITNESS: Yes. The Albanian government formed an Emergency
18 Management Group which was charged with trying to handle the influx of so
19 many hundreds of thousands of people into Albania. In addition to the
20 refugees crossing directly into Albania, Albania was also receiving a lot
21 of refugees being transferred from Macedonia. So the EMG was charged with
22 this charge. They did not begin work immediately. Their work really
23 began in ernest in mid-April/early -- mid-April. And at that time they
24 began releasing daily or even semi-daily reports of the number of refugees
25 crossing. When they did that there was a period of about a week or ten
1 days. We can refer to my published data, it's shown there, in which both
2 the EMG and the UNHCR continued to publish daily reports, but at some
3 point, they got together and said that's enough and UNHCR stopped
4 publishing those reports and then the EMG took over and did it.
5 JUDGE BONOMY: Thank you.
6 THE WITNESS: I use both series because they cover -- although
7 they overlap, they cover different parts of the period. They both drew on
8 same sources, however, which were the OSCE border observers who were
9 hand-counting and the Albanian border guards.
10 MR. STAMP:
11 Q. You said you do not believe the new data from UNHCR has any
12 substantive impact. Impact on what? Could I ask in respect to your major
13 findings, in respect to population flows and the three surges followed by
14 troughs, do they have any impact on that finding at all?
15 A. One might argue, although I think it would be on weak grounds, one
16 might argue that the trough between the 6th of April and the 10th of April
17 is not as low as it is in the UNHCR data as it is in the other data. Now
18 it's not clear to me that that would make a significant substantive change
19 in our argument and furthermore, as I argued earlier, I think that's based
20 on an artefact of the way the data was interpreted in Geneva. One would
21 be very hard-pressed to explain how refugee flow could appear from the
22 point of view of Geneva to have been identical for five days in a row;
23 that seems very unlikely to me.
24 However, that would be the only difference it makes. From all the
25 other points in our argument, the patterns are substantively identical.
1 There is, for example, a difference in very early April where the series
2 close to the ground, in particular, the Albanian guards at the border and
3 the UNHCR in Kukes and Tirana show a slight decline, whereas on the same
4 two-day period, the UNHCR shows a peak. Well, UNHCR Geneva shows dips
5 both before and after that. Again, that looks to me like they aggregated
6 the data from before and after and then post hoc reported it as a single
8 I think one would be hard-pressed to put more credibility in data
9 that was collected or was reported from farther away from the conflict
10 when it shows so many anomalies as the Geneva data does, relative to the
11 data that was captured at the border or very near to it, as was the case
12 with the Albanian guards' data, the UNHCR Kukes data and the EMG
13 government, not Albanian government data.
14 Q. Thank you. Could we move to figure 1, figure 1 which is on page
15 5. I'd like you to explain figure 1 in the context of the statement that
16 you make at page 4. I'll read it to show you where I want you to direct
17 your explanation.
18 "Both the 2002 and 2007 analyses used conservative coding rules in
19 order to give the KLA and NATO hypotheses the fairest possible test, that
20 is, if there were any possibility that the KLA or NATO could have caused
21 migration or killings, the municipality was quoted as inconclusive or
22 coinciding. The coding scheme that produced the categorisation in the
23 table above 1, that is table 1, which I will get to later, may have
24 yielded excessively conservative results for these reasons."
25 In that context, can you just briefly in layman's terms explain
1 figure 1?
2 A. Yes. Our objective is to consider in more depth the argumentative
3 device that was discussed at some length in my earlier testimony before
4 the ICTY, in which we discussed how we might interpret a statistical
5 pattern as a cause or not a cause of another pattern. And one of the
6 fundamental points that we argued at the time was that in order to be a
7 cause that the purported causal act had to precede the purported result.
8 So if NATO bombing, for example, would cause migration, it had to precede
9 it. If NATO bombing followed migration, it could not said to have caused
10 the earlier migration. However, the question is: How long would it be
11 between the NATO bombing and the migration before that purported cause
12 would have dissipated? And for the purposes of the earlier discussion, we
13 said it would either have to coincide saying it occurred in the same
14 two-day period or it would have occurred in the previous two-day period;
15 that is to say, 0 to 4 days in the past. If a bombing occurred 0 to 4
16 days in the past, it was considered to have coincided with a peak in
17 migration or killing.
18 If, however, the air-strike occurred more than four days in the
19 past, that is to say more than four days before a peak in migration or
20 killing, it was coded as inconclusive, which is that it could have
21 possibly been a cause but that we were skeptical. We would ask: Why
22 didn't people leave sooner?
23 What we find, looking at figure 1, is that we divide the peaks in
24 killing and migration between those that occur before the 14th of April or
25 after the 14th of April. And we then say of the peaks in killing and
1 migration that occur before 14th of April -- the 14th of April, which is
2 most of them, by the way, we find that nearly 60 per cent of them occur
3 with no prior NATO or KLA activity. A very small percentage on the order
4 of 10 per cent are inconclusive, which is to say there was KLA or NATO
5 activity but four or five days prior. And there's a intermediate fraction
6 on the order of 30 or 35 per cent which coincide, which is to say that KLA
7 and NATO activity occur within four days prior to the peak in killing or
9 After the 14th of April, we see a somewhat different picture.
10 There's a much greater proportion of inconclusive events, and this we
11 separate out to highlight, that as the conflict wears on, inconclusive
12 events, according to this definition, will necessarily accumulate simply
13 because there's more time in the past available for a NATO or KLA activity
14 to have occurred. This does not make that NATO or KLA activity a more
15 likely cause; it's simply that there's more time. If the conflict's been
16 going on for three days, it's impossible to have an inconclusive event
17 because our window is four days long. If the conflict is 20 days old,
18 then there are 20 minus 4, 16 days, in which an inconclusive event may
19 have occurred. By the time the conflict is 60 days old, we now have a 60
20 minus 4, 56 possible days, in which a conflict may have occurred -- an
21 inconclusive - excuse me - event may have occurred. So as time goes on,
22 we believe that this notion of inconclusiveness becomes spurious and what
23 we would like to do is raise to the Court's attention the weakness of this
24 kind of analysis. It is for this purpose that statisticians use the kind
25 of analysis we used in our original report, which is multiple regression,
1 because it takes these sorts of relationships into account.
2 Q. The result of this excessive conservativism which would result in
3 excessive inconclusive findings we see represented here in table 1, can
4 you comment on table 1. In a general sense and in layman's terms, does
5 the factors noted in table 1 coincide -- does not coincide inconclusively.
6 What do they indicate in terms of the cause that might have affected the
7 killing and population movement during the relevant period?
8 A. Yes.
9 Q. Do you understand what I'm asking?
10 A. Yes, I think so. As one examines the cells of table 1, there are
11 three values in each of the cells, three possible values. So considering
12 the upper left-most cell for the municipality of Decani, we find that the
13 relationship between the KLA and killing of Kosovar Albanians is
14 inconclusive. What this means is that the peak of killings in Decani is
15 preceded by one or more interactions between the KLA and the Yugoslav
16 authorities; however, those activities occur five or more days prior to
17 the peak of killings. That's what inconclusive means in this context.
18 If we move left -- rightward - excuse me - in the cell, the
19 word "coincides" indicates that the peak of killings in Decani, killings
20 of Kosovar Albanians, coincides with a NATO air-strike that occurs 0 to 4
21 days prior to the peak of killings. If we move to the end of that row, we
22 see that if we look at the peak number of migration, people migrating out
23 of Decani, we find that there is no prior NATO air-strike. There is no
24 NATO attack that precedes that peak. So those are the three possible
25 values available in the cells of this table.
1 As we observe in our analysis, peaks are not the only time that
2 people move. There are in some cases smaller peaks or plateaus, and so we
3 do not recommend to the Court that this kind of analysis of simple peaks
4 be overinterpreted; rather, the broad patterns of rising and falling
5 should be the focus because peaks can be surrounded by other similar peaks
6 or they can be isolated alone in a field of 0s.
7 For examples of that, I recommend that people review the 29 graphs
8 which follow at the end of this report, where we show municipality by
9 municipality killings, migrations, KLA activity, and bombing, and we see a
10 bewildering array of possible combinations. The combinations are, indeed,
11 bewildering even to me, and I've been looking at these data for more than
12 six years. That's why statisticians reduce them to more direct analyses.
13 Q. If would you look at this table, table 1 overall, is this
14 consistent with or inconsistent with a pattern of killings or migration
15 being caused by KLA activity or NATO activity -- NATO bombing?
16 A. I believe that for many municipalities it is inconsistent; for
17 other municipalities it is inconclusive. I do not consider it an adequate
18 test of the hypotheses. We have done it because in prior questioning it
19 was a source of considerable discussion, and so for purposes of complete
20 transparency we have presented it here.
21 Q. Okay.
22 [Prosecution counsel confer]
23 MR. STAMP:
24 Q. When you say it is inconclusive, can you explain that a little bit
25 further. I understand when you say it is inconsistent with the hypotheses
1 that the refugee movements and killings could have been caused by NATO
2 bombing or KLA activity, but when you say it is inconclusive --
3 A. Okay.
4 Q. -- just elaborate a little bit on that. Is it something that --
5 or would the table indicate that the movements are not related or are
6 unlikely to be related?
7 A. I think this is -- the notion -- let's see. How can I explain
8 this? Finding that patterns coincide with each other is a seductive -- is
9 a seductive finding. What is necessary in statistical argument is not
10 simply that patterns coincide when they both occur, but rather, that they
11 also coincide when neither occur. This analysis only allows us to examine
12 the former.
13 So, for example, as a metaphor, we could say that it is true that
14 on every 24-hour period that an air-strike occurred in Kosovo during this
15 period, the sun rose. We would find a perfect coincidence between the
16 sun's rising and an air-strike occurring. However, there were many days
17 on which the sun rose and there was no air-strike. Therefore, it's clear
18 that the sun does not cause the air-strike; rather, they are irrelevant
19 and spuriously related. It is unfortunate that analysis such as that
20 presented in table 1 is unable to distinguish between the spurious
21 coincidences, such as that I describe in the rising sun metaphor, and real
22 patterns such as those that would be found in the regression analysis
23 presented in the original report in 2002.
24 As an example, we can consider Pec. Pec in table 1 on page 3
25 suggests that for all four of the relationships of interest, there is a
1 coinciding relationship between NATO's potential impact on killing --
2 excuse me - KLA's on killing, NATO on killing, KLA on migration, and NATO
3 on migration. If we refer -- I'm sorry to whoever is creating the images.
4 I'd like to turn to page 18. At the top there's a graph of all four
5 patterns in Pec. If we expand now just to go to the top. Thank you.
6 That's perfect. We can see that there is a sequence of KLA activity
7 throughout the period, some of which is followed by migration or killing,
8 some of which is not. Similarly with killings, some KLA activity is
9 followed by killings; in other cases the killings precede the KLA
10 activity, such as the peak at the end.
11 Bombings follow a similar pattern. In this case table 1 would, in
12 my opinion, mislead us to thinking that there was a relationship between
13 killing and migration and the KLA and NATO activity, whereas a more
14 detailed examination shows that, rather, Pec was simply a very active -- a
15 very active site of both NATO bombing and KLA activity and, perhaps
16 coincidentally or perhaps not, migration and killings.
17 Q. Very well. Now, the conclusion from that, I take it, is that the
18 use and analysis of the new additional information does not in any
19 significant way affect the patterns that you found in your principal
20 report of 2002?
21 A. That is correct.
22 Q. And to a great extent, the data used confirm the accuracy -- the
23 new data confirm the accuracy of the data used and estimations made in
24 respect to the principal report?
25 A. The new data, if you are referring to the data used for killings
1 that we discussed -- I'm sorry. I've lost the page. It's probably page
3 Q. ICRC, ICMP, OMPF.
4 A. Correct. If you are referring to those data, yes, they, I
5 believe, confirm the validity of our earlier analysis. Validity,
6 precision, and accuracy are terms of art in statistics and so -- that have
7 slightly different meanings. So for us, we would say they confirm the
8 validity of the earlier analysis, yes.
9 Q. And --
10 JUDGE BONOMY: Dr. Ball.
11 THE WITNESS: Yes.
12 JUDGE BONOMY: Sorry to interrupt, but there's something I have
13 been trying to work out here which I can't.
14 THE WITNESS: Yes.
15 JUDGE BONOMY: Just a little while ago you said, when you were
16 referring to that table 1, that for many municipalities it was
17 inconsistent and for others it was inconclusive. I do not consider it an
18 adequate test of the hypothesis. We have done it because in prior
19 questioning it was a source of considerable discussion. Now, what is it
20 that you've done --
21 THE WITNESS: Yes.
22 JUDGE BONOMY: -- that you don't consider to be an adequate test
23 and please try to give me a brief answer to that.
24 THE WITNESS: Sure. So I think that if we can refer to the graph
25 on Pec, it provides an excellent example. If we can go back to that on
1 page 18, because Pec is an especially complicated place. Okay. You'll
2 notice that both of the lines on the table have a series of different
3 peaks so that for migration we see a very -- a strong, early peak which
4 turns out to be the dominant one and then two later peaks, one immediately
5 before the 7th of April and one late in May, both smaller peaks. The
6 analysis in table 1 is focussed only on that first peak. It treats all
7 the other peaks as though they're meaningless, as though they're
8 irrelevant, and it does so because of the nature of this analysis where
9 we're examining the coincidence of the peak as though there's only one,
10 but, in fact, there are three.
11 In other graphs, if we were to examine other municipalities, the
12 relationship between peaks is even more complex, and we could go into
13 those in detail if it were of interest to the Court. But I think we can
14 examine the line for killings in Pec and find an equally complex story,
15 where we have a series of small peaks in early April, a larger peak in
16 mid-April, and then the largest peak, actually, at the very end of April.
17 The problem with the analysis in table 1 is that it can only take
18 into consideration the dominant peak, the single highest peak, rather than
19 taking into consideration the series of peaks which, in my opinion, better
20 represent the pattern in any given location, in any given specific place.
21 JUDGE BONOMY: What is it that's in table 1 that causes us
22 difficulty? I'm sorry. I'm lost on this completely at the moment.
23 THE WITNESS: Yeah.
24 JUDGE BONOMY: I do not understand this.
25 THE WITNESS: That's in table 1.
1 JUDGE BONOMY: I'm not understanding what it is about table 1 that
2 confines it to one peak and why, indeed, you did this at all. I just
3 don't understand what's going on here. I'm sorry about this, but I'm
4 completely baffled by this.
5 THE WITNESS: No, my apologies for that. I -- this was actually
6 originally requested of me in the -- in my prior testimony here, and so we
7 have presented it for completeness since there was no additional
8 opportunity to present it in the previous case --
9 JUDGE BONOMY: What you're saying is you're presenting something
10 misleading for completeness and I'm having difficulty with this concept
11 why one would do that in the first instance.
12 THE WITNESS: Well, I guess only because I was asked. I don't
13 really agree with it and I would prefer to look at a more complex measure,
14 such as that presented in the second appendix of my first report; I
15 believe that is the measure which takes into consideration all of this
16 variation and all of the necessary complexity to understand the
17 relationship among these four patterns.
18 JUDGE BONOMY: [Microphone not activated]
19 THE WITNESS: Which one?
20 THE INTERPRETER: Microphone, please.
21 JUDGE BONOMY: The same report, the page immediately before the
23 THE WITNESS: Okay.
24 JUDGE BONOMY: And in the largest paragraph there which starts "In
25 the first analysis ..."
1 THE WITNESS: Mm-hmm.
2 JUDGE BONOMY: I see in there that this is in fact confined to one
3 peak only, and that's the bit I was missing I think.
4 THE WITNESS: Yeah.
5 JUDGE BONOMY: That I assumed that this was simply a repetition of
6 the same type of analysis as before, but this is confined to only one peak
7 for each municipality. That's the difference.
8 THE WITNESS: This is the same as that presented in the original
9 report in figures -- let me give you the precise cites. The original
10 report showed these as figures 8 and 9 in the 2002 January report. And we
11 presented these after we had shown the original patterns and just prior to
12 the figures that we believe are conclusive, figures 10 and 11. And this
13 is in the 2002 -- excuse me.
14 JUDGE BONOMY: In the report, your latest report --
15 THE WITNESS: Yes.
16 JUDGE BONOMY: -- you deal with how you proceeded before at the
17 foot of page 1 and into page 2.
18 THE WITNESS: Mm-hmm.
19 JUDGE BONOMY: Now, is that an example or is that a process
20 different from the one set out in the paragraph we're looking at further
21 down the page so that it would identify various peaks? Or is that again a
22 description of the same exercise confined to one peak carried out in 2002?
23 THE WITNESS: The latter, Your Honour. It is confined to one
24 peak. We were asked specifically to explore figures 8 and 9, which are
25 confined to one peak. We have a very brief comment on this weakness in --
1 on pages 5 and 6, but it is very brief.
2 JUDGE BONOMY: When I read this report, I -- where does it say
3 that this is unsatisfactory? I've read this thinking this is a genuine
4 update of your conclusions, but I'm now getting from you that this is an
5 exercise that you're not happy with. Now, where do I see in this report
6 that you're unhappy about this?
7 THE WITNESS: I think that if we read the section on page 5, which
8 I'm looking for now. I will point it to you directly. When we stress
9 over and over again the excessively conservative nature of this report,
10 that is what we are saying. We are saying that we believe that this
11 overstates -- that this analysis overstates any potential relationship
12 between NATO and KLA activity and potential killing and migration.
13 JUDGE BONOMY: Sorry. And I really hadn't got from this the
14 message that this was unsatisfactory, but thankfully it's been clarified
15 in the oral evidence.
16 Mr. Stamp.
17 MR. STAMP:
18 Q. The primary task of this report was to incorporate the -- or
19 compare the additional data that you were presented with to the earlier
20 data you used and I understand from what you are saying is that the
21 results of that analysis does not in any way change the analyses and
22 conclusions in your principal report of 2002?
23 A. That is correct.
24 Q. Without going over that report in detail -- but turn to it and
25 just refresh briefly. The report at figure 2, I think, if you could just
1 briefly elaborate on that in layman's terms briefly, because remembering
2 that you have done so to some degree before. It indicates that there are
3 three distinct surges or patterns in the killings and refugee flows, and
4 these distinct patterns in the killing flow also coincides with the
5 distinct patterns in the refugee flows.
6 A. That's correct.
7 Q. And that is -- is indicated in figure 2. Do you have that before
9 A. Yes.
10 Q. And in figure --
11 JUDGE BONOMY: In case you're in doubt, I have no difficulty
12 following that. It's the simple -- the simple part I think are -- if
13 anything here's simple, that part I had grasped.
14 MR. STAMP:
15 Q. And that part I take it is the pith and gravaman of your findings
16 in respect to the patterns?
17 A. Yes, this is the basis for our conclusion in the first hypothesis
18 that there is likely a common cause of both killing and migration, which
19 was the basis for the subsequent analysis.
20 Q. And at figures 10 and 11 you used those graphs to demonstrate the
21 conclusions that the killings and refugee flows were unrelated to NATO
22 bombings or KLA activity.
23 A. More precisely, what we tried to show in figures 10 and 11 is that
24 if we control for the effect of what is known about KLA actions, KLA
25 interactions with Yugoslav authorities and NATO air-strikes, the
1 underlying pattern of killings in figure 10 and refugee flow in figure 11,
2 those underlying patterns remain substantially the same. That is, if we
3 used the statistical relationships between KLA activity and NATO activity
4 and killings in figure 10, we are unable to explain the pattern. We can
5 only remove a small amount of its shape. The remainder of the shape, the
6 residual, in statistician's talk, shows the pattern after the KLA and NATO
7 activity have been controlled. Since the patterns are the same, our
8 interpretation would remain the same and we would discard the hypotheses
9 that either the NATO or KLA, together or in combination in this case, can
10 explain the pattern and consequently could be plausible causes of the
11 killings. Figure 11 is very similar, although the pattern is slightly
12 less clear. We also in the appendices present these analyses by region,
13 for the east, west, south, and north. Or maybe not. I'm sorry. Let me
14 look for them. Mm-hmm. No, they're presented only as tables, not as
15 graphs. My apologies.
16 Q. In your report at page 15 --
17 A. I'm sorry, figures 20 and 21 show them by -- show the killings and
18 residuals by time and region. Yes. Pages 58 and 59 -- or -- sorry. 59
19 and 60, figures 20 and 21. Thank you.
20 Q. The report at page 15 --
21 A. I'm sorry?
22 Q. In your report at page 15 --
23 A. Yes.
24 Q. -- section 5.4, you speak of a circumstantial link between
25 Yugoslav Army activities and the observed pattern of killings and refugee
1 flow. The extreme decline in the number of killings and refugee flow
2 observed in the period of 6 and 7 April coincides with the unilateral
3 cease-fire by the Yugoslav authorities in recognition of Orthodox Easter.
4 Can you elaborate upon that very briefly?
5 A. Yes.
6 Q. What I want to know is the extent to which, in statistical terms,
7 is there an indication of a causal link.
8 A. There is no indication of a causal link. Let me be quite clear
9 that observational statistics do not permit an affirmation of cause.
10 Observational statistics are at their limit disconfirming hypotheses or
11 rejecting hypotheses. As we have done here, rejecting the hypotheses of
12 NATO or KLA causality with respect to killing or migration. However, we
13 observe coincidences, and coincidences may be evocative and helpful for
14 other determinations of fact.
15 In the case alluded to here, in section 5.4, we note that on the
16 evening of the 6th of April the Yugoslav government declared a unilateral
17 cease-fire in recognition of Orthodox Easter. This cease-fire was not
18 respected by NATO or the KLA. They increased their operations, doubling
19 and tripling them, according to the databases published by the Yugoslav
20 government in particular, although with -- supplemented by other sources
21 as we discussed in our appendices. What is interesting to us and what
22 I've mentioned several times earlier this morning is that during the
23 period of the cease-fire, from the early morning hours of the 7th of April
24 until Saturday, the 10th, killing and migration in most regions declines
25 to its lowest point throughout the conflict to that point. It declines in
1 many places to 0, in other places to near 0.
2 So the Yugoslav government declares a cease-fire, killing and
3 migration go down towards 0, NATO and KLA activity go up, doubling and
4 tripling, but no activity -- no more killing and migration. On Saturday,
5 the Yugoslav government announces that because NATO and the KLA are
6 ignoring the cease-fire, they will resume operations. They resume
7 operations and we observe that killing and migration begin to increase
8 again. Not immediately to a peak. The peak occurs a few days later
9 sometime between the 13th and the 18th depending on which of the series we
10 observe, but it does begin increasing on Saturday the 10th. This is a
11 coincidence. This is not proof that the Yugoslav government was the agent
12 of killing and migration, but it is, I believe, an important statistical
13 coincidence which could be useful in the balance of evidence.
14 Q. Thank you very much, Doctor.
15 [Prosecution counsel confer]
16 MR. STAMP: Thank you very much, Your Honour. There is nothing
18 JUDGE BONOMY: Thank you, Mr. Stamp.
19 Mr. O'Sullivan.
20 MR. O'SULLIVAN: Your Honour, the order will be: General Ojdanic,
21 General Pavkovic, Mr. Sainovic, General Lazarevic, General Lukic, and
22 Mr. Milutinovic.
23 JUDGE BONOMY: Mr. Sepenuk.
24 MR. SEPENUK: Yes. Thank you, Your Honour. Before I start, Your
25 Honour, a number of my colleagues are going to cross-examine Dr. Ball, and
1 we'd like to ask about the rules on cross-examination here. I'm not sure
2 that the rules apply in any event to an expert, but if they do, we'd ask
3 Your Honour for some leeway for a number of reasons. First, Dr. Ball, in
4 the original 65 ter estimate, was going to be a six-hour witness.
5 Secondly, there had been additions to his report. And thirdly, unless
6 we've misinterpreted the Trial Chamber's ruling respecting Dr. Ball's
7 testimony, we seem to get some indication from that that the Court might
8 provide additional time for us on cross-examination.
9 [Trial Chamber confers]
10 JUDGE BONOMY: We think the indications that have been given of
11 the likely lengths of cross-examination are reasonable in the
13 MR. SEPENUK: Thank you, Your Honours.
14 JUDGE BONOMY: Mr. Sepenuk.
15 MR. STAMP: Can I just indicate, just briefly, that I -- Dr. Ball
16 traveled a long distance to get here and got here not in the best of
17 health. So I would ask the Court to grant him whatever dispensations he
18 might need in terms of time maybe just to step out for a while. He could
19 indicate if he needs to.
20 JUDGE BONOMY: You will do that, will you?
21 THE WITNESS: Yes.
22 JUDGE BONOMY: Thank you.
23 Mr. Sepenuk.
24 MR. SEPENUK: Thank you, Your Honour.
25 Cross-examination by Mr. Sepenuk:
1 Q. Good afternoon, Dr. Ball. I'm Norman Sepenuk and I'm an attorney
2 for General Ojdanic.
3 Before I get to your statistical analysis, I would like to spend a
4 small amount of time on background and context questions so we can better
5 understand your testimony, and I want to start with your report, your 3
6 January 2002 report, Prosecution Exhibit 1506, which is killing and
7 refugee flow in Kosovo March/June 1999, 3 January 2002, and this report
8 was presented to the International Criminal Tribunal for the former
9 Yugoslavia. Is that correct?
10 A. Yes.
11 Q. And I notice that on the front of the report there are two
12 particular organisations which are mentioned, the American Association for
13 the Advancement of Science and the American Bar Association Central and
14 East European Initiative. And can I ask you why these two particular
15 organisations appear on the front cover?
16 A. Because they were the sponsoring organisations for the production
17 of the report. At the time, I worked at the American Association for the
18 Advancement of Science. My co-author Wendy Betts worked at the American
19 Bar Association/CEELI. A number of the other authors were associated as
20 consultants with those two organisations and we had our organisational
21 homes there.
22 Q. Thank you. And you note on the second page of the report, though,
23 that these organisations -- that the opinions expressed in the report are
24 not the opinions of the organisations but rather you and your fellow
25 authors. Is that correct?
1 A. That's correct.
2 Q. And in your acknowledgement to the report, you state: "US
3 government support was essential to the project and support was provided
4 among others by the United States Department of State and the US Agency
5 for International Development." Is that true?
6 A. That support was provided to ABA/CEELI, that's correct.
7 Q. Thank you. And in your acknowledgement you also state that the
8 report would not be possible without the collaboration of a number of
9 organisations and individuals. I won't name them all.
10 A. Mm-hmm.
11 Q. But I notice among them are the Human Rights Watch, the
12 Organisation for Security and Cooperation in Europe, and the International
13 Criminal Tribunal for the former Yugoslavia; correct?
14 A. That's correct.
15 Q. And in your acknowledgements you also mention that your earlier
16 report which you referred to in your direct testimony published in 2000
17 called "Policy or Panic, the Flight of Ethnic Albanians" that -- and you
18 mention also a number of individuals and organisations who contributed to
19 that project and one of them is a Mr. Fred Abrahams who worked for Human
20 Rights Watch. Is that correct?
21 A. That's correct.
22 Q. And as a matter of fact, as I understand it, it was Mr. Abrahams
23 and a gentleman named Eric Stover who originally invited you to go to
24 Albania in mid-April 1999 to undertake the project?
25 A. They invited me considerably earlier and independently of each
1 other, but, yes, I ended up there in early April.
2 Q. And in "Policy and Panic" in your acknowledgements in "Policy and
3 Panic" you state, "Fred Abrahams from Human Rights Watch and Eric Stover
4 of the Human Rights Centre of the University of California Berkeley
5 originally invited me go to Albania in mid-April 1999. Subsequently, they
6 introduced me to the project's co-director Fron Nazi, East-West Management
7 Institute for the Policy on Legal Studies. Fred, Eric and Fron oriented
8 me to the region, told me what I needed to read --"
9 THE INTERPRETER: Please, if the speakers could slow down while
10 reading. Thank you very much.
11 MR. SEPENUK: I'll slow down.
12 Q. "-- provided many feedbacks, helpful feedbacks on many drafts and
13 gave me access to their interviews for confirmatory qualitative evidence."
14 Correct, sir?
15 A. Correct.
16 Q. Is it fair to say that at least in these early stages of your
17 research, Mr. Abrahams, for example, was sort of a mentor to you?
18 A. No, Dr. Scheuren was a mentor to me. I'm a statistician. Fred's
19 a human rights analyst. Fred and Fron and Eric showed me information,
20 provided data, told me stuff about Albania that I could use, but mentor is
21 too strong a word.
22 Q. All right. But at least you respected Mr. Abrahams's factual
23 knowledge of the area and his views on various subjects?
24 A. I did.
25 Q. So as I understand it, you went to Kosovo in mid-April 1999 to
1 gather data for a potential report.
2 A. I went to Albania originally to figure out if there was a report
3 that could be done. We found one, but at the time it was pretty up in the
4 air. We weren't sure what we would find.
5 Q. Okay. Fair enough. And how long did you stay in Albania in the
6 early stages?
7 A. In the first visit about two weeks, about two weeks or a day or
9 Q. And you went back when?
10 A. I went back to the US in early May.
11 Q. And then back to Albania?
12 A. About two and a half weeks later.
13 Q. Okay. And then what? Just give us very briefly again the
15 A. I began recruiting a team of Albanians and Kosovar Albanians who
16 would act as a survey group. My original intent was to build a team of
17 people who would survey refugee camp residents on a probability sampling
18 basis, that is, we would sample probablistically households from tents and
19 then interview them about their experiences before and after their
21 Q. And if I can just interrupt you briefly.
22 A. Mm-hmm.
23 Q. I'm trying to get some timing here. At least you were still doing
24 your research, let's say, in July of 1999; correct?
25 A. No, we were long done by then.
1 Q. You were long done by then?
2 A. Yeah.
3 Q. Okay.
4 A. Because when we realised that NATO was going to intervene, we had
5 to radically restructure what it was we were doing. We weren't going to
6 be interviewing people in camps, because the camps were going to be empty.
7 Q. But I take it by that time, let's just say July 1999, you hadn't
8 yet reached any conclusions?
9 A. None.
10 Q. Okay.
11 A. I had actually -- at that point I was not at all sure what the
12 project was going to look like. I didn't--
13 THE INTERPRETER: Interpreters kindly request for both speakers to
14 slow down.
15 THE INTERPRETER: Could the speakers make pause between question
16 and answer for interpretation.
17 MR. SEPENUK:
18 Q. [Previous translation continues]... is it fair to say --
19 A. I'm sorry. May I -- I was interrupted by the interpreters. May I
20 complete my answer?
21 Q. Certainly.
22 A. We had, really, very little idea even what data we had collected,
23 because things were very hurried in the last weeks of June, mid and last
24 weeks of June, and so we had no interpretation whatsoever of that material
25 in early July.
1 Q. Right. And the only thing I'm trying to get at is, at least by
2 that period, let's say July of 1999, you hadn't yet reached any
3 conclusions based on your statistical analyses?
4 A. No.
5 Q. But is it also fair to say that aside from not reaching any
6 conclusions based on your statistical analysis, you had already reached
7 the conclusion at this early stage that the Serbian defence establishment
8 had a plan to expel ethnic Albanians from Kosovo; hadn't you already
9 reached that conclusion, Dr. Ball?
10 A. That was a working hypothesis, but by no means a conclusion.
11 Conclusions are reached by data and not by anything else.
12 Q. Okay. And what was your working hypothesis?
13 A. My working hypothesis was that there had been some kind of process
14 that had driven people from their homes. What would that be? Clearly,
15 one of them was that there had been a process of ethnic cleansing. Others
16 were that NATO bombing or KLA activity had caused the migration and the
18 Q. Wasn't that a --
19 JUDGE BONOMY: Mr. Sepenuk, you really both need to --
20 MR. SEPENUK I'm sorry.
21 JUDGE BONOMY: -- allow time, because Mr. -- Dr. Ball is answering
22 you very quickly.
23 MR. SEPENUK: Thank you, Your Honour. I will.
24 JUDGE BONOMY: We need some sort of gap between the question and
25 answer to enable the interpretation to catch up. That's why you had that
1 problem a moment ago.
2 THE WITNESS: Yes, Your Honour.
3 MR. SEPENUK: Thank you. I'll be very much aware of that.
4 Q. When you said a working hypothesis, what I'm asking you is:
5 Hadn't you already concluded that there was a plan by Serbian forces to
6 expel ethnic Albanians from Kosovo before you ever even started to do your
7 statistical study?
8 A. Okay. I'm going to read your question on the screen here, and
9 you -- if I understand correctly, you've asked: Had I already concluded
10 such a thing. And I can tell you: No.
11 Q. Okay. Then let's look at a transcript of a radio programme that
12 you had.
13 A. Mm-hmm.
14 Q. In July 9th, 1999.
15 A. Mm-hmm.
16 Q. It was radio talk, "Talk of the Nations Science." Remember?
17 A. Yes, I do.
18 Q. Iris Flatow --
19 A. I do remember.
20 Q. -- was the moderator. And you appeared on that program?
21 A. I did, and we talked about Operation Horseshoe.
22 Q. Well, that's what I'm trying to get at, sir. You haven't
23 mentioned that yet. And let me go on, please.
24 A. Mm-hmm.
25 Q. And in that programme after you had said that you hadn't yet
1 concluded the results of your statistical analysis, you said:
2 "What's interesting is that the Yugoslav defence establishment
3 made public a plan for ethnic cleansing. Well, it wasn't a plan about
4 ethnic cleansing. It was a plan, a strategic plan, to attack the KLA, the
5 Kosovo Liberation Army, back in October/November of last year, the defence
6 establishment was very clear that this was what they were going to do. It
7 was called Operation Horseshoe."
8 Now, tell us -- first of all, you said it was made public. Tell
9 us about that.
10 MR. STAMP: This is not an objection, but since we're referring to
11 a document, could I just have the reference?
12 MR. SEPENUK: Certainly. 3D518 - excuse me - 3D518, page 377 in
13 the e-court.
14 THE WITNESS: Are we ready?
15 MR. SEPENUK: I'm waiting for it to come up on the screen. Is
16 there a problem with finding it? Here we go. You're right at the right
17 spot now.
18 Q. Do you see what I just quoted from?
19 A. Mm-hmm.
20 Q. So tell us at that time in July of 1999 what you knew about
21 Operation Horseshoe and how you happened to know it.
22 A. Well, there was a lot of talk about it among all the journalists
23 who were in Albania at the time, and I certainly heard their talk. This
24 influenced my formulation of hypotheses. I thought if these people are
25 right, what we should see would be a horseshoe. If you look at my
1 subsequent published material, you'll find that's not at all what we
2 found. That's a hypothesis which was rejected.
3 Q. And you say this was -- your understanding is this was made public
4 by the Yugoslav authorities? That's what you say in the radio programme.
5 A. I believe the German Defence Minister, if I recall correctly, make
6 remarks about this for which he was later in some trouble, but it was
8 Q. Did you later find out that that report was completely fake? Did
9 you find it out either yourself or from Mr. Abrahams who wrote about it in
10 Human Rights Watch?
11 A. I heard that it was fake, yes, but as you'll note, we didn't use
12 it. I mean, I think that this confirms the scientific process. Even when
13 you find something that's fake, when you go to the data and you don't find
14 it, you discard it.
15 Q. All right. Let's go on here what you said in the radio
16 programme. You said:
17 "Our hypothesis is that if ethnic cleansing matches, the pattern
18 of refugee flow matches that plan, that will be very interesting evidence
19 that, in fact, the refugee flow is a product of that plan. We can't say
20 precisely why. We'll then go back to witness and refugee testimony. We'd
21 go back to forensic evidence to make those links more tightly, but this
22 is, as I said, the sort of topographic map, as it were, of this process,"
23 correct. So you already had -- before you even had any results of this
24 statistical analyses, you had a so-called topographic map based on what
25 you now say is an erroneous notion about Operation Horseshoe which
1 started, even if it did exist, even if a plan did exist, against the KLA
2 and then transferred to a plan to expel all ethnic Albanians. That was
3 your notion at the time. Correct?
4 A. No, sir, that's not correct. In fact, you're taking a series of
5 subjunctive statements and making them indicative. What I said is that
6 one could imagine that we could have a hypothesis and, if these patterns
7 were to match, then things would be interesting. What I'm stating here
8 are a series of hypotheses to be investigated. I stated those hypotheses
9 before we had evidence to investigate them -- well, we had the evidence,
10 but it wasn't analysed. And when we finally did analyse it, as is clear
11 from the published record, we discarded those hypotheses because they were
12 inaccurate. I believe this is the process of all science, that all
13 science begins with hypotheses, Most of which end up being discarded in
14 the face of evidence.
15 Q. Dr. Ball, when you wrote --
16 THE INTERPRETER: Can the speakers please make pause between
17 question and answer? Thank you very much.
18 MR. SEPENUK:
19 Q. [Previous translation continues]... previously talked about,
20 didn't you already have in mind then or hadn't you already concluded then
21 that there was this pre-existing plan by Serb forces to expel ethnic
23 A. I'm sorry. Can you repeat the question?
24 Q. Yes. At the time you wrote "Policy or Panic," which is the year
25 2000, didn't you --
1 A. Mm-hmm.
2 Q. -- hadn't -- pause for your answer, sir. My colleagues are
3 continually reminding me of it, and I apologise for not doing that, to
4 you, to the interpreters, and to the Court, and I'll try from now on out
5 not to mess up anymore.
6 And my question is: At the time you wrote "Policy or Panic" in
7 the year 2000, isn't it fair to say that you had already concluded, aside
8 from any statistical analysis, that there was a pre-existing plan by Serb
9 forces to expel ethnic Albanians from Kosovo? Yes or no, sir, if you
11 A. When I published "Policy Or Panic" in 2000, I had found that we
12 could reject the hypothesis that NATO bombing was a plausible cause of the
13 migration. We at that time had very limited data on -- well, I had no
14 data on the KLA, and I had very, very limited data on killings, so our
15 sketchy conclusions about other topics were, again, inaccurate, but our
16 finding about the -- excuse me -- about bombing and migration has been
17 sustained throughout our later work.
18 Q. All right. Let me -- I'm sorry. Let me just read to you
19 specifically what you said in "Policy and Panic."
20 A. Sure.
21 Q. And this is 3D516 at page 232, and maybe you want to wait until it
22 comes up so you can see it.
23 MR. SEPENUK: 3D516, page 232. And if you just -- no, the second
24 paragraph, please right there.
25 Q. You say, if you could follow long: "The findings in this report
1 suggest that the Yugoslav military, police, and paramilitary apparatus may
2 have used the bombing as a broad political excuse to conduct a previously
3 prepared operation to evict ethnic Albanians."
4 That's what you said, correct?
5 A. Mm-hmm. That's correct.
6 Q. And that previously prepared operation, was that the erroneous and
7 non-existent horseshoe plan that you have previously testified about?
8 A. Well, let's observe two things about this sentence. First, we use
9 the term "may." We do not use the term "prove" or "did" and I think
10 that's important. So I'll pause that the interpreters can translate the
11 subjunctive. Second, I think it's essential to note that we do not
12 link -- or in this case, I, because I was the sole author of this
13 particular work, I did not link the claim about a previously prepared
14 operation to any specific operation, such as that you've just referred to.
15 So, no, by this time the whole operation horseshoe thing had been clearly
16 discarded, and what we're suggesting here is essentially the same kind of
17 language which appears later in my reports when I observe that there's a
18 coincidence on the night of the 6th and 7th of April that when Yugoslav
19 forces cease operations, migrations and killings also cease. That, too,
20 may suggest that Yugoslav forces are responsible. However, let me draw
21 your attention to the lack of the word "proof" here. As statisticians
22 working with observational data, we do not make such claims.
23 Q. Thank you, sir. We'll get to that later.
24 A. Okay.
25 Q. Thank you. Now, as you say -- you say that Mr. Abrahams
1 originally invited you in April of 1999, Mr. Abrahams and two others, to
2 go to Kosovo and undertake this project and one of the things that these
3 folks told you, they told you what to read. Do you remember that? That
4 was -- say yes. You have to say yes or no.
5 A. Yes.
6 Q. And did Mr. Abrahams suggest that you read his article of August
7 15th, 1958 [sic] in the International Herald Tribune in which he stated
8 that the first priority of the United States policy should be the
9 indictment of Milosevic. Did he direct you to that?
10 A. I suspect --
11 JUDGE BONOMY: Hold on. I can remember, Mr. Abrahams, and I doubt
12 if he was physically capable of writing that article in 1958.
13 MR. SEPENUK: Did I say that, Your Honour? No, no, August 5th,
15 JUDGE BONOMY: Thank you.
16 MR. SEPENUK: Thank you, Your Honour.
17 THE WITNESS: I do not specifically remember what he asked me to
18 read, but I think it probably was more Human Rights Watch reports than
19 media reports.
20 MR. SEPENUK:
21 Q. Did you read that media report or did you discuss his personal
22 views about Mr. Milosevic?
23 A. Certainly I discussed his personal views of Mr. Milosevic as I
24 have with dozens of other people, probably hundreds of other people,
25 including, in fact, Mr. Milosevic himself.
1 Q. And you of course shared Mr. Abrahams's views about Mr. Milosevic;
3 A. I really didn't know. I never worked on the Balkans and my
4 expertise was Central America and South Africa, so I really was more
5 interested in getting into the story than sharing anyone's views. Fred
6 and I were colleagues but didn't know each other that well when we started
7 working together in 1999.
8 Q. When you say you really didn't know, you had certainly well prior
9 to your testimony in the Milosevic case, you had formed some certain
10 opinions about Mr. Milosevic, had you not, sir?
11 A. Yes, but you're eliding three and a half extremely turbulent years
12 of research. In 1999, when I first went to Albania, I didn't really know
13 what I was getting into.
14 Q. But I'm talking now about 2001 --
15 A. When I'm writing --
16 Q. July 14th, 2001, when you were at that hacktivism conference?
17 A. Yes, I certainly do.
18 Q. Your remember that, sir? In Las Vegas, Nevada?
19 A. Yes.
20 Q. And it was presided over by Reid Fleming from The Cult of the Dead
22 A. I sure do.
23 Q. And The Cult of the Dead Cow being what?
24 A. They are a self-proclaimed hacktivist group.
25 Q. And were you on their board of advisors?
1 A. I'm on the board of advisors of their hacktivist group, not of The
2 Cult of the Dead Cow itself.
3 Q. And I may have missed it and if I did, I apologise to you. Is
4 that in your updated curriculum vitae?
5 A. No, I haven't done any work with them for years. Would you like
6 to know what they do, sir?
7 Q. Yes, please.
8 A. Right now they are trying to produce, and they have been for some
9 years, produce software that would enable dissidents in China to make
10 their views known throughout the world without repression by the Chinese
12 Q. And at that conference you said as follows: "Of course now on the
13 basis of crimes of war we have tribunals. Tribunals which try people who
14 have violated international humanitarian law and send them to prison. It
15 would be very nice if we had a round of applause for the extradition of
16 Slobodan Milosevic to The Hague last week." And then the record indicates
17 applause. "And I hope you're all excited as I am by the prospect of his
18 very pleasant and drawn-out trial which will begin in about six months."
19 You said that, didn't you?
20 A. I did indeed.
21 Q. And were you also excited about the prospect of going to The Hague
22 some months later to testify at his trial?
23 A. I had no idea at that time that I would be invited.
24 Q. You had no idea. Okay. And you said later on in that conference,
25 and I quote directly, you're talking about human rights, and you
1 said: "Let's work with those rights and then push them. If you want
2 more, let's keep pushing, absolutely, keep going, no reason to stop now,
3 we've got some momentum. We've got one of the worst bad guys ever in
4 jail, at least in my lifetime."
5 You said that, correct?
6 A. Mm-hmm.
7 Q. And you were referring to Mr. Milosevic, weren't you?
8 A. I do not recall. It could have been Pinochet but it's in the same
10 Q. Is that serious testimony?
11 A. What?
12 Q. Are you serious about this? You really don't think it was
13 Milosevic. I mean, the whole talk was in the context of Kosovo and
14 Mr. Milosevic. You're not saying that you weren't referring to
15 Mr. Milosevic there, were you?
16 A. I don't recall exactly who I was referring to. The talk was not
17 all about Milosevic. The talk was actually all about technology and I had
18 several sections of it about Guatemala and there were also several
19 sections, if you recall, about domestic US situations.
20 JUDGE BONOMY: Remind me if you would when Pinochet was in jail if
21 at all.
22 THE WITNESS: No -- he wasn't really in jail. I don't remember.
23 But this is an activist conference so I would have been making fairly
24 off-the-cuff remarks.
25 MR. SEPENUK:
1 Q. Off-the-cuff remarks but truthful remarks do you think?
2 A. Truthful remarks.
3 Q. And --
4 A. I'm willing for the record to say that I was referring to
5 Milosevic, yes.
6 Q. Yes, that's all I wanted to know. Thank you for that.
7 JUDGE BONOMY: Is that a suitable time for us to interrupt,
8 Mr. Sepenuk?
9 MR. SEPENUK: Yes, Your Honour, yes, Your Honour.
10 JUDGE BONOMY: Lunchtime, Dr. Ball. The usher will show you where
11 you can wait -- well, where you should go. We'll resume in about an hour
12 at quarter to 2.00.
13 [The witness stands down]
14 --- Luncheon recess taken at 12.47 p.m.
15 --- On resuming at 1.45 p.m.
16 JUDGE BONOMY: Judge Chowhan will be absent on other authorised
17 Tribunal business this afternoon, and we have decided that it's in the
18 interests of justice to continue with the sitting in his absence.
19 [The witness takes the stand]
20 JUDGE BONOMY: And this afternoon, every time the interpreter
21 interrupts to remind anyone about the failure to observe the pause, we
22 will deduct five minutes from the time for cross-examination.
23 Mr. Sepenuk.
24 MR. SEPENUK: Thank you, Your Honour, and I wasn't going to
25 mention it but I will mention it now. I'll mention another fault of mine
1 and that is when I switch from this mike to the mike in the middle, I
2 thought it was on the same channel, 4. It wasn't. It was on a different
3 channel, so I actually never heard the interruptions of the interpreter.
4 Now, I didn't hear it because I had it on the wrong channel. That's no
5 excuse, particularly, but that's the reason and hopefully that won't
6 happen again. Thank you, Your Honour.
7 Q. Dr. Ball, when we broke, you were frank enough to state for the
8 record that it was indeed Mr. Milosevic you were referring to at the
9 hacktivism conference when you said we've got one of the worst bad guys
10 ever in jail at least in my lifetime. Is that right?
11 A. That's correct.
12 Q. And I take it, though, that you didn't feel or don't feel that
13 your personal feelings about Mr. Milosevic in any way prevented you from
14 being objective in your testimony at his trial?
15 A. What I testified to then and now have to do with the statistics,
16 not with my opinions.
17 Q. Okay. And actually the main reason -- or at least one of the
18 reasons I'm asking you that was -- it's based on an article that appeared
19 in the report Wired News of July 16th, 2004, and the article's
20 entitled "Hacktivism and How it Got Here." It's Exhibit 3D519 at page
21 387. And if you can put that up, I'd appreciate it, so you can read along
22 with me. Yes. And it's beginning: "Ball spoke." Do you see that?
23 A. Mm-hmm.
24 Q. "Ball spoke about hacktivism at a hacker conference in the summer
25 of 2001 and during his talk made a disparaging comment about Slobodan
1 Milosevic, former president of Serbia and of the Federal Republic of
2 Yugoslavia. Ball later testified against Milosevic at the war crimes
3 trial in The Hague. When Milosevic cross-examined Ball, one of the first
4 questions he asked him was, Who was This Dead Cow Cult?" And you
5 said: "My under oath spin to Slobo," Slobo, I take it meaning
6 Mr. Milosevic, correct?
7 A. Yes.
8 Q. "My under oath spin to Slobo was that hacktivism is an opportunity
9 for engaged, young programmers to do cool and socially beneficial stuff
10 with their technical skill and curiosity instead of getting in trouble,
11 said Ball, and I actually believe that."
12 My question to you, Dr. Ball, is what did you mean by "your under
13 oath spin"?
14 A. People -- the question that underlay that was: How could you say
15 that the Cult of the Dead Cow was about cool and socially beneficial
16 stuff. And I said: I wasn't talking about the Cult of the Dead Cow, I'm
17 not part of the Cult of the Dead Cow. I'm not part of the hacktivismo
18 group which you and I clarified earlier. And I said: What I meant to
19 Slobo, which I used in the slang term spin, was that hacktivism is a way
20 to try to keep what I think is the most the brightest and most engaged
21 generation of my time out of trouble.
22 Q. My question is actually a little bit broader than that. Maybe I
23 didn't make it clear. I take it you agree with me when you're under oath
24 you don't spin?
25 A. That's correct.
1 Q. But you did say "my under oath spin." So at least at that point
2 apparently you did spin a bit; didn't you?
3 A. No. What I did is explain something in a slang term to a
5 Q. Okay. Thank you.
6 A. Mm-hmm.
7 Q. Now, I want to discuss with you some of the difficulties of
8 understanding a massive phenomenon that you've been testifying about such
9 as the ethnic Albanian refugee flow from Kosovo during the period March to
10 May of 1999. Now, I assume you would agree with me that purely
11 qualitative methods such as refugee interviews have their limits. I think
12 that can be answered yes or no.
13 A. Yes.
14 Q. And the first limit, the obvious one, is that there are far more
15 people involved than can realistically be interviewed, yes?
16 A. That's a resource question; that's not a yes or no question.
17 Q. All right. Then answer it however you want.
18 A. Okay. If you had sufficient resources and you had a small enough
19 context, you could interview everyone.
20 Q. Okay. But in this case probably unrealistic, correct?
21 A. Probably unrealistic in this case.
22 Q. And then second, there's the inherent problem of whether the
23 interviewees, the people interviewed, are telling the truth or telling
24 their story accurately and, as a corollary, whether the interviewers are
25 getting the information down correct. True?
1 A. That's actually pretty tricky. That's a tricky question because
2 there are many inaccuracies which can enter an interviewing process which
3 turn out to be irrelevant.
4 Q. All right.
5 A. So, for example, if you were taking a deposition of a witness
6 about a particular event, but the witness provided you with extensive
7 details about some other event and she were to get many of the details
8 about this other event wrong, that would actually not affect the validity
9 or the usefulness of the interview for your purposes since the event
10 you're interested in she got right. So it's actually pretty complicated.
11 The validity of an interview depends not on the interview itself but
12 rather the purposes to which you intend to put it.
13 Q. My question might have been a little simple-minded. All I meant
14 what that interviewers and interviewees are human beings and sometimes
15 they make mistakes? That's all.
16 A. Sometimes they make mistakes, of course.
17 Q. And third, and perhaps most importantly, is that these
18 interviewees -- there was no random sampling of the interviewees; correct?
19 A. No, that's unfortunately not correct. Some of the interviews were
20 conducted by random samples; others were not. Physicians for Human
21 Rights, which was the basis of much of our migration analysis, was
22 conducted by random sample. The first two components of the IPLS survey
23 that we conducted in camps, which we discussed earlier, particularly in
24 late May and early June, were conducted by random sample.
25 Q. [Microphone not activated]
1 Wouldn't you agree that a substantial number, though, of the
2 interviews, were not conducted by random sample? For example, Human
3 Rights Watch.
4 A. Well, Human Rights Watch was the smallest of the sets. I'm afraid
5 I don't know what you mean by a substantial number.
6 Q. Well, let me be more specific. You advised Mr. Abrahams at Human
7 Rights Watch on the statistical analysis violation chapter in their book,
9 A. Yes, I did.
10 Q. And there's a chapter in that book and it's Prosecution Exhibit
11 438, e-court page 444, which talks about limitations of the data and it
12 says as follows -- if we can put it up, Prosecution 438, page 444. There
13 it is.
14 And at the top, if you'll bear with me, in the first paragraph
15 there it says -- again, you were an advisor on this chapter. It
16 says: "Most importantly, Human Rights Watch did not randomly sample the
17 interviewees. On the contrary, researchers purposely sought out not only
18 the victims and witnesses of violations but specifically those with
19 knowledge of the most serious violations such as torture, sexual violence,
20 and executions. Therefore, these data cannot be extrapolated to general
21 findings for Kosovo as a whole."
22 You don't disagree with that statement, do you?
23 A. No.
24 Q. As a matter of fact, I represent to you that Sandra Mitchell of
25 OSCE also testified here to the same effect. I just mentioned that to
2 A. That would be logical.
3 Q. Now, as I understand it, you considered three -- because --
4 because there -- these interview techniques have these shortcomings, you
5 undertook -- that -- one of the reasons is that you undertook your
6 statistical analysis, correct, in view of these shortcomings of the purely
7 interviewing techniques, one reason that you proceeded was that reason,
8 shortcomings in interviewing techniques; therefore, you wanted to come up
9 with a more authoritative study based on statistical analysis.
10 A. I don't think "authoritative" is the appropriate word.
11 Q. What would it be?
12 A. We wanted to use a method which would be able to account for the
13 spaces or under-reporting which necessarily occur in the sorts of surveys
14 or interview processes that the four samples we had available to us.
15 Q. All right. Fine. And you considered three possible causes --
16 A. Mm-hmm.
17 Q. -- of the refugee flow?
18 A. Mm-hmm.
19 Q. And you're quick to point out in your report, that's Prosecution
20 Exhibit 1506, I'll read it, I don't think there's any dispute about it, so
21 I don't think we have to put it up on the board. You said in the report:
22 "It is beyond the capacity of statistical analysis to prove that
23 any of these hypotheses is the definitive cause of the patterns seen in
24 the two forms of violence."
1 A. That is correct.
2 Q. Okay. Now, let's go to the first cause here, and that is the NATO
3 air-strikes. Now, I understand you restricted yourself to just two
4 sources, at least when you first started to undertake that research, there
5 were going to be two sources, one, Yugoslav and press reports and press
6 reports from Yugoslavia, correct, which is ultimately what you used. And
7 you were also going to US Department of Defence information; correct?
8 A. That was an original idea, yes.
9 Q. And you, as I understand it, on the department -- there were two
10 things you were going to do. One, the Serb sources and, two, the
11 department of defence. As I understand it from your testimony in the
12 Milosevic trial, you didn't proceed with the Department of Defence data or
13 trying to get that data because as you put it: "My two phone calls to the
14 US Department of Defence went unanswered."
15 A. Mm-hmm.
16 Q. That was the extent of your trying to get information about NATO
17 bombings, two phone calls?
18 A. Yes. As a human rights activist, I do not have extensive contacts
19 at the Department of Defence.
20 Q. Well, who did you call there?
21 A. The public relations office I got from the internet.
22 Q. From the internet. And you never got a live person? What did you
23 do; leave a message?
24 A. No. I spoke to someone who said they would get back to me, and
25 they didn't.
1 Q. And then what did you do? You followed up on it?
2 A. Yeah.
3 Q. And that was the end of it?
4 A. Yeah.
5 Q. You didn't make any effort to go there to talk to anybody?
6 A. No, I did not.
7 Q. You didn't write a letter?
8 A. No, I did not.
9 Q. You didn't send a fax or an e-mail?
10 A. No, I did not.
11 Q. Okay. Don't you think it would have been at least interesting to
12 have that information other than the information from Serb authorities?
13 A. Well, it's interesting you should ask that, because several years
14 later I ended up at a conference with a couple of guys from I think it was
15 the Air Force War College on bomb damage assessment in Kosovo, and when I
16 told them this, they laughed at me. They said, of course they didn't call
17 you back, they were never going to give you that data. We would never
18 release information like that, not even years after the event. So I think
19 that while it may be an interesting point to debate here now seven years
20 after the fact, I think that my initial assessment has been borne out by
21 discussions with people who told me that I was never going to get that
22 data, no matter how hard I tried for it.
23 Q. [Microphone not activated] -- and that concerns the diligence of
24 your efforts, but we won't argue. Now, after the NATO bombing, you -- as
25 I understand it you next considered the hypothesis whether actions by the
1 Kosovo Liberation Army motivated Kosovar Albanians to leave their homes;
3 A. That's correct.
4 Q. And the information on KLA activity, again as I understand it from
5 your report, was obtained from interview accounts and a variety of
6 non-governmental reports summarised and provided to the project by the
7 ICTY; correct?
8 A. Some of them came from that. There were also web site reports,
9 again from the Yugoslav government.
10 Q. And as a matter of fact, you attached an appendix, appendix 3,
11 "Additional Sources of KLA and NATO Activity"?
12 A. Yes.
13 Q. Let's first look at the -- at the interviews, and what interviews
14 are you referring to here? You said "interview accounts."
15 A. Well, the interview accounts would have been those given to the
16 ICTY, principally, since the ICTY was conducting investigations and they
17 provided me with a database of those materials.
18 Q. And did you -- and so the interviews are primarily, if not
19 exclusively, of Albanians. Is that correct?
20 A. I don't know. As I said, they came from the ICTY, so at that
21 point I was unaware of what the original sources were. All I knew is that
22 they had come from the ICTY.
23 Q. Did your interview -- did your research and the general knowledge
24 you gained about the area ever indicate to you that ethnic Albanians
25 generally are hesitant, to say the least, to implicate the KLA in
1 activities for fear of reprisals?
2 A. I'm sorry. That's outside of my competence.
3 Q. So you never learned that, for example, ethnic Albanians who
4 collaborated with Serbs or even did business with Serbs were being
5 abducted and killed on a regular basis? You never learned that?
6 A. No, I can't say that I've heard of a systematic report about that.
7 Q. Forget about systematic report. Did you hear about it at all?
8 A. I've heard plenty of rumours about lots of violence in the region.
9 Q. But you didn't try to do anything to track down what you thought
10 to be rumours and confirm it was really fact?
11 A. I relied on the sources that are listed in appendix 3.
12 Q. And how about Serb civilians, were they interviewed?
13 A. I'm sorry. I have to refer to my earlier answer. I've answered
14 that already. The interviews came to me through the ICTY database,
15 principally. The other materials here are documentary sources that are
16 available that you can refer to if you're of continuing interest. The
17 interview materials, however, were provided to me by ICTY, and I do not
18 know what or who the original interviewees were.
19 Q. Now, you say they also came from a variety of non-governmental
20 reports, for example. Could you give us some examples of that?
21 A. There were reports by the European community monitor -- they're
22 listed here in appendix 3. I'm not sure which -- if you want me just to
23 read it or --
24 Q. No, no, no, not at all. I have them and I -- but you said
25 non-governmental, whereas in this appendix, there are various governmental
1 sources, for example, Federal Republic of Yugoslavia Ministry of Defence,
2 Federal Republic of Yugoslavia Ministry of Foreign Affairs.
3 A. Right.
4 Q. Were they consulted?
5 A. Their websites were consulted.
6 Q. And What did you find on their websites?
7 A. I found listings of moments of conflict between Yugoslav
8 authorities and KLA authorities in Kosovo during the period of study, and
9 we coded those as attacks, killings, and bomb attacks, where there was no
10 exchange of gun-fire.
11 Q. Did you remember or did you -- did you remember seeing a so-called
12 White Book published by the federal Ministry of Foreign Affairs in March
13 2002 containing some 958 pages -- and that's defendant exhibit 3D182 --
14 which gave detailed accounts of KLA attacks, KLA terrorist groups
15 attacking civilian, army, and police both before and during the war, do
16 you remember that 985-page book?
17 A. No, did not look at it.
18 Q. Did not look at it?
19 A. No.
20 Q. Don't you think that would have been a helpful source of
22 A. It may well have been. On the other hand, it may well have simply
23 reprinted the material that's available in the rather copious other
24 Yugoslav government sources that are here.
25 Q. You don't know that. You're speculating?
1 A. I'm speculating. Do you think that they came up with a great deal
2 of additional evidence in that period?
3 Q. Believe me, sir, if I could testify in this case, we would spend
4 the rest of the day here? Let me just ask you questions.
5 A. Right.
6 Q. Did you consult the publication issued by the VJ, the army General
7 Staff, in June 2001, defendant exhibit 3D133, setting forth again 206
8 pages of KLA terrorist attacks on the army, MUP, and civilians, both
9 before and during the war? Do you remember that at all, sir?
10 A. No, I do not.
11 Q. All right. Now, as I understand it, you didn't -- you also didn't
12 consider in your analysis, data on whether NATO and the KLA were working
13 together in various combat operations, did you?
14 A. We did not consider qualitative evidence of that kind. We
15 considered some kinds of quantitative evidence of that kind, but as I
16 reported, I believe I testified in the 2002 appearance, the statistical
17 data does not support the claim that there was an interaction between the
19 Q. Nor did you consider, as I understand it --
20 A. I'm sorry, sir, you're mischaracterising what I just said. What I
21 just said is that I did consider it, okay, let's be clear. I did consider
22 it, but there was no statistical basis for continuing.
23 Q. All right. I accept that. Thank you. Nor, as I understand it,
24 did you consider that there may have been advanced warnings to the KLA of
25 NATO bombings so that civilians could flee the area. Was that a potential
1 possibility that you thought about or considered?
2 A. I'm sorry. I'm puzzled by your question. Can you restate it
3 without starting with "nor" since that's the point we just clarified four
4 lines above.
5 Q. Did you consider whether -- whether there were advanced warnings
6 by NATO to the KLA to -- of bombing so civilians could be evacuated from
7 the area?
8 A. As is clear from my report, we considered only violent clashes
9 between KLA and Yugoslav authorities. We did not consider any kind of
10 verbal warnings by anyone to anyone else about any activity.
11 Q. That's all I'm asking.
12 Did you consider data that indicated that ethnic Albanians
13 sometimes left their villages on orders or instructions of the KLA?
14 A. I believe that's a restatement of the previous question and the
15 answer is --
16 Q. And the answer is --
17 A. -- no.
18 Q. Okay. Let's get to the third hypothesis and that is that refugee
19 flows and killings resulted from actions by Yugoslav forces. All right.
20 You state in your report that the statistical analysis and correlations
21 cannot prove that the Yugoslav forces were the external influence
22 responsible for the observed patterns. However, the findings of this
23 study are consistent with the hypothesis that action by Yugoslav forces
24 was the cause of the killings and refugee flow.
25 Did I state that accurately?
1 A. That's roughly accurate, yes.
2 Q. Okay. Now, similarly, and perhaps more clearly in the Milosevic
3 case, it's 3D521 at page 19967 --
4 MR. SEPENUK: By the way, Your Honours, the Prosecution has now
5 added the 2003 Milosevic testimony to their exhibit, but they only did
6 that yesterday. So prior to that, we had made a separate exhibit out of
7 it not knowing that the Prosecution was going to add it. So this is
8 contained in both the Prosecution exhibit and the Defence exhibit that I
9 just mentioned, 3D521. It's transcript 19967.
10 Q. And you said as follows: "We include no data on Yugoslav forces
11 or force movements. It is for that reason that there is no statistical
12 analysis in this report of Yugoslav activity."
13 Do you remember that? Testifying to that?
14 A. Yes.
15 Q. Okay. And in your most recent report, that's the report of
16 January 28th, 2007, revisiting killings and migration in Kosovo, you
17 said: "These three hypothesised causes are KLA activity, NATO
18 air-strikes, or a systematic campaign conducted by Yugoslav forces.
19 Because of available data, we're only able to consider directly the first
21 Okay. And finally on this point, so there's no mistake about it,
22 I'd like to quote from your article, Statistics and Slobodan, written by
23 you, I believe, with Jane Asher in 2002. This is after you testified on
24 the Milosevic case. Statistics and Slobodan is the name of the article,
25 using data and analysis and statistics in the war crimes trial of former
1 President Milosevic. And by the way, this is one of the publications
2 included in your updated curriculum vitae.
3 And in this report you said: "There was no data available to us
4 that detailed Yugoslav government force deployments or movements during
5 the period in question. We could not, therefore, directly test the
6 hypothesis that Yugoslav forces were responsible for ethnic Albanian
7 deaths and migration."
8 Do you remember that, writing that?
9 A. Yes, I do.
10 Q. Okay. Now, could you -- I may have missed it in all your prior
11 reports, Policy and Panic and this report, and so forgive me if I have,
12 but could you tell us why you could find no data that detailed Yugoslav
13 government -- the deployment of Yugoslav forces. Why couldn't you find
14 any data? I -- please enlighten us on this.
15 A. We found some data, but it was too skimpy, too scarce, too patchy,
16 to be considered sufficient. We found, for example, that there were huge
17 bases in some places and there were some supposedly armoured groups in
18 other places. But those were far too -- at too gross a level to provide
19 us with information. They were not individual attacks, the way the KLA
20 data was detailed. They were not specific air-strikes, the way the NATO
21 attacks were detailed. So we were unable to do this calculation.
22 Q. I guess -- I'm sorry. I guess what I don't understand is you were
23 able to find the KLA data without apparent difficulty --
24 A. Well --
25 Q. Well, let's say with difficulty. I don't want to minimise what
1 you did. You were able to find the KLA data. Okay.
2 A. Mm-hmm.
3 Q. And why -- you said you were able to find some data on Yugoslav
4 force deployment, but it was too skimpy. Where did you look? Who did you
5 talk to? How did you go about doing this?
6 A. I talked to some military analysts that I had met in Washington
7 and they didn't know much. They said that the Department of Defence
8 probably had it and that would have been one of the things I would have
9 asked if I had gotten through to anybody at public affairs --
10 Q. You mean if you had completed that phone call? Go ahead. Please,
11 go ahead.
12 A. But they were very skeptical that I would find anything. And the
13 KLA and NATO data was actually not that hard to find, counselor, because
14 it was on the Yugoslav government websites.
15 Q. How about the 958-page report put out by the Ministry of Foreign
16 Affairs well before your reports were written. Couldn't you have
17 consulted that?
18 A. One of the things that I think is interesting to consider when
19 we're looking for statistical evidence -- there may be a parallel idea
20 about this in law. You'll have to tell me. I don't really claim to
21 understand the law --
22 Q. I hope you don't think I understand statistics.
23 A. Don't worry, counselor. Is that evidence that goes against
24 interest is much more weighty than evidence that goes in favour of
25 interest. So the evidence that was presented in realtime by the
1 ministries -- the Yugoslav ministries on their websites was far more
2 interesting to me and continues to be far more interesting to me than
3 anything that would have been published by them subsequently, because
4 anything that was published subsequently, would have had the opportunity
5 to be sanitized, whereas things that were published in realtime, had a
6 very clear focus. It was written on the website. The websites were
7 titled with names like The Denouncement of Criminal NATO Air-strikes and
8 so forth. So the intent of those websites was quite clear. I felt
9 confident with that material because it was evidence against interest. By
10 using the evidence published by the Yugoslav authorities, that struck me
11 as the strongest possible case of the test of NATO or KLA involvement.
12 Using Yugoslav government information on their own troop movements post
13 hoc would have, and to this day, would strike me as highly suspect.
14 Q. How about KLA troop movements, what did you use for that?
15 A. KLA troop movements are not part of the analysis. What's part of
16 the analysis, sir, is interactions between the KLA and the Yugoslav
17 authorities. Bombings, injuries, casualties of Yugoslav authorities,
18 those are all denounced rather loudly in the websites that were published
20 Q. Sir, isn't it a fact that the reason you didn't get any
21 information on Yugoslav force deployments was that you were unable to do
22 that because you couldn't find any action of Yugoslav forces independent
23 of interactions with the KLA. Isn't that a fact?
24 A. No, in fact, that strikes me as absurd.
25 Q. Well, then let me refer to your article "Statistics and Slobodan."
1 A. Mm-hmm.
2 Q. 3D520, page 394, 3D520, page 394, if you'd kindly put that up,
3 please. And could you blow it up on the left side there. Yes, there it
5 What you just said was absurd, Dr. Ball --
6 A. -- mm-hmm.
7 Q. -- is exactly word for word what you said in this article. "We
8 were unable, however, to obtain data on Yugoslav Army activity independent
9 of interactions with the KLA."
10 Can you reconsider your testimony?
11 A. No, because that statement is out of the context of this
12 discussion. "We were unable, however, to obtain data on Yugoslav Army
13 activity independent of interactions with the KLA from the sources" - and
14 I'm sorry we didn't include this in the article - from the sources that
15 were used for this analysis.
16 Q. Isn't it a fact, sir, that the actions by forces of the KLA and
17 Yugoslav forces was so closely related in both time and place that it's
18 virtually impossible to separate them. Isn't that a fact, Dr. Ball?
19 A. No, that's not a fact at all, not to my knowledge.
20 Q. And given the dynamic that we've been talking about, isn't it fair
21 to conclude that the most that can be said is that ethnic Albanians fled
22 the area where combat was occurring between the army or between Serb
23 forces and the KLA?
24 A. I'm sorry. You've just assumed that my rebuttal to you was an
25 affirmation. Perhaps you should review. What I said is that, no, that's
1 not a fact at all, to my knowledge, and then you restated it as if I'd
2 affirmed it. Should we go over that again? I don't think we've
3 understood each other.
4 Q. Okay. Well, then, give me --
5 A. Go ahead.
6 Q. Give me your statement.
7 A. My statement is that I have -- I do not believe that it's the case
8 that actions of forces by the KLA and Yugoslav forces are so closely in
9 time and place, it's impossible to separate them.
10 Q. Do you accept it as one reasonable alternative, though?
11 A. Not necessarily. I think we would need a lot more information
12 about Yugoslav force activity before such a thing could be said.
13 Q. Which you never looked into?
14 A. I looked into and wasn't able to find, but let's consider a
15 thought experiment, Counselor. One way to think about --
16 THE INTERPRETER: Kindly slow down, please. Thank you.
17 THE WITNESS: The way a statistician would consider this problem
18 is to think about the relative sizes of the two forces. If the two forces
19 were in some way similar in size, then the proposition just made might be
20 reasonable. However, all the evidence suggests that the two sizes, the
21 size of the Yugoslav forces and the size of the KLA were radically
22 different --
23 MR. SEPENUK:
24 Q. Can I just interrupt you there for a moment. Have you ever heard
25 of a gentleman named Bislim Zyrapi, who was the commander of the KLA, who
1 testified in this courtroom that the KLA numbered about 18.000 people --
2 A. Mm-hmm.
3 Q. -- and had turned from a guerilla army into a regular --
4 A. Mm-hmm.
5 Q. -- army or more or less regular army with a guerilla capacity and
6 supported by a good percentage of the Albanian population. Would that
7 change the dynamic, as far as you were concerned?
8 A. Well, my analysis looks at the amount of interactions between the
9 KLA and the Yugoslav forces. Notwithstanding any testimony that has been
10 given here, looking at that data indicates that there was relatively
11 little going on between the Yugoslav forces and the KLA. Consequently, it
12 seems to me highly unlikely that there was either parity in size or an
13 extensiveness in interaction through this period. Consequently, I
14 disagree with the statement which you put into my mouth -- well, it's
15 scrolled off the screen now -- in which you said that -- well, we'll have
16 to scroll back to it.
17 Q. Well, what I'm respectfully suggesting to you, Dr. Ball, is that
18 like you, those fleeing refugees were also -- just like you couldn't
19 distinguish activity independent of interactions with the KLA and Yugoslav
20 forces, that refugees were also not able to distinguish actions of Serb
21 forces and the KLA and they were simply fleeing the combat area where
22 there was actions between two opposing forces, and I'm saying, don't you
23 at least accept that as a reasonable possibility?
24 A. Well, no, I don't, because part of the point of this entire
25 analysis, sir, is to show that over time and space, the interactions
1 between the KLA and Yugoslav forces do not coincide with the flows of
2 refugee movement or of killings. If your supposition were to be correct,
3 we would see interactions between the KLA and the Yugoslav authorities
4 at --
5 Q. Which you never got any data on; correct?
6 A. I'm sorry. Perhaps we should review. In appendix 3 we have two
7 pages of the lists of them, and I have subsequently published online the
8 database of the KLA and Yugoslav force interactions. Have you reviewed
9 those? Those might answer your question.
10 Q. Continue, please. I'm not here to answer your questions.
11 A. No. I'm just trying to clarify. I'm sorry. But the -- in that
12 on-line material that's been online since my 2003 appearance and here in
13 appendix 3 we detail our sources of KLA data, that KLA data is on
14 interactions between the KLA and the Yugoslav forces, precisely the kind
15 of conflict which -- hello? I'm sorry. There's been a --
16 Q. I think it's okay. I think you can go on.
17 A. Okay. Precisely the kind of conflict which, if I understood your
18 question correctly, you imply people would confuse and therefore flee.
19 Well, whether or not that's true, those conflicts came after most of the
20 killing and migration had been completed. That's the point of the 29
21 graphs in the recent report, that's the point of the regression analysis
22 in the January 2002 report.
23 Q. And my point was only that the actions of the Yugoslav forces - in
24 your article it says Yugoslav Army, but I think you meant Yugoslav
25 forces - activity independent of interactions with the KLA, you could not
1 find, and our contention is they were inextricably interwoven but, again,
2 this is not the place to argue. It is up to the Trial Chamber for later
4 Let me just turn to my last point here, and that is what you
5 call -- you say that you can't directly test the hypothesis that Yugoslav
6 forces were responsible for the migration and killings of ethnic
7 Albanians, but you do say that there's one important circumstantial factor
8 and that is the -- what you call the -- it's on section 5.4 of your
9 report, Prosecution Exhibit 1506, and that is -- you say:
10 "One of the findings of the study shows a circumstantial link
11 between Yugoslav Army activities and the observed pattern and killings and
12 refugee flow," and then you cite "the extreme decline in the number of
13 killings and refugee flows observed during the period 6/7 April, 1999,
14 coincides with the unilateral cease-fire declared by Yugoslav authorities
15 in recognition of Orthodox Easter."
17 A. That's correct.
18 Q. Now, Dr. Ball, may I suggest to you that an alternative reason for
19 the drastic decline in refugee flow is that immediately after the
20 announced cease-fire at 3.00 a.m. on 7 April 1999, the Yugoslav
21 authorities shut Morina border point and the daily number of refugees
22 entering Albania fell from tens of thousands to under 2.000 until the
23 border re-opened on the morning of 10 April 1999. I suggest to you that
24 that's another -- at least another alternative reason why the refugee flow
25 declined to almost 0. Isn't that a fair statement?
1 A. It is a fair statement, and we considered it at length.
2 Q. Okay. Thank you.
3 A. Would you like to understand the reasoning why we do not consider
4 it an important --
5 Q. All right. Sure.
6 A. -- factor?
7 Q. Go ahead.
8 A. Okay. Well, for two reasons: First, because killings match the
9 pattern precisely and it's not my impression that closing the border at
10 Morina would have stopped the killing pattern -- so the killing -- the
11 decline was not only in migration; it was also in killings. So that's the
12 first of the two reasons why we did not consider that especially
14 The second is that our analysis is of people crossing the border.
15 Our analysis is of people leaving home, and in the "Policy or Panic"
16 report, we -- or at that time I explain in rather tedious statistical
17 detail how we projected backward, people leaving -- from people crossing
18 the border into people leaving their homes. So we are not measuring
19 people crossing the border, which would be immediately vulnerable to
20 precisely the argument you raise but, rather, people leaving their homes,
21 which is the measurement of much greater interest.
22 We do not, upon looking at the transit times that people
23 experience in mid-April, find a dramatic change. It's not as though we
24 found that people had been held up all of a sudden for three or four
25 additional days in mid-April compared to earlier April. Rather, people
1 simply weren't leaving their home during that period, so it's not as
2 though it hasn't been looked at. It's been looked at and discarded.
3 Q. Yeah. Well, people, of course, were leaving their homes and there
4 were long lines of refugees. Do you not know this, that there were long
5 lines of refugees, when the border was closed --
6 A. -- mm-hmm.
7 Q. -- long lines of refugees were turned back by --
8 A. Yes.
9 Q. -- Serb authorities and told to go home.
10 A. That's right. Those were people who had left their home before
11 the 6th of April. What we're interested in here is people leaving their
12 homes between the 6th of April and the 10th of April.
13 Q. I don't think you know that, do you? Do you know that?
14 A. What's the "that"? I'm sorry. You've lost me.
15 Q. You say these are people who left their home before 6th April?
16 A. Well, if they were turned back on the night of the 7th of April,
17 it seems a fair guess that they left their homes on the 6th or before.
18 Q. Do you remember being interviewed or -- by the "Guardian," the
19 "London Guardian" on March 21st, 2002? This is 3D522.
20 A. I don't remember this specific interview, but you'll no doubt put
21 it up on the screen.
22 MR. SEPENUK: Your Honour, it's pointed out to me that there's no
23 answer on page 103, line 21.
24 JUDGE BONOMY: Sorry, there is an answer. Or is that the -- is
25 that your question in both lines?
1 MR. SEPENUK: I don't know. Maybe we can clarify it now. I'm not
2 sure we should go back, Your Honour.
3 JUDGE BONOMY: What I don't understand about this, Mr. Ball, and
4 perhaps you can very simply explain it. I understood perhaps wrongly that
5 you gathered your statistical information from people who actually crossed
6 the border.
7 THE WITNESS: That's correct.
8 JUDGE BONOMY: So if the border's closed, what's the source of
9 information from which you can work back to the point where people leave
10 their homes?
11 THE WITNESS: That's a great question and we spent months trying
12 to think that through. What we had done is conduct several thousand
13 interviews among people who were in refugee camps principally in Albania
14 but also in Bosnia and in Macedonia. And among the questions that we
15 asked them in several different surveys, some conducted by my teams and
16 others conducted by Physicians were Human Rights or Human Rights Watch,
17 were when did you leave your home in Albania and when did you cross the
18 border? Okay. So those are two different dates. Now, very often those
19 were the same dates; other times they were different dates.
20 When they were different dates -- when they were the same dates,
21 we calculated the transit time, the time people were on the road between
22 their homes and the border. We used that pattern to create a model of how
23 long it took people to cross the border and that model varied by when
24 people were leaving their homes and when they were crossing the border.
25 So we were able to change the model as necessary to reflect the differing
1 conditions that people faced through the course of the conflict. This is
2 all detailed in the appendix to the Policy or Panic report. And the
3 point -- the precise point of this goes to the question at issue here,
4 which is, what happens when the border closes? Do people stop leaving
5 their homes or do they simply queue up in a long line? My understanding
6 is that, for the most part, there were a lot fewer people leaving their
7 home; perhaps, not 0, there may have been some, but there was still a lot
8 more people leaving their homes both before April 6th and after April
10 JUDGE BONOMY: I'm sorry. I -- it's no doubt me again, but I
11 don't understand what the source of that information is.
12 THE WITNESS: The --
13 JUDGE BONOMY: I don't understand how a model can assist you in a
14 situation where the point at which the information is being gathered is
16 THE WITNESS: Because we weren't gathering the information from
17 people who were specifically at the border. We were gathering the
18 information retrospectively in late April and May from people who were in
19 refugee camps in Albania and Macedonia. And we were asking people in late
20 April/May, When did you leave your home in Kosovo? And they might say, On
21 the 2nd of April. And then we'd say, Well, when did you cross the border
22 leaving Kosovo into Macedonia, Albania, or Bosnia? And they would say,
23 The same day or a day later or two days later or we went to stay with our
24 cousin and then came six days later or what have you.
25 JUDGE BONOMY: Did that exercise not throw up a block of answers
1 that demonstrated that no one was crossing the border when it was closed?
2 THE WITNESS: Yes, that's what we found and that's documented in
3 some detail in the Policy or Panic report.
4 JUDGE BONOMY: I see.
5 Mr. Sepenuk.
6 MR. SEPENUK: Yes, just to finish up here, Your Honour.
7 Q. That 3D522 on page 489. Down at the bottom there, when you were
8 explaining what happened after the cease-fire and you
9 said, "Coincidentally enough, the number of refugees and the number of
10 people being killed falls to zero, bang. Right to zero like the door was
12 And I simply suggest to you, Dr. Ball, that at least another
13 reasonable explanation for the decline in refugee flow was just that, the
14 border was closed?
15 A. Mm-hmm.
16 Q. Thank you, sir.
17 MR. STAMP: Presumably that was a question, so could the witness
18 be allowed to answer it.
19 JUDGE BONOMY: Mr. Stamp wants to swap chairs with me, obviously.
20 That answer -- or, sorry, that quotation from you, which is that
21 the number of refugees falls to zero, can you explain what that means,
22 bearing in mind the two different points that are relevant, that's the
23 point of crossing the border and the time of leaving home.
24 THE WITNESS: Mm-hmm. This was a very -- I remember this
25 interview vaguely. It was a very long interview and I don't think that
1 this is the complete context of what I told the journalist. The number of
2 people crossing the border and the number of people being killed does fall
3 to zero or towards zero. Falling to zero is a term of art. It doesn't
4 mean that things go completely to zero but it does go pretty close to
5 zero. Could the door be closed, is that an explanation? It's a possible
6 explanation, but it fails to account for the number of killings following
7 that pattern almost identically. It also fails to account for our
8 transition -- excuse me, our transformation of the data from people
9 crossing the border to people leaving their homes, via the model I
10 explained earlier. This model, transforming border-crossing time into
11 home-leaving time, was a major focus of the Policy or Panic report and I'd
12 be happy to elaborate on it if it would please the Court.
13 JUDGE BONOMY: Thank you.
14 Mr. Sepenuk, in the cross-examination by Mr. Milosevic, there were
15 points made about the point of crossing the border and whether it was, in
16 fact, the Morina crossing point that the information was being gathered
17 at. Is that no longer a matter of controversy?
18 MR. SEPENUK: I think it could be, Your Honour, but in the
19 interests of this -- a lot that's going to be covered today. Quite
20 frankly, I have only the vaguest notion of what my colleagues are going to
21 cover but that could very well be covered in subsequent
23 JUDGE BONOMY: Very well.
24 Mr. Ackerman.
25 MR. SEPENUK: Thank you, Your Honour.
1 Cross-examination by Mr. Ackerman:
2 Q. Good afternoon, Dr. Ball. How are you doing?
3 A. I'm well. Thanks.
4 Q. You're holding up all right?
5 A. Sure.
6 Q. You can make it another hour?
7 A. Sure.
8 Q. Because I fly here from Texas and that's seven hours so I know
9 what you're feeling like I think.
10 When you started your testimony today you told the Prosecutor that
11 you were a sociologist and a statistician and the first question I have is
12 in relation to your educational background. According to your CV, you got
13 your bachelor's of degree in 1988 from Columbia University in New York in
15 A. That's correct.
16 Q. And then in 1991 you got a master's degree in sociology?
17 A. That's correct.
18 Q. And in 1998, you got a doctorate of philosophy degree in
20 A. That's also correct.
21 Q. And wrote a thesis, Liberal Hypocrisy and Totalitarian Sincerity,
22 the Social and Idealogical Origins of the National Human Rights Movement
23 in El Salvador, Pakistan, and Ethiopia, which just from its title seems to
24 me to have nothing to do with statistics, but pure sociology?
25 A. Sociology at the University of Michigan always nearly always has a
1 substantial quantitative component and I think if you refer to chapter 5
2 you'll find plenty of statistics there.
3 Q. In that dissertation?
4 A. Correct.
5 Q. You don't have any degrees in statistics though, do you?
6 A. No, sir.
7 Q. I assume you've taken a course or two though?
8 A. A course or two. Nine semesters, if I recall correctly.
9 Q. Well, I was being a little bit facetious when I said a course or
10 two. In the analysis that you did in this case, I take it you attempted
11 to be logical in that process?
12 A. That's correct.
13 Q. And logic is important to the analytic kind of thinking that you
14 do, isn't it?
15 A. Yes, it is.
16 Q. And I take it you studied some logic in college?
17 A. Yes, I did.
18 Q. And you studied the important logical principles?
19 A. Yes, I did.
20 Q. And in your analysis in this case you attempted to apply some of
21 those logical principles that you studied in college?
22 A. Yes, I did.
23 Q. And one of the principles you applied was the principle of the
24 undistributed middle?
25 A. In what context do you refer?
1 Q. In context of your analysis of the data in this case.
2 A. I'm not sure how you mean.
3 Q. Well, tell the Chamber what your understanding is of the principle
4 of the undistributed middle?
5 A. I don't know the principle.
6 Q. You do not know the principle?
7 A. No, I don't.
8 Q. Do you know that it was first -- that it was first annunciated as
9 long ago as Plato in his dialogue?
10 A. No, I don't.
11 Q. Never read those?
12 A. I read some of them.
13 Q. Well, let me give you a syllogism that --
14 MR. ACKERMAN: Your Honour, my microphone over there keeps coming
15 on and I don't understand why. I've turned it off three times now.
16 THE INTERPRETER: The interpreters kindly ask that the other
17 microphone is also switched on because we don't hear the counsel.
18 JUDGE BONOMY: I think your voice is not perhaps being picked up
19 as clearly as it had been.
20 MR. ACKERMAN: They've told me they need it.
21 Q. You're familiar with a syllogism like this, I'm sure. All the
22 goats have long hair, Jim has long hair, therefore Jim is a goat?
23 A. Yes.
24 Q. And that illustrates the fallacy of what is called the
25 undistributed middle, isn't it?
1 A. That illustrates a fallacy, yeah.
2 Q. Now, there's also a scientific discipline, is there not, which
3 analyses data using fuzzy logic?
4 A. Yes.
5 Q. And you're familiar, I suppose, with the proposition that fuzzy
6 logic was first suggested, I guess, I say suggested, and -- in a paper by
7 a scientist by the name of Zadik?
8 MR. STAMP: Your Honours, I wish to object to this line of
9 question. I think it's really a matter for the Court to control but it's
10 going too far now. The cross-examination should be focussed on the
11 reports of the witness and not generally as to syllogisms or ancient
13 JUDGE BONOMY: Mr. Ackerman.
14 MR. ACKERMAN: Well, Your Honour, one can always question an
15 expert about the techniques and systems that he used to arrive at the
16 conclusions that came from his analysis. And right now I'm exploring
17 those with him to see how many of those he paid attention to, how many of
18 them he used.
19 JUDGE BONOMY: Well, I think you can trust us not to require the
20 entertaining prompting that you've been good enough to favour us with so
21 far, but we'll be able to get to the crux of it fairly quickly and it
22 would help, I think, if you did, too.
23 MR. ACKERMAN: I'll get there quickly, Your Honour. I promise.
24 Q. I want to know if you applied fuzzy logic in your analyses in this
1 A. No.
2 Q. Do you know what the best-selling statistics book of all time is?
3 A. I don't know, but I suspect I can guess.
4 Q. Would you guess Darrel Huff's book?
5 A. Probably, either Darrel Huff's or "Misused Statistics" by Spirer.
6 Q. Darrel Huff's book has sold over half a million copies. It's
7 entitled, "How to Lie With Statistics"; isn't it?
8 A. Yes.
9 Q. And you're familiar with it?
10 A. I've glanced at it a few times. I know what it contains.
11 Q. You know that one of the things he writes about in that book are
12 misleading graphs; don't you?
13 A. Of course.
14 Q. And do you recall this, that he says with regard to misleading
15 graphs, that they can result from unintentional bias on the part of the
17 A. I don't remember that particular quote, no.
18 Q. Well, according to him, if he's correct, that even if you don't
19 intend to allow bias to creep into your research, that a misleading graph
20 can result.
21 A. That would be a truism, yes, sir. I mean a truism in the sense
22 that it's unfalsifiable.
23 Q. Mm-hmm.
24 A. The old --
25 Q. I think we can use an example and I don't -- let's look at P2678
1 on page 7, please.
2 A. Which report do you refer to, sir?
3 Q. It's your most recent one. This is a graph we've already looked
4 at today, so you'll recognise it immediately when it comes up.
5 MR. ACKERMAN: Now, if we can just enlarge that middle so we get a
6 good shot of that graph.
7 Q. Now, we've already discussed this at length, but what you're doing
8 here is you are taking new material from ICMP, ICRC, and OMPF, and
9 superimposing it over a graph that contains earlier material that you were
10 using at the time of your early testimony in Milosevic. Is that a fair
12 A. That's correct.
13 Q. And the way it's displayed here, it appears, if you just glance at
14 this graph without knowing -- without getting some kind of a serious
15 explanation about it, if you just glance at this graph, it looks like the
16 data in many ways matches up fairly closely with the earlier data, if you
17 look at where the peaks are and the height and stuff like that, doesn't
19 A. Indeed it does.
20 Q. And that's because this graph was constructed not as a stand-alone
21 graph but as an overlay, even though it doesn't say that it's an overlay,
22 that's what it is, isn't it?
23 A. I'm sorry. I don't know what you mean by an overlay.
24 Q. Let me tell you what I mean. You use one scale for the old data
25 where the high point in it is 1226 [realtime transcript read in
1 error "526"] and you use another scale for the new data where the high
2 point is 439, but when you overlay them, the high points wind up at the
3 same place, therefore making them look like they track each other?
4 A. I'm afraid we have some slippage on what we mean by track.
5 Q. Mm-hmm.
6 A. If I can explain, the point of this graph is not to show that the
7 scales are the same, but to show that the peaks and valleys are the
8 same --
9 Q. Well --
10 A. May I finish?
11 Q. Go ahead.
12 A. Any correlation analysis will disregard differences in scale.
13 This is an analysis of the correlation between the two series. Therefore,
14 the difference in scale is not relevant. In fact, scaling them to be the
15 same enables us to see the correlation. That's the point of the graph.
16 Q. But I wouldn't know that if you didn't explain it to me. If I was
17 just glancing at it, I would be struck by the similarities, especially in
18 the height of that first peak.
19 MR. ACKERMAN: Your Honour, it's just been pointed out to me that
20 page 113, line 19, where it says 526, what I really said there I think was
22 JUDGE BONOMY: Yes, thank you.
23 MR. ACKERMAN: It makes a difference.
24 Q. I want to have the usher show you a really crude little graph that
25 I've drawn and I emphasise crude. This isn't a beautifully
1 computer-constructed or anything like that. What I've tried to do is use
2 1226 as the high point and put that new data kind of where it belongs. We
3 can probably put this on the ELMO and then we can all see it.
4 Now, what I've done is used your 1226 scale to put in the 439
5 data, basically, that you see kind of in red down there and that's very
6 inartfully drawn, but in terms of raw numbers, that gives you a more
7 accurate view of what the raw numbers look like between the two sources,
8 doesn't it?
9 A. Well, what's our objective? If the objective is to see the
10 total -- the absolute values, that's correct. If our objective is to ask
11 the question: Would adding new data to the old data change the results?
12 Then I'm afraid that your graph tells us relatively little. What we might
13 ask if we wanted to use both graphs is what would be the correlation
14 coefficient between the two series. It's not reported here, but I can
15 calculate it for you rather trivially and I remember anecdotally that it's
16 on the order of 0.75 which is extremely high. So the graphs are not
17 different and I don't dispute that your graph is also informative, however
18 I do dispute that it is the appropriate graph to use given the question
19 we're asking in this particular context.
20 Q. Well, one of the things that I know you are concerned about is
21 data integrity and this shows a rather wide variance of data from
22 different sources, doesn't it?
23 A. No, not at all. The point is that we have count data, which is
24 simply the data that's available to be counted in the three new sources as
25 opposed to estimate data which takes into account the data which was not
1 countable. Therefore, they are not series which should be directly
2 compared; they are series which should be compared in terms of their peaks
3 and valleys which is done in figure 2. They should not be compared in
4 terms of their absolute values which you've done here in your, I might
5 say, quite elegantly done, hand graph. If you wished to take your elegant
6 hand graph and compare it more appropriately to the new data, I might
7 refer you to figure 1, appendix 2 of the 2002 report which I mentioned
8 this morning. That is the graph which contains the original count data
9 from the original four sources and would be a more appropriate comparison
10 to what you've shown here.
11 Q. All right. I'm through with this.
12 MR. ACKERMAN: I need to have this marked as an exhibit, Your
13 Honour, and it's going to be 4D something but I have no idea what and then
14 perhaps we can somehow get it into the system.
15 THE REGISTRAR: That will be IC121, Your Honours.
16 MR. ACKERMAN: That's okay, too.
17 Q. Now, this analytical process, when you started the process, you
18 started it with I think what might be characterised as a general
19 understanding that there had been a fairly large number of Kosovo Albanian
20 deaths. Is that correct?
21 A. Well, no. I started with the observation that there had been a
22 lot of refugees. I didn't know how many deaths there were, and, in fact,
23 even after I'd written my first report, it was not at all clear to me how
24 many deaths there had been.
25 Q. Well, there was a time when you must have suspected there'd been a
1 large number of deaths and then you started work on that and you applied
2 statistical methods to arrive at a number which you believed might fairly
3 reflect the number of deaths?
4 A. There was such a point, yes.
5 Q. And in your report at paragraph 1.3, that's in your original
6 report to the OTP, you indicate that you arrived at an estimated number of
8 A. I'm sorry, can you tell me the paragraph?
9 Q. 1.3.
10 A. Oh, yes.
11 Q. Did you find it?
12 A. Mm-hmm.
13 Q. It says you arrived at an estimated number of 10.356. Correct?
14 A. That's correct.
15 Q. The first thing that strikes me about that is that seems to be a
16 pretty precise number for an estimate.
17 A. Well, point estimates are of a point. When we interpret the
18 number, we interpret it within a margin of error, as it's commonly called
19 or, in more technical language, a confidence level. Throughout the report
20 you'll find us analysing that number in terms of the confidence level, in
21 terms of the error bands around it, both around this specific point
22 estimate and in terms of the trends and patterns over time that we work
24 That's why all the graphs in the appendices are accompanied by a
25 similar graph or table which indicates the amount of error that we
1 estimate to be associated with each point over time and space. For
2 example, figure 12 in appendix 3 is one such graph. Figure 11 also -- in
3 appendix 2. Figure 11 in appendix 2 of the same report is another such
5 Q. That was all interesting stuff. I don't think it had anything to
6 do with my question. My question was: When you're doing an estimate, how
7 do you arrive at such a precise number as 10.356? Why not 10.300 or
8 10.400? That sounds more like an estimate to me. That's the question I'd
9 like to have answered.
10 A. Right. Okay. An estimate is not a number that we guess at, but,
11 rather, a result that comes from fairly precise calculations. We report
12 the exact results of the calculations, but we also report our uncertainty
13 about the precision of those calculations. That's how statistics work.
14 So if we have a number of people who were drawn from this room, if we
15 select half the people in this room and I measure their weights, I can
16 come up with an exact average of their weights, but if you ask me then to
17 apply that average of weights to all the people in this room, I'll say:
18 Well, the precise measure in the sample is, you know, 175 pounds, but it
19 would be plus or minus 20 or 30 pounds. Okay.
20 But there will be an exact point estimate of the average. Here
21 there is an exact point estimate of the total deaths; however, no
22 statistician would work with that point estimate as an exact figure.
23 Instead, we would work with the confidence interval, the margin of error
24 around that figure, and the margin of error is described in some detail in
25 the appendix and how we calculated it.
1 Q. Well, I understand and I agree that 10.356 is not a number you
2 just guessed at; it's a statistically derived number from other numbers
3 you guessed at. Isn't that true?
4 A. No. It's a statistically derived number based on the counts
5 generated by our underlying sources and that -- to which we applied a
6 commonly used technique called multiple systems estimation which produced
7 this number.
8 Q. Mm-hmm. And we'll talk about that a little later.
9 A. Mm-hmm.
10 Q. To show us and demonstrate to us how careful you were in arriving
11 at this number of 10.356, you tell us in footnote 4 that all of the
12 statistical programming connected to the estimation of the results was
13 done independently by two analysts using separate computers and different
14 software and their results were identical; correct?
15 A. Yes.
16 Q. So these two independent analysts, not sharing, apparently,
17 software or computers or what they were doing or anything else --
18 A. Well, we shared what we were doing, certainly.
19 Q. You shared your data, for sure?
20 A. We shared the initial data and we shared the reasoning and the
21 methods we applied.
22 Q. And both of them came up with the number 10.356?
23 A. Within rounding, yes.
24 Q. Well, that's what we just went through. I said how did you get
25 that precise 10.356. Wouldn't it make sense for it to be 10.300 or
1 10.400? Now we get here where you say they come up with the exact same
2 number --
3 A. We truncate it --
4 Q. Now, wait a minute. And I asked you if it's 10.356 and you said
5 well, with rounding. Now, that makes no sense compared to the answer you
6 gave me just a moment ago.
7 A. We truncate it to an integer. I'm sure we can recover the
8 underlying fractions, if that's of interest to the Court.
9 Q. I want to know how it would have affected your conclusions
10 regarding killings if that number had come out, say, 20.742.
11 A. I'm sorry. I've just lost my earphones. There we go. It depends
12 where the 20.742 were distributed. It might not have affected them at
13 all; it might have changed them. It depends on how over time and space
14 those 20.742 or whatever number you choose had been distributed.
15 Q. So is it fair to say that there could have been no change or a
16 profound change or something in between?
17 A. There were a variety of other possible models which we report in
18 appendix 2, some of which reached as high as 20.000 or more. I checked
19 the patterns of all those models and none of them were substantively
20 different from the model we had here. So if there were a much, much
21 larger estimate and it were produced by the same techniques and
22 methodologies used in this paper, I am skeptical that the results would
23 have been different. If some radically different method were used, then
24 you could potentially have used it -- you could have come up with a
25 different result.
1 Q. Well, then the obvious question becomes this: Why go through this
2 very elaborate statistical process where two people are doing it
3 independently of each other on different computers to arrive at this
4 number 10.356 when it didn't make any difference; you could have just
5 picked a number out of the air. Let's use the number 7.000 or let's use
6 the number 12.000. It won't make any difference.
7 A. Because that's not how science is done.
8 Q. Well, I know that's not how science is done, but what you've just
9 told me, it probably won't make any difference if that number is double
10 what you found or half what you found.
11 A. I'm a bit baffled by the question. The question is: If you think
12 you know the answer, can't you just make it up? Is that roughly what
13 you've asked me?
14 Q. No. What I'm asking you is, does this number have any
15 significance to the analysis you performed by using it? I think you've
16 just told me that it doesn't.
17 A. The absolute number is not key to the analysis, no.
18 Q. Yeah.
19 A. What is key to the analysis is the distribution of that number
20 across time and space, that's crucial.
21 Q. So you can use any number as long as you distribute it the same?
22 A. Well, you don't distribute it. As long as that number is
23 distributed the same. The analyst does not do the distribution. It's --
24 the structure of that sentence is very important.
25 Q. Does the programme do the distribution?
1 A. No. The underlying data does the distribution.
2 Q. And the underlying data that gave you information about killings
3 didn't include numbers anywhere near 10.356, did it?
4 A. It included 4.400.
5 Q. That's right. Are you aware that the current best number on
6 Kosovar Albanian deaths is 4.200?
7 A. I think that may be the best number of documented deaths. Again,
8 the point of this analysis is to include those that are not documented
9 and, in all likelihood, will never be documented.
10 Q. Well, if one's taking into account all the bodies that have been
11 found and identified plus all the people that are missing and have not
12 shown up and you arrive at a number of 4200, you probably can't change
13 much from that. And if that's the current number, if that's where we are
14 today, then that's quite a ways from your number of 10.356 --
15 A. Well --
16 Q. -- and whether that's correct or not, I can't testify to, and I
17 won't. I'm just asking you hypothetically if it's in that range, 4200,
18 based on actual real-world figures, your statistically derived estimates
19 are off by more than 150 per cent, aren't they?
20 A. Well, you've embedded a fascinating hypothesis in the question, so
21 let's go back and unpack that hypothesis and maybe we can get to the
22 question. The hypothesis is that because people have only identified
23 4.200 bodies -- identified or claimed as missing, that that's all there
24 are. Now, I've worked in a lot of countries where there have been mass
25 atrocities from Peru to Guatemala to South Africa to East Timor, and in my
1 experience identifying even this many is an astonishing accomplishment.
2 It is very rare that very many people after a mass atrocity are
3 identified, for dozens of reasons; among those, that the families don't
4 see any point to report, there's not much -- they have little to gain by
5 reporting anything, because all the witnesses have either fled or gone.
6 Families may have left the country, what have you. There are many
7 reasons, and each context has its own reason why relatively few people
9 This said, as we discussed in the previous line of questioning,
10 whether the number is 10.356 or 4200 turns out to make relatively little
11 difference to the analysis being given in this report. Again, referring
12 you to the graph, figure 1, appendix 2, the unadjusted graph with the raw
13 data, we find the same pattern. So we can debate the validity of the
14 statistical estimate of the total, but the underlying argument is the same
15 whether or not you make the estimate or not. We made the estimate in
16 order to correct for biases we were concerned about in the captured data.
17 Q. Are you through?
18 A. Yes.
19 Q. Every answer you give raises new questions. I'm reasonably
20 certain you have absolutely no data from Kosovo about the incidents that
21 were reported. You're guessing about that.
22 A. The incidents of reporting, do you mean the reporting rate?
23 Q. Yeah, the reporting rate for people missing. I think you have no
24 information about that.
25 A. Well, actually, the reporting rate can never be analysed directly,
1 sir. The reporting rate always has to be estimated, so you're correct, I
2 never have direct evidence of the reporting rate, but neither does anyone
3 else for any other country on earth because the only way the reporting
4 rate can be analysed whether it's in Sweden, the best registry in the
5 world, or in some very-difficult-to-document place is by estimating the
6 total and then looking at your report totals and dividing into that
7 estimated total. By it's nature, the reporting rate is unmeasurable in
8 any direct sense, so you're right that I don't have that data but that's a
9 word game.
10 Q. You must know that one of the major issues in Kosovo today and
11 yesterday and for the last several years has been the issue of missing
12 persons and where are they and have they been found and every time a mass
13 grave is found there's a major interest in identifying who was there. And
14 people are constantly reporting: My Joe is still missing. Was he found
15 there? And we're now eight or nine years down the road from that war.
16 Don't you think we have a pretty good idea of how many people are missing?
17 A. Well, let's see, I've been working in Guatemala for the last few
18 months --
19 Q. I'm not talking about Guatemala. I'm talking about Kosovo. Don't
20 you think we have a pretty good idea of how many people are missing?
21 A. No.
22 Q. No, you don't. Now, if the number 4200 just, hypothetically, is
23 an accurate number, then the statistical analysis you did that put that
24 number at 10.000 was off by 150 per cent, wasn't it?
25 A. Well, hypothetically, we can agree to anything, so --
1 Q. We can agree to that, can't we?
2 A. If we agree hypothetically that the number is 10.356, then my
3 estimate was off by 0 per cent. I think that's a truism of the question.
4 Q. But you can answer my question that hypothetically it's off by 150
5 per cent if the right number's around 4.000?
6 THE WITNESS: Your Honour, is it a well-formed question?
7 JUDGE BONOMY: I think it's a well-formed question but it's been
8 answered because you've already told us that you don't think it will make
9 a significant difference to the outcome of your work.
10 But Mr. Ackerman is intent upon pursuing it and I'm not going to
11 stop him.
12 THE WITNESS: Yes, then.
13 MR. ACKERMAN:
14 Q. And that causes one to wonder about the integrity of the rest of
15 your statistical calculations. If you can be off by as much as 150 per
16 cent in this statistical analysis, what if you're off by 150 per cent or
17 200 per cent or 800 per cent with regard to the rest of your work? It
18 puts the rest of it in some question, doesn't it?
19 A. Well, I mean, we started off with a pretty bizarre hypothetical.
20 If you divide by zero you can prove anything and that seems to me the
21 structure of this questioning. I'm sorry. It doesn't strike me as very
23 Q. I'm going to change to a different subject now.
24 A. Okay.
25 Q. And we'll talk about something else.
1 JUDGE BONOMY: Just to be clear. There may be a miscommunication
2 here because of your understanding of what hypothetical means. Your
3 thinking of hypothesis being posed and then carrying out an exercise to
4 demonstrate whether or not they're right or not. But here, all that is
5 being put to you is that it may actually be the fact that there were only
6 4.200. Now, you don't accept that, but assuming that is the case, I
7 understand your answer is it doesn't make any difference -- or it's
8 unlikely to make any difference to the outcome of what you did.
9 THE WITNESS: That's correct.
10 JUDGE BONOMY: Now we can move on, I think.
11 MR. ACKERMAN:
12 Q. You were asked by the Prosecutor to assist in these matters to
13 help determine the effects of KLA activity and NATO bombing on refugee
14 movements and killings. That's basically what you were asked to do,
16 A. No -- I mean I was asked by the Prosecutor to do that after I'd
17 written two reports and then I wrote one for them, but I began this work
18 independently and carried it on for two years independently.
19 Q. We already knew that. I'm getting to the point of when you
20 started talking to the Prosecutor about what they asked you to do and
21 that's what they asked you to do?
22 A. They asked me to write up what I had presented in oral reports to
23 them from my earlier work.
24 Q. I take it you'd agree -- in fact, I wonder if you've ever heard of
25 a statistician from the University of Wisconsin, now retired I think, John
1 Steele. Do you know John Steele?
2 A. It vaguely rings a bell.
3 Q. Well, a question almost anybody would ask if you were to pose
4 what's going on here today to them just out of the blue, it would be:
5 Isn't the best way to determine why people left, what caused people to
6 leave, is just talk to them and ask them. Why are you using -- why are
7 you using the rather inexact methods of statistics to try to answer a
8 question that you could just ask the refugees about and get the specific
9 answers to. I mean, it makes one wonder why you're here, doesn't it?
10 A. Well, lots of people do that. This is what I do. This is a
11 different way of answering the same question.
12 Q. But somebody decided it was important for you to come and give
13 this testimony instead of getting that information from the people who
14 were there and who knew why they left and how long it took them to get to
15 the border.
16 A. That's a question for the OTP.
17 JUDGE BONOMY: Well, earlier in this case, Mr. Ackerman and his
18 colleagues spent a great deal of time and effort trying to prevent the --
19 and successfully to some extent, the Trial Chamber from considering the
20 answers to these particular questions as they might be posed. And in
21 place of that, what I think we have is accounts from a number of people
22 about what happened to them. And I expect - although, no doubt, I may be
23 looking at it too simply - I expect that the Prosecution might try to link
24 the statistical exercise to the statements that we've heard from
25 witnesses. They may not be able to do so of course but I would be
1 surprised if they weren't at least trying to.
2 MR. ACKERMAN:
3 Q. I wonder, Dr. Ball, if you were told by the Prosecutor when asked
4 to pursue this inquiry regarding what you and I have just been talking
5 about, if you were told by them that there was a very significant concern
6 regarding the truthfulness of refugee accounts and of what had happened in
7 Kosovo and that's why you were asked to perform this kind of an analysis.
8 Did they tell you that?
9 A. No. No one from the OTP has ever said any such thing to me. But
10 now that you mention it, actually, that's something that former President
11 Milosevic raised more or less continuously throughout his
12 cross-examination. So that's certainly a place I've heard it.
13 Q. And you've heard it other places, too. It's something, I think,
14 you're quite familiar with, aren't you?
15 A. In terms of Albanian refugees being liars? That's pretty much
16 come from former President Milosevic, yeah.
17 Q. Don't you recall speaking at a conference for journalists
18 sponsored by the crimes of war project in the year of 2000, May 5 and 6 of
19 2000, before you wrote your report in this case?
20 A. Yes, I do.
21 Q. And don't you remember that journalists there were expressing
22 their concern that Kosovo refugees were relating stories that had been
23 planted by the KLA?
24 A. Well, you use journalists in the plural. There was a journalist
25 there who told a story of a refugee who had misled her.
1 Q. That was Nancy Durham?
2 A. I believe so. I don't remember her in particular. And the
3 discussion came up how my methods would handle people lying to us and I
4 discussed it and I explained how we would deal with that.
5 Q. Nancy Durham showed a film at that presentation and said that she
6 decided to spill the beans about what was going on there at the border in
7 terms of KLA influence on the stories of refugees, didn't she?
8 A. That's not how I recall her presentation. She presented it as
9 having been misled by a single person and that's my recollection of the
10 story. Perhaps you have spoken to her more recently, I don't know.
11 Q. You don't recall her talking in terms of spilling the beans?
12 A. No.
13 Q. When you were -- you've talked today already about the methods you
14 used. I think you used two when you were talking about the movement of
15 refugees. The statistical analysis methods you used I think basically
16 boils down to a data correlation process where you're correlating one body
17 of data against another body of data to see what that shows you. Is that
18 a fair statement?
19 A. No. Data correlation is a very specific technique and I only used
20 it for exploratory purposes here. That's not accurate.
21 Q. Okay. This morning at page 33, line 20, you were being asked by
22 the Prosecutor about your November -- some changes to your November 2002
23 report, I think based upon some additional information you received
24 regarding NATO air-strikes.
25 A. What?
1 Q. And you indicated that you noted that the NATO air-strikes could
2 be interpreted to have caused a reduction in killings of Kosovar
3 Albanians --
4 A. Yes, from 2002, yes.
5 Q. Yeah. And you say: "We interpreted that result as coincidental."
6 A. Yes.
7 Q. Right?
8 A. That's --
9 Q. After you had explained that, saying you interpret it as
10 coincidental. I don't think I need to read all of your words at this
12 A. Mm-hmm.
13 Q. But I think you heard Judge Bonomy say: That might make some
14 plausible sense without trying to make a judgement of it. It may be an
15 action that because it engages the Serbs distracts them from killing
16 Albanians that I can see. You heard Judge Bonomy say that?
17 A. Yes, we discussed it.
18 Q. And I didn't notice that you made any effort to correct him
19 regarding his understanding of what you had said, did you?
20 A. I didn't interpret it as a question.
21 Q. But didn't you understand that he had misunderstood what you had
23 A. Well, it's a plausible interpretation. It's not the
24 interpretation we drew in this report, but it could be the case.
25 Q. Well, it is the case when you testified on the 2nd of May, 2003,
1 in Milosevic about this same thing. What you said in a question from
2 Mr. Nice regarding this data was this:
3 "Most of the NATO air-strikes occur in late May and early June.
4 We therefore think that although the statistical models will see that as a
5 negative correlation because much later in the period there's much more
6 air-strikes and fewer killings, more air strikes and fewer migrations."
7 And then you say this: "We believe it's illogical to argue that
8 air-strikes that occur in late May and early June can somehow be
9 interpreted as having prevented or stopped killings that occurred weeks
11 A. That's right. And this morning what I said was, we believe this
12 is a coincidence. It should not be interpreted -- it should not be over
14 Q. Well, coincidence is probably a term of art that most of us don't
15 understand. And I think you left Judge Bonomy with the belief that what
16 you were saying is that the NATO air-strikes were reducing the death rates
17 in Kosovo, and that's not what you were saying?
18 A. No, that's not what I was saying, and if I left that
19 misinterpretation, my apologies.
20 Q. And, in fact, you used this same statistical analysis that you
21 used to arrive at the other things you're telling us to come up with these
22 results which you then -- the only thing you can do when they are just
23 obviously wrong is say: Well, okay, this is a coincidence. And because
24 you can't now rely on them because they're just obviously wrong.
25 MR. STAMP: That is, with respect, engaging now in argument and
1 not proper questions to an expert witness about his analysis and
3 JUDGE BONOMY: Can you answer the question?
4 THE WITNESS: Yes. In our original report we cut off the analysis
5 on May 11th precisely to avoid this, because we knew that after May 11th
6 there were relatively few killings and relatively little migration. We
7 did not want to continue the regression past that point, because we were
8 concerned that the high level of bombings after that point would create
9 such an artefact as you have expressed concerned about.
10 We were asked to go ahead and extend the model. We did so. We
11 then covered it in caveats, as you have noted, but when we are asked to do
12 something, we do it, and then we interpret it as best we can. So if we're
13 asked to do something which we feel is not perhaps the best appropriate
14 scientific technique, we will nonetheless carry out the analysis but we
15 will try the best we can to explain why the analysis is perhaps not the
16 best statistical approach to the question.
17 MR. ACKERMAN:
18 Q. Well, is it -- you know, I'm just very confused, because it seems
19 to me what you're saying is, if the model we use supports our hypothesis,
20 then it's valid; if it doesn't support our hypothesis, then it's a
21 coincidence and it's not valid. And it seems to me it either works or it
22 doesn't work. So I think what you're proving here is that it doesn't work
23 and that your data is flawed and your analysis is flawed and we shouldn't
24 pay attention to it. Am I wrong?
25 A. Yes, and I must say your question is so long it would take quite a
1 while to unpack all the reasons you're wrong --
2 Q. Take your time.
3 A. But I will, yes.
4 Q. You find it humorous, do you?
5 A. No. I guess in California we smile a lot.
6 Q. And giggle?
7 A. When models are constructed, we construct them carefully so that
8 they are within the bounds that are appropriate for the piece of analysis
9 necessary at hand. In particular, we are concerned about a phenomenon
10 called outliers. Outliers are data points that are well outside the model
11 but that may change the model in some radical way because they're part of
12 a different pattern.
13 What we judged and what we argued in this report is that there is
14 a fundamental shift in the conduct of the conflict in mid-May, that in
15 mid-May there are very few killings, very few migrations that occur after
16 that point, after May 11th. There are, however, quite a few NATO
17 bombings. In fact, the NATO air-strikes intensify throughout the late
18 May/early June period. It would be in some sense tipping -- putting our
19 thumb on the scale unfairly to include the period during which there are
20 no NATO - excuse me - there are no killings or migration, but there are
21 NATO air-strikes. It's not so much that the model is invalid but, rather,
22 that it would overstate the impact of NATO's actions.
23 Our exercise throughout here, our impulse, our efforts throughout
24 this project have been to be as conservative as possible; and by
25 conservative, in this sense I quite specifically mean that we have
1 attempted in all cases to give the best possible case to the contradiction
2 of our hypothesis. Including the later period in May is a way of
3 strengthening our hypothesis greatly, as is shown in the November 15th
5 To the contrary, we wish to weaken our hypothesis as much as
6 possible, weaken the data that we are using to argue our hypothesis so as
7 to subject it to the strictest possible test. Adding the data for late
8 May is not conservative and we did not do it on our on. We did it because
9 we were asked to. Once we did it, the predicted result evolved, which is
10 that the NATO air-strikes seemed to have this effect of creating negative
11 killings, that is, stopping killings. However, it was an
12 anti-conservative move and we would avoid it.
13 Q. Well, you know, that was an interesting answer. The problem, of
14 course, is this: Us lay people who don't understand statistics like you
15 do can look at those subsequent reports that you say you were asked to do
16 that provided data that was not reliable and we can just read them over
17 and over, and I know it's in there. I know you could find it and show it
18 to us, but we can read them over and over and we don't see anything in
19 there that says: Please don't rely on this, this is not valid, or
20 anything like that.
21 A. Okay. We do have a confusion. Nowhere do I say, nor would I ever
22 say, that the data are unreliable. What I say is that the conclusions
23 should be taken with a grain of salt. And, as in the conclusion, I think
24 that it's not particularly difficult to find, that the conclusion of --
25 maybe we can put it on the --
1 Q. What page are you on?
2 A. I'm on page 2 of the 15 November corrigendum, 15 November 2002.
3 MR. ACKERMAN: That's P1394.
4 THE WITNESS: Yes. The conclusion's quite clear. None of the
5 conclusions alters our conclusions --
6 MR. ACKERMAN:
7 Q. Do you see it on your screen?
8 A. Yes, I can.
9 MR. ACKERMAN: For some reason, I don't have it.
10 JUDGE BONOMY: We've got hard copies, so we can carry on.
11 MR. ACKERMAN:
12 Q. Okay. Go ahead. What's the conclusion?
13 A. "None of the corrections alters our conclusions from the 3 January
14 2002 reports."
15 And that's it. That's the bottom line. We discuss some of the
16 minor changes, which are marginal, including this anomalous one, which in
17 the paragraphs above we don't treat it -- we say it's not serious, we
18 interpret these results as coincidental, but in the conclusion, the first
19 sentence of the conclusion, none of the corrections alters our conclusions
20 from the 3 January report. Now, in respect to that being hidden somehow,
21 let's refer to the introduction to the same report, which is, to be fair,
22 one and a half pages long, so it's not an especially long read. Okay.
23 In the introduction we say that the statistical -- that each of
24 the conclusions -- each of the individual corrections below was small and
25 discussed earlier their combined impact was negligible, and that's the
1 first sentence in the corrections, section two, middle of the page. So
2 it's not as if we've somehow hidden this in a welter of complexity;
3 rather, these two conclusions in which we say nothing's changed are pretty
4 much front and centre.
5 Q. Well, just up at the top of the page - it's going away right now -
6 you said that the conclusions should be taken with a grain of salt. Can
7 you point that to me?
8 A. We interpret these results as coincidental.
9 Q. Where does it say "with grain of salt"?
10 A. I'm sorry, sir. We're writing this in formal language. I was
11 trying to interpret for you, because you expressed some difficulty in
12 interpreting the formal language.
13 MR. ACKERMAN: Your Honour, I'm at a good stopping point.
14 JUDGE BONOMY: Yes. I hope that cross-examination is being
15 informed by the advice of someone, an expert in this particular area,
16 because very often, pure layman's cross-examination on technical areas
17 like this can be unproductive; on the other hand, it can in certain parts
18 be helpful. So hopefully, these are going to be distinguished in
19 tomorrow's continuation of this cross-examination.
20 That brings us to an end of today's proceedings. We shall resume
21 at 9.00 tomorrow. Please do not discuss your evidence with anyone, either
22 the evidence you've given or the evidence you might give with anyone
23 overnight and we'll see you back here tomorrow at 9.00. You may leave now
24 the courtroom with the usher. Thank you.
25 THE WITNESS: Yes, Your Honour. Thank you.
1 [The witness stands down]
2 --- Whereupon the hearing adjourned at 3.31 p.m.,
3 to be reconvened on Wednesday, the 21st day of
4 February, 2007, at 9.00 a.m.