¹⁹⁹⁴¹

Friday, 2 May 2003

[Open session]

[The accused entered court]

--- Upon commencing at 10.22 a.m.

JUDGE MAY: We are starting late, I'm afraid, because the case we were hearing overran from yesterday afternoon and we had to finish it. The upshot is that time is foreshortened, but we think that the most expeditious way we can use it, having regard to the time which is available and having regard to the requirement of all sides for time this afternoon, is this: We will sit for an hour and a half now; we will adjourn for half an hour; and we will then sit until 2.00 p.m. Yes, Mr. Nice.

MR. NICE: The application is that we may interpose Mr. Ball to deal with the corrigendum to his report, which was part of the Kosovo evidence. The possibility of interposition was raised yesterday when Mr. Ball's personal problems were identified. Those problems are that he has a flight back to America booked for tomorrow and has commitments in Chicago and California, I think, for next week to fit in with family commitments and will not be able to meet those commitments if he doesn't travel tomorrow.

JUDGE MAY: And --

MR. NICE: We notified the amicus and the accused via his associate yesterday, and neither of them raised an objection at that stage. I left a message with your legal officer. I believe the telephone system here has had some problems, so the message may not have been ¹⁹⁹⁴² received immediately.

JUDGE MAY: He has come from the United States, flown across the Atlantic to give evidence; is that right?

MR. NICE: He's certainly come from -- whether he's come direct, I don't know. He travels all around the world. But the next limb is back home for a day and then to Chicago and California.

JUDGE MAY: And how long do you anticipate being with him in chief?

MR. NICE: Ten to fifteen minutes, I should think.

[Trial Chamber confers]

JUDGE MAY: Very well. We will interpose Dr. Ball for a short period, and then we will return to the evidence which we were hearing.

MR. NICE: Before the witness comes to court, may I do something I've done certainly once before, on I think the 22nd of April last year and possibly since then. It relates to the possibility of using the provisions of 92 bis in whole or part for all future witnesses in the case. It was done certainly in April 2002 in respect of the Kosovo part of the proceedings, but really I think expressly for all witnesses, and the Chamber will recall that the provisions of 92 bis have been used in part during some of the Croatia segment witnesses. So at this stage I give notice under 92 bis (E) that we may seek to rely on the provisions of 92 bis for every forthcoming witness insofar as it is appropriate to do so. We will seek whatever relevant shortening of time periods may be appropriate if we find that within the next 14 days there are witnesses who we wish to give under the provisions in whole or ¹⁹⁹⁴³ part of 92 bis for their evidence in chief.

But this notification relates to witness whom -- witnesses whom we accept should properly be cross-examined. The position of witnesses who we will be applying should be taken under 92 bis without cross-examination will be the subject of the usual full written applications. The Chamber may recall that on earlier occasions, having given a general notice and working on the basis that there may be opposition of the kind forecast under 92 bis (E) by the accused, the Chamber has then dealt on a case-by-case basis with witnesses being provided by us with either the statements in paragraph numbers or by the statements and summaries so as to identify what parts of the evidence it is we seek to give by 92 bis and what part we seek to give live. The timetable problems on us are such that it is absolutely essential we optimise the use of time, and it seems to us that the time has come when we simply must use 92 bis wherever possible, and we have in mind the Chamber's recent ruling on the possible use of witness statements as evidence in chief where the Chamber reminded us of the potential to use 92 bis. The first witness who might qualify for this procedure is indeed the next witness. Now, we may start him today, in which case he'll be started live and that will be that. But if we don't reach him today, Mr. Gusic is a witness who might be susceptible to 92 bis but we haven't yet prepared his materials in a way that would identify precisely which parts we would suggest would be acceptable under 92 bis and which parts, for example, should be given live. It may be that there would be written notification of our position on that by the end of the day; alternatively, ¹⁹⁹⁴⁴ first thing on Monday morning, if -- if we don't start him.

JUDGE MAY: Yes.

MR. KAY: Can I deal with one matter? Because there's a danger when you wrap everything up in that way that injustice can be caused, because 14 days' notice is required with a particular witness, whether 92 bis provisions are to be applied, and for our part, we need to consider whether the rule should properly apply or not, which takes a degree of research and time to consider the matter.

JUDGE MAY: We'll not consider it in a vacuum. Let us receive an application and then we can start dealing with it.

MR. KAY: I'm much obliged.

MR. NICE: And Your Honour, I think I've already said that I'm going to be improving and regularly updating the witness list that has been attached to the latest -- to the only application for time that we've made, and that falls for determination in due course. In the newer versions of that list - perhaps there will be one as soon as next week - we'll have a column which I hope will identify prospectively what witnesses may and what witnesses may not be susceptible to 92 bis in part. So I think that will also be helpful.

Can I respectfully request from the Chamber a ruling in respect of one witness whose name appears on the latest running list of witnesses, in respect of whom we've sought leave that he be called for whom no decision has yet been made. You'll see the name, and I've got the list. I can tell you which one it is without reading it out publicly. It's -- it's on the second page, and it's the second name on the second page. ¹⁹⁹⁴⁵

[Trial Chamber and legal officer confer]

MR. NICE: Second page, second name.

JUDGE MAY: We think the matter is in hand, but that can be checked.

Yes, Mr. Kay.

MR. KAY: There's one matter of concern about that witness, as I can see from the table, and the Court will appreciate. This is a big witness indeed, and --

JUDGE MAY: What's the matter of concern?

MR. KAY: We've received no statement or materials, of course, at this stage in relation to that witness. There's a date for his testimony which is apparently fixed as only being one particular date on which he is available. Looking at the timetable of disclosure, adequate time for preparation by the accused in relation to such a witness, I can register concerns about a witness suddenly coming into the proceedings in that way without adequate disclosure. The trial --

[Trial Chamber confers]

JUDGE MAY: Let us deal with it on this basis. At the moment - again, it's purely hypothetical - we've merely got this list. There will be an issue, of course, as to disclosure -- well, there may be an issue. We'll have to see what happens when the Prosecution propose to call the witness.

MR. KAY: Adequate time is a very important consideration.

JUDGE MAY: Mr. Nice, you have that point. Seeing the witness and seeing who it is, that obviously must be right. ¹⁹⁹⁴⁶

MR. NICE: Yes. But he's only being called on a very narrow basis, a limited amount of material. I think five pages of notes will be served and can probably be served pretty well immediately. And we are certainly going to ask where we take any witness on a particularly narrow basis, as we would take this witness in the short period of time, that that doesn't open up the floodgates for endless cross-examination, because to do so imperils our ability to prove the case within the time limits that, whether amended or not, are imposed on us. And of course, it's always -- take this particular witness, without naming him. If, for example, we call him on a particularly narrow issue for an hour's evidence or less and he's cross-examined and dealt with on that, there's nothing to stop the accused calling the same witness, if there's a whole lot more he wants, in his own case and in his own time. And with witnesses who are capable of covering such a wide -- capable of covering such a wide part of the history, that may actually sometimes be an appropriate course. We desire him only for limited purposes and we will call him in a short period of time.

JUDGE MAY: Well, I notice the day is some way off that you propose to call him. Has the statement been disclosed yet?

MR. NICE: No. We've been waiting for a determination on whether he can be added to the list. But it can be done forthwith. It's not a statement, it's actually notes, but it's the same as.

JUDGE MAY: We must resolve this. It's been granted, I'm told.

MR. NICE: Thank you very much.

JUDGE MAY: The decision is still being filed. ¹⁹⁹⁴⁷

MR. NICE: Thank you. We'll disclose today.

JUDGE MAY: Very well.

MR. NICE: Thank you very much. May the witness come in. The Chamber will probably want to be able to refer to the report that the witness produced on the last occasion, together with the submission of the corrigendum. I do not intend to revisit the nature of his testimony in any detail but must take it as -- as it were, read and recalled.

[The witness entered court]

MR. NICE: But it may be the Chamber, after this interval of time, will or not want the witness to be re-sworn. I don't know.

JUDGE MAY: He needn't be re-sworn. Dr. Ball, you -- Dr. Ball, you are, of course, under the declaration which you made on the last occasion.

THE WITNESS: Yes, sir.

WITNESS: PATRICK BALL [Recalled] Examined by Mr. Nice:

Q. Dr. Ball, you've produced a corrigendum which has text covering just one and a half sides, and you then produced on pages 3 to 13 various figures, charts, and graphs which are reruns, under different technical procedures slightly, of graphs produced on the pages cited and referred to in the earlier report. So that on page 3 of the addendum, to find that graph, we have to go to page 2 -- figure 2 on page 6 of the original report; is that correct?

A. That's correct. ¹⁹⁹⁴⁸

Q. In fact, I'm only going to ask you probably to look at about two figures. Before doing that, your corrigendum arises from the following: First, did you find available to you, after giving evidence, further material from open sources and from other material published in Belgrade and elsewhere maybe, about NATO airstrikes occurring through early May and early June of 1999?

A. Yes.

Q. Although your conclusions material to this case really related to an earlier period, did you think it appropriate to add this later NATO airstrike material?

A. Yes, we did.

Q. Does it have an effect on one of the figures that we can look at? For example, figure 11, page 14, shown on page 9 of the addendum and of course on page 14 of the original report.

A. Yes. You can see very small effects there.

Q. Which is the -- which is the chart that gives a better demonstration of the effects?

A. Figure 19 has, I think, the clearest expression of those effects. It's page 11 of the addendum.

Q. But I want to stay with the graph first.

A. All right.

Q. Because your expertise is only ever going to be understood by laymen in a different way from the way you would understand it, and I want to demonstrate something through this chart, which could be conveniently shown on the overhead projector if there's one available. But I'm not ¹⁹⁹⁴⁹ going to take time while we do that.

MR. NICE: Have you got a spare copy of the report, page 13 of the addendum page 9?

Q. If the Chamber looks, while we're putting it on the overhead projector, at page 13, figure 10 -- page 9 of the corrigendum -- we don't have a spare copy of the original report at the moment, so this may be an incomplete exercise.

But looking at page 9 of the corrigendum and page 13 -- sorry, page 14 of the original report, we can see here what to the untrained eye will look an identical top line; correct?

A. Yes, the top line is identical.

Q. People leaving their homes.

A. Yes.

Q. The bottom line, which is called "Residuals," is different on the part post the 11th of May, i.e., the period for which you got additional material about NATO airstrikes. That's visible to the naked eye. Also visible to the naked eye, if one looks, for example, at the period the 24th of March to the 6th of April on the Residuals line, is the fact that that line is a bit different. Not very much different but a bit different.

JUDGE KWON: I'm sorry, but are we looking at figure 10 or 11?

MR. NICE: Figure 11.

JUDGE KWON: Figure 11.

MR. NICE: Page 14 on the original. And figure 11, page 14, shown on page 9 of the corrigendum. ¹⁹⁹⁵⁰

Q. So even the -- put it this way, Residuals line for the -- for phase 1, 24th of March to 6th of April, seems to have changed a bit, but you didn't have any additional material about NATO bombing because there wasn't any for that period.

A. Mm-hm.

Q. Can you just explain in layman's terms why it is that some material that came to you about the period post the 11th of May, or thereabouts, affects the look of the line as early as 24th of March?

A. Yes. A statistical model is an equation or set of equations derived from data, and the residuals here are the result of the statistical model that we built. Adding data for this later period changes the equations and so also changes the estimates that we would make for the earlier period. Once the equations have changed, that change will affect the entire line. So the line itself is not -- is the product of the changes throughout the -- throughout the model that result from changes in data only at the very end.

Q. Thank you. I'm going to come to the result of your additional materials later, but let's go to the second piece of additional material that you had to deal with. Following your giving of evidence, did it become clear that some of the data upon which you relied was subject to Rule 70 of the procedures -- of the Rules of Procedure and Evidence of this Tribunal?

A. Yes.

Q. There being no release from the provisions of Rule 70, that is therefore material that although you had had access to it, you shouldn't ^{19951 BLANK PAGE 19952} have had access to?

A. That's correct.

Q. Did you rerun your tests, deleting that material?

A. Yes, we did.

Q. Was there a total of 17 records that had to be deleted?

A. That's correct.

Q. But was it only three of those that had any information of relevance to your analysis?

A. That's correct.

Q. A third change can be understood by those with the appropriate technical expertise at page 50 of the original report. We've got a spare copy of that. We'll lay it on the overhead projector to see exactly what you're saying, just page 50.

The line that the Judges need to look at is the line in the middle of the page where you dealt with the procedures you adopted and in particular how you'd determined to make a model choice for two-day estimates. And in the middle of the page we see the line, "Choose the model with the lowest adjusted Pearson chi-squared statistic." On review of all your materials and tests, did you discover that that line was to a degree inaccurate?

A. Very slightly, yes.

Q. Because?

A. Instead of the adjusted chi-square statistic, we had used a raw chi-square statistic to make the selection for that component of the estimates. ¹⁹⁹⁵³

Q. We're not going to attempt in the time available a detailed understanding of what either a Pearson or a raw chi-squared statistic is, but in layman's terms, can you describe what this statistic did in the process of preparing this materials, and then we'll have a look at the consequences of the change of choice.

A. Mm-hm. As we did these estimates, we had to produce many different estimates for each point in time, each two-day period. We did so in order to explore the particular intricacies of data that had given rise to that estimate. We then had to select the best model there. There are many ways to do that. We felt that the best would be to express a scientific value which states that if you have several different possible explanations, scientific explanations for material, then all else being equal, the simplest of those explanations is to be preferred. It's called Occam's razor. More formally it's called parsimony. The adjusted chi-squared statistic expresses that scientific value. It incorporates this notion of parsimony or simplicity into the calculation. We would then prefer that to other measures for selection.

Q. Having stated in the report at page 50 that you had chosen the model with the lowest adjusted Pearson chi-square statistic, you then discovered that in fact ...?

A. We had used the raw chi-square statistic for that component of the calculations, for the piece that's described here on page 50 for the two-day estimates.

Q. Given that for reasons you've identified, namely parsimony, as you describe it, the other statistic was your preferred choice, did you rerun ¹⁹⁹⁵⁴ the materials with the other statistic in place?

A. Yes, we did.

Q. Did that bring some change to the results?

A. There is a change. It's extremely small. It cannot be seen easily in the graphs, but it can be seen in the numeric tables.

Q. And that's, therefore, the last page we'll look at, which in the corrigendum is on page 11. And that itself tells us we can find the original chart for which this is a replacement, as it were, at page 58 of the report.

And the Chamber may recall, looking at page 58 of the report, that the various figures revealed as regression coefficients were statistically of no consequence save where they have asterisks beside them. Would that be correct?

A. That's correct, yes.

Q. The Chamber may, I think, find it annotated its papers halfway down the page, opposite "KLA (battle)" with an annotation to the effect that the third entry, 2.728,6 should have an asterisk on it because, broadly speaking, the figure in brackets is less than half of the figure 2.728. Is that correct, Dr. Ball? I certainly marked mine as being asterisked whereas --

A. Yes. Yes, that's correct.

Q. It may be that the Judges will have done the same thing. But in any event, it was only those with an asterisk beside them, and although the figure is not precise - and you can explain this again in the way that you did before - these figures only become significant if ¹⁹⁹⁵⁵ they are very approximately twice the figure in brackets, which is the -- tell us, what's the figure --

A. The figure in brackets is the standard error. Twice the standard error -- plus or minus twice the standard error is sometimes called the margin of error. In order for something to be -- a measure to be meaningful, it has to be sufficient to exceed the margin of error. The margin of error must not overlap 0. If it does, then in formal language, we cannot distinguish the measure from 0 and so therefore must draw no interpretation from it.

Q. So to remind the Judges, looking at the original report, page 58, figure 19, and going to the left of the figure on which I've concentrated, 13.3 being nothing like as much as twice 12.2, is of no significance; 34.7 being nothing like twice 32.1 is also of no significance.

A. Right.

Q. When you reran the model using the different chi-square statistic, you produced the chart that we have on page 11 of your corrigendum.

A. Yes. It incorporates all three of the changes we've discussed this morning.

Q. And what is it that we now notice about the numbers beside which there are asterisks?

A. There are several new -- newly significant numbers that result, I think primarily from the addition of the NATO data after the period when migration and killing occurred. These are these three numbers here. It is of interest that they are all negative, indicating a negative relationship between the occurrence of NATO airstrikes and the occurrence ¹⁹⁹⁵⁶ of killings in those periods and regions.

JUDGE KWON: Could you point at the three numbers once again.

THE WITNESS: This one, this one --

JUDGE KWON: Yes.

THE WITNESS: -- this one.

MR. NICE:

Q. Is there any way we can see the consequence of this on any of the graphs? You say the consequence is very slight.

A. Yeah, it's very difficult to see. The -- excuse me. There is a very close comparison of figures 10 and 11. Comparing the bottom lines in each of those against the originals, we'll find that these two figures -- to these two figures, the bottom lines in figures 10 and 11, very slightly, more closely approximate the top lines than they do in the original report, but it's, as I said, very, very small.

Q. I think that your drawing that to our attention provides a suitable route to really the last question I want to ask you, which is to remind us of what a residual is and what its closeness to the top line, which is the -- in this case, people leaving their homes, if we're looking at figure 11, what the significance of vis-a-vis its being close in shape to the top line.

A. As I said earlier, a statistical model is a set of equations, one or more equations, derived from data which we use to make predictions, numeric predictions about something we observe. In figure 10, for example, we've built statistical -- a statistical model which predicts for each two-day period how many people were killed. The Residual line ¹⁹⁹⁵⁷ underneath is the result of subtracting the prediction from the top line. The purpose of such an analysis is to see if the prediction has been a good one. Has it approximated -- has it told us something about the top line? Finding that the residual pattern closely tracks the original line indicates that the prediction has been a poor one; it has not well described the original.

When we say that we have failed to explain the original pattern, we make that conclusion on the basis of the residual being essentially, in an interpretive sense, identical to the original pattern. So we have failed to explain the original pattern. Since the statistical model incorporates our knowledge -- our statistical knowledge about KLA and NATO activity, what we say then is that KLA and NATO activity failed to explain, in this case, the number of killings in each two-day period. It is therefore that the model gives us the basis for rejecting the hypotheses that the KLA activity or NATO activity in our model could be a cause of killing. So the closeness of the two lines is a way of saying that the data used to create the bottom line does not explain the top line.

Q. I said that was the last question, but I think, looking at page 2 of your corrigendum and recognising that many people here or viewing will not have had the opportunity of going with you over the techniques you employed in more detail, it may be helpful to put this material in even more lay terms.

A. Mm-hm.

Q. Did the new material in raw terms apparently say something about ¹⁹⁹⁵⁸ whether the number of deaths increased or declined as NATO airstrikes increased or declined?

A. The -- adding the data from late May and early June creates in the model a negative correlation between NATO activity and killings. Formally the interpretation would be that on a given two-day period, on average there would be 7.8 or approximately 8 fewer killings for each NATO airstrike that occurred in that two-day period. And there would be 14 fewer killings in the same region, in the same period -- excuse me, in the subsequent period as a result of each additional NATO airstrike. Overall, the interpretation is that at times and places where there are more killings, there tend to be fewer airstrikes, and conversely, when there are more airstrikes, there tend to be fewer killings.

Q. Now, for anyone seeking to argue that the NATO airstrikes were irrelevant to the killings, this raw finding or raw conclusion would seem to be favourable to that conclusion, but you make the point contrary to that, that it's really coincidental.

A. Yes. It's clear from the various graphs that most of the migration killing occur in late April -- excuse me, late March through early to mid-April. Most of the NATO airstrikes occur in late May and early June. We therefore think that, although the statistical models will see that as a negative correlation, because much later in the period there's more airstrikes and fewer killings, more airstrikes and fewer migrations, we believe that it is illogical to argue that airstrikes that occur in late May and early June can somehow be interpreted as having prevented or stopped killings that occurred weeks earlier. So we ¹⁹⁹⁵⁹ interpret this as a statistical coincidence.

Q. In sum, do the -- does the additional material about airstrikes, the deletion of the three pieces of material that were Rule 70, or the rerun using the different coefficient change your original conclusions at all?

A. No.

Q. Thank you.

JUDGE MAY: Yes, Mr. Milosevic. Cross-examined by Mr. Milosevic:

Q. [Interpretation] Mr. Ball, the title page of your report of the 15th of November says the following: "The American Association for the Advancement of Science, American Bar Association, Central and East European Law Initiative." That's right, isn't it?

A. Yes.

Q. However, already on the next page - and the number of that page is 03079044 of the B/C/S version, but it is page 2 anyway of your text - it says the following: "The materials contained herein should not be construed to be the view of the American Association for the Advancement of Science, Science and Human Rights Programme, or this third institution, the American Bar Association," et cetera. That's right, isn't it?

A. That's correct.

Q. Is it also correct that on that same page you state that interpretations and conclusions do not represent the positions of the organisations mentioned and that they have not been allowed either by the board of governors of the American Bar Association, et cetera? Is that ¹⁹⁹⁶⁰ roughly what it says? "Views of the AAAS board council, the CSFR, or the members of the association."

A. That's correct.

Q. And at the end it says, "Nothing contained --" and I'm quoting from the B/C/S version here. "Nothing contained in this publication is to be considered as the rendering of legal advice for specific cases, and readers are responsible for obtaining such advice from their own legal counsel. This publication and any forms and agreements herein are intended for educational and informational purposes only." Is that so, Mr. Ball?

A. That's correct.

Q. Now, tell me this, please: As you're a man of science, what value, in view of all these very explicit reservations that have been proclaimed here, do you consider that your material has in proving the indictment by the other side here?

JUDGE MAY: That's not a relevant question. What you're limited to is asking this witness about the evidence he's given, rather than a polemic of that sort.

THE ACCUSED: [Interpretation] Very well, Mr. May. If the question is not relevant, let's move on to others which you might consider relevant.

MR. MILOSEVIC: [Interpretation]

Q. Mr. Ball, you placed the focus of your research exactly on the period of 78 days, the 78 days of the NATO aggression. That's right, isn't it? ¹⁹⁹⁶¹

A. [No audible response]

Q. From the 24th of March, that is, until the first days of June; right?

A. Yes, from late March to early June.

Q. Yes, that's right. And then you divide that period of time into phases, stages, and you established movements, values, estimates of the numbers of people leaving their homes or who were killed; is that right?

A. That's correct.

Q. In doing so, you use -- for all the casualties of the war, you used the term "killed persons"; right?

A. For the deaths of ethnic Albanians, we used the term "killed persons." Not for all casualties of war. That's explicit in both reports, I believe.

Q. Fine. Very well. Now, as you set this framework for yourself from the 24th of March to the first few days of June, and then you go on to observe that looking at movements within those frameworks your basic conclusion represents the following: That the activities of the KLA and NATO were not the cause of human casualties. That is your main message; right, Mr. Ball?

A. The conclusion is that activities of the KLA and NATO are not the cause of the deaths of ethnic Albanians, as represented in these estimates, nor are they the cause of the mass migration patterns also described here.

MR. NICE: Your Honour, before the --

THE ACCUSED: [Interpretation] That's fine. ^{19962 BLANK PAGE 19963}

MR. NICE: Before the accused goes on, I'm reminded helpfully by Mr. Saxon and I should have reminded you earlier, that the leave given for cross-examination by your order of the 25th of February of this year is very narrowly expressed, and it's that he should be available for cross-examination on three issues: Source, authenticity, and reliability of the new data; two, reason for and the precise effects of the using of the adjusted Pearson chi-squared statistical method; and three, the basis for the different figures provided in the revised tables at pages 313 of the corrigendum -- 3 to 13 of the corrigendum. So that is the order of the Chamber.

JUDGE MAY: Thank you. Yes. Mr. Milosevic, you've heard that. Move on to some relevant matter.

THE ACCUSED: [Interpretation] Very well, Mr. May. I assume that these are relevant things.

MR. MILOSEVIC: [Interpretation]

Q. Now, tell me this, please, Mr. Ball: Regardless of the fact that the opposite side interrupted us, I nevertheless am sure that you're following what I'm asking you, and this is it: How do you know from the aspects of statistics -- because you have made this framework for yourself from the end of March to the first days of June, that is where you place your research, within that framework. From the statistical standpoint, how do you explain the fact that before the war began there was -- there were no casualties of this kind?

JUDGE MAY: That's not a matter for the witness. He's purely produced this report, and your cross-examination should be based -- if ¹⁹⁹⁶⁴ you've got any questions about the changes in the report, the corrigendum, you can ask them. The point is you've already cross-examined him once extensively, so we can't go over that again. Now, have you got any questions -- do you have any questions you want to ask about the corrigendum?

THE ACCUSED: [Interpretation] I do. As the corrigendum has been provided for this purpose, then I assume, Mr. May, that it is quite a legitimate question to ask the author these questions. Despite all the restrictions and limitations we have noted, he stands by the fact that this publication can be used in these proceedings, the ones conducted here and now.

MR. MILOSEVIC: [Interpretation]

Q. That's right, isn't it?

A. Yes.

Q. And in that regard, it is not vital that the casualties have all been situated during a time of war, whereas before that these casualties did not occur.

JUDGE MAY: These are not matters for the witness.

THE ACCUSED: [Interpretation] Very well. Fine. So just the corrigendum, is that it, Mr. May?

JUDGE MAY: Yes.

MR. MILOSEVIC: [Interpretation]

Q. On page 03079045, point 2 of the corrigendum, you state that each of the individual corrections described below was small, and as we discussed it earlier, their combined impact was negligible. That's what ¹⁹⁹⁶⁵ you state; right?

A. Yes.

JUDGE MAY: This is so we --

MR. MILOSEVIC: [Interpretation]

Q. But you also --

JUDGE MAY: This is page 1 of the report.

THE ACCUSED: [Interpretation] Yes. Very well.

MR. MILOSEVIC: [Interpretation]

Q. Then you go on to say the following. It is in section 2.2. There are three dots, and the second dot reads as follows: "Pursuant to Rule 70 of the ICTY Rules of Procedure and Evidence -" of this institution, that is - "we were asked by the OTP to drop 17 records from the data they provided to us about activity by the KLA because the information came from confidential sources." Is that so?

A. Yes.

Q. Now, answer me this, please: By including these information -- this information in your analysis, would you in any way whatsoever have disclosed the identity of the sources of that information? Because here you're dealing with statistical indices, a statistical representation, in fact, of what you are following in terms of numbers. So had you introduced into the analysis what you were not allowed to do so by the opposite side over there, because it was confidential data and sources, as you say in your corrigendum, in order to disclose the identity, would the identity of the source information be disclosed had you used what the opposite side did not allow you to use? ¹⁹⁹⁶⁶

JUDGE MAY: Can you answer that, or not, Dr. Ball?

THE WITNESS: It's -- all I can say is that the OTP provided me with the identification numbers of the records to delete; I deleted them. Three of those records were relevant to this analysis.

MR. MILOSEVIC: [Interpretation]

Q. All right. Fine. Thank you. That will be sufficient as a response to my question.

Now, the substance and essence of your corrigendum in fact relates precisely to this piece of information: That according to Mr. May's assessment, we did not use adequately but I assume you'll confirm that that does represent the substance in the period you're dealing with, that is to say, the time of the NATO aggression. So in that period of time, the relationship and ratio between killings and NATO bombings is reversely proportionate; that is to say, the number of the killings of Albanians is reduced with the intensity of the bombing as it increases and vice versa, is increased with the decrease of NATO aggression and bombings. Is that what you're claiming?

A. I'm a bit confused by the explanation. Perhaps I can -- can clarify. We interpret this as coincidental. Our observation is that there is a negative association between the patterns if one interprets the statistics literally. As I explained during my direct examination, however, we believe this to be a coincidence of when these two processes occurred, with NATO bombing occurring weeks later than the bulk of killing and migration.

Q. So you consider this to be coincidental. ¹⁹⁹⁶⁷ Now, could you give us the source you used for your information when you conducted this analysis and the conclusions you arrived at. What source?

A. The same sources were used as in the original report. They are listed in the original report. I'll give you the page. It's Appendix 3, page 63 and 64.

Q. And tell me this: In those points, which data did you include into your analysis relating to the distribution, composition, types of units of the Ministry of the Interior and the Army of Yugoslavia at the material time and material period in the given regions? Because you have northern, eastern, southern, and so on, those different regions. So which data of that type did you use and incorporate?

A. We include no data on Yugoslav forces or force movements. It is for that reason that there is no statistical analysis in this report of Yugoslav activity. We did observe one coincidence, which was that after the Yugoslav government said on television on the evening of April 6th that they would observe a cease-fire, we see that killing and migration in the subsequent four days after that announcement are at their lowest point since the beginning of the conflict. At that same period, NATO and KLA activity increase greatly relative to their earlier rates. However, as we observe in the original report, that is suggestive but by no means proof of -- of this activity, that the Yugoslav government was the cause of killing and migration. We simply observe that there is a coincidence there that is interesting.

Q. An interesting coincidence. Right. But it's not proof of the ¹⁹⁹⁶⁸ assertion that the activities of the Yugoslav authorities were the reason for the migration; is that what you say, Mr. Ball?

A. I think more formally we should state that the coincidence of the Yugoslav announcement of the cease-fire with the dramatic reduction in killing and migration is consistent with the argument that the Yugoslav government was the cause of killing and migration. It is not proof. Statistics do not offer positive proof. Statistics can reject hypotheses. But when we find statistical patterns that are consistent with hypotheses, we say just that, that they are consistent with, or sometimes the formal language is that they are suggestive of agreement with the hypothesis or argument.

Q. All right. Now, movements of numbers of people who were fleeing from the war-inflicted areas, the number of refugees and people fleeing and so on, was that one of the topics of your analysis as well? Were you asked to look at that topic?

A. I was not asked to look at it, but yes, we did look at it. That was the basis of an earlier report, which we discussed during my testimony a year ago, entitled "Policy or Panic." The analysis of migration is also to be found in many of the graphs in our report to the Tribunal last year, as well as in, for example, figure 2 on page 3 of the report we are discussing this morning.

Q. Let's take any page. For example, page 2 -- or rather, figure 2, the estimated total refugee migration and killings over time. It's on page 3. And then you can see phase 1, phase 2, and phase 3. And you say, "People leaving their homes." That's what it says here; right? ¹⁹⁹⁶⁹

A. Yes.

Q. And then according to statistics, in phase 1, where according to you was the lowest intensity of the NATO bombing; right? Phase 1 was the lowest NATO bombing intensity; is that right?

A. We have not argued that, but there is lower intensity during that period than in the later periods.

Q. All right. That's very good. So in phase 1, people flee the most. That's when most of the migrations and fleeing takes place. Most people -- lots of people are leaving their homes during phase 1; right?

A. That's correct.

Q. You know, Mr. Ball, I'm sure, what "Homo sapiens" means.

JUDGE MAY: Mr. Milosevic, let's move on. You've got a quarter of an hour more. That will give you, we think, sufficient to cross-examine on this corrigendum.

MR. MILOSEVIC: [Interpretation]

Q. Mr. Ball, the beginning of the bombings, is it logical to assume that the beginnings of this would trigger off the effect where the largest number of people who are afraid and who cannot know what the intensity of the bombings will be in future --

JUDGE MAY: You misunderstand his evidence. His evidence is mathematical. He has come to -- he has done various tests and he's come to various conclusions. You can ask him about that. The questions you are asking him are matters which you'll have to address to us in due course, as to why you say this refugee flow occurred. What may seem common sense or something like that is not the point. The point is to ask ¹⁹⁹⁷⁰ him about his own model and his evidence. It's an opinion. He's an expert. We will have to consider what weight to give to it. Now, if you've got any more questions about his report, you can ask them, or particularly about the corrigendum; otherwise, we'll bring this to an end.

THE ACCUSED: [Interpretation] Very well. Mr. May, here we have a question about models, or rather, the logics of thinking that Mr. Ball applied.

MR. MILOSEVIC: [Interpretation]

Q. And I'm asking you, Mr. Ball, did you take into consideration in your model and the data and information you present fear, the fear that bombing produces in people?

JUDGE MAY: That's precisely the sort of question which I've told you you can't ask him. He's not here to give evidence about psychology. He's giving evidence about a report.

Now, it sounds as though you've really got no more to ask him on the corrigendum, so we might as well move on and not waste further time.

THE ACCUSED: [Interpretation] Well, let's move on. I have not used up all my time yet, Mr. May.

MR. MILOSEVIC: [Interpretation]

Q. Mr. Ball, tell me this, please: In any one of your figures, graphs, in addition to the curves that we see according to your measurements show the trend of migration, that is to say, the people leaving their homes, do we also see a line which would indicate the intensity of the NATO bombings themselves? ¹⁹⁹⁷¹

A. In these graphs, there is no such observation.

Q. Why not?

A. Because we have incorporated the analysis of the NATO patterns in the statistical model, which I discussed during my direct examination. So, for example, the effect -- the statistical effect of those patterns, of both the NATO airstrikes and the KLA activity, can be seen in, for example, figure 10 in the Residual line, which we have already covered; similarly in figure 11, also in the Residual line, which we have covered; and then in more precise technical detail in figure 19, page 11, entitled "Regression coefficients." It is in particular the coefficients for each of the KLA measures and the NATO measure that the effect on killing or migration of the KLA or NATO activity can be seen.

Q. All right. And tell me this, Mr. Ball: What were your data sources for KLA activities?

A. We covered this extensively in my original testimony last year, but to review, those sources are shown in the Appendix 3 -- I'm sorry, I guess I've already given you that page number. And perhaps we can look at them one at a time, if they're of interest, but they're listed there.

Q. As documents exist which indicate KLA activity, which is in possession of the opposite side as well, there are also documents which indicate the orders given by the KLA to the population, telling them to leave the territory of Kosovo and move to Macedonia, Albania, and Montenegro. Did you take into consideration those documents too when you conducted your analyses of the effects of KLA activity?

THE ACCUSED: [Interpretation] I'm asking Mr. Ball, Mr. May, ¹⁹⁹⁷² whether he took into consideration certain documents. That's quite a legitimate question, I assume.

JUDGE MAY: You cross-examined extensively a year ago, and we've heard all this before. But yes, since it's been asked, perhaps you would deal with it.

THE WITNESS: We did not take into consideration any documents from any party that expressed intention, in statistical terms. The only intention we considered was to -- was not statistical but, rather, just to observe the dramatic decline in killing and migration that occur on the night of April 6th/April 7th.

MR. MILOSEVIC: [Interpretation]

Q. All right. In your conclusions, you state that you're continuing your investigation of the data using more complex methods; is that right?

A. Yes, it is.

Q. This observation of yours, can I interpret it as your view and assessment that not even from your standpoint is this analysis comprehensive and all-embracing, nor that you used certain more complex and reliable methods?

A. For scientific publications, the conclusions -- the substantive conclusions here I believe are no longer in question. We've already published these results in a peer review journal, so that's not the question.

However, the data are uniquely interesting from a scientific perspective and invite substantial additional kinds of experimentation. We have published the data on the Internet specifically to invite the ^{19973 BLANK PAGE 19974} scientific community to continue considering these -- these phenomena, but I consider it unlikely in the extreme that any change in the interpretation of the core hypotheses we have proposed here should emerge. Quite the contrary. Our initial and much more detailed kinds of analysis that we are beginning for scientific publication, we hope, have found precisely the same results that are found here.

Q. Let's just clarify what we're talking about, not to speak of two different things. When you say that you published that, observed that, and so on, this relates to the meaning that was put forward on page 1 of your report that the information contained and the positions are exclusively used for education and information purposes, in order to educate and inform; is that right? It's the last sentence, actually, which says that: "This publication and any forms and agreements herein are intended for education and informational purposes only." Before that you say that it cannot be considered to be legal advice for specific cases. And before that again, it states that it does not represent the positions of the institutions you mention.

So when you say that you have concluded this portion, you're talking about the scientific aspects of your analysis and checking out certain statistical methods which you applied; is that correct?

A. I am a scientist; my colleagues are scientists, and of course what we are doing is for scientific purposes. I believe that's my purpose here today.

Q. Very well. And tell me, finally, to what extent the type of sources of information influences any analysis, including your own. The ¹⁹⁹⁷⁵ different types of resources and sources of information, to what extent that has an influence on an analysis, any analysis, including your own.

JUDGE MAY: Can you answer that?

THE WITNESS: Sorry?

JUDGE MAY: Do you understand that question?

THE WITNESS: I think it requires a course of methodology to respond to. It's not an -- not a question that can be answered easily. I can try to give a very brief answer, if that would be useful to the Court.

JUDGE MAY: Well, since you've been asked, a brief answer, please.

THE WITNESS: Of course the sources of data and their type influence our choice of statistical methods that should be used to analyse them. In particular, when we have multiple projects that attempt to enumerate deaths, we can use the method that's used here to compare those multiple projects, determine which deaths have been reported in common across the projects, and use that information to make an estimate of deaths that have not been reported to any project. That is the basis for our analysis of killings.

The purpose of using a method of that kind is explicitly to understand how many people have not been reported -- how many people's deaths have not been reported to any project. So this is one example of considering the sources of data, in this case interviews by the ABA/CEELI, Human Rights Watch, the Organisation for Security and Cooperation in Europe, and the forensic reports of bodies exhumed and identified. These four sources can be used with this method called "multiple systems estimation." ¹⁹⁹⁷⁶

MR. MILOSEVIC: [Interpretation]

Q. So only those four sources. Now, tell me, please, Mr. Ball: These numbers, as far as I was able to understand from something you said during the examination, they apply exclusively to Albanians; isn't that right?

A. These four sources are used to estimate killings. Let's distinguish that from data and analyses used to estimate migration patterns.

To the second part of the question, yes, we have estimated only the deaths of ethnic Albanians. That's, I believe, clearly explained in the first report, where we explain the entire process by which we matched the deaths across projects to determine which deaths are reported in more than one of the projects. That's Appendix A in the first report.

JUDGE MAY: You can ask one more question.

MR. MILOSEVIC: [Interpretation] Very well. I'm sorry that I was not able to find my way straight away. Because you say here people leaving their homes and people killed, I assumed that you were referring to all the people, not just ethnic Albanians. But do you think -- don't you think that it would be essential, in order to see the effect of everything that was going on, to cover all people? For example, 100.000 Serb refugees from Kosovo, or how many Serbs were killed in Kosovo, and not just ethnic Albanians. In order to provide at least to some extent a more objective presentation of the statistical data that you have produced?

JUDGE MAY: Can you answer that or not? ¹⁹⁹⁷⁷

THE WITNESS: I believe that using the data we have used is sufficient to address the hypotheses we addressed, which is what caused the killing and migration of ethnic Albanians.

JUDGE MAY: Very well. Mr. Kay?

MR. KAY: Nothing further to add to previous cross-examination.

JUDGE MAY: Anything from the Prosecution? We ought to -- we ought to exhibit the corrigendum, before that's forgotten.

MR. NICE: Yes.

JUDGE MAY: It's 67 ...?

THE REGISTRAR: Your Honours, the corrigendum will be marked Prosecutor's Exhibit 67.1, and the B/C/S translation will be 67.1A.

MR. NICE: Thank you. Re-examined by Mr. Nice:

Q. Dr. Ball, you've been asked questions about the source of your material. For NATO bombing, did you turn to material provided by Yugoslavia itself?

A. Nearly all -- nearly all the data -- I'm sorry. Nearly all the data for NATO bombing comes from Yugoslav sources.

Q. Any particular reason for using that, as opposed to any other source?

A. My calls to the US Department of Defence were unanswered, and the Yugoslav sources were published in press briefings and on the Internet, so they were easily accessible. We also felt - and I believe we discussed ¹⁹⁹⁷⁸ this last year - is that using the Yugoslav sources on airstrikes would be the strictest possible test of the hypothesis that the NATO activities had caused killing and migration, since it was some of those same sources that had advanced that idea.

JUDGE MAY: Dr. Ball, that concludes your evidence. Thank you for coming back again to add to it. You are free to go.

[The witness withdrew]

JUDGE MAY: Now, Mr. Nice, is there anything to raise before we go back into closed session?

MR. NICE: I don't think so, no.

JUDGE MAY: We'll go into closed session.

[Closed session]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted] ¹⁹⁹⁷⁹ Pages 19979-20037 - redacted - closed session

²⁰⁰³⁸

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

[redacted]

--- Whereupon the hearing adjourned at 2.08 p.m., to be reconvened on Monday,

the 5th day of May, 2003, at 9.00 a.m.