Appendix 5 The use of statistical evidence
An age of uncertainty
Inquiry into the treatment of individuals suspected of people smuggling offences who say that they are children
Appendix 5: The use of statistical evidence
1 Introduction
This short paper is concerned, first, to examine the significance in a particular case of statistical evidence; secondly, to examine what it means to prove a fact on ‘the balance of probabilities’; and finally to examine the significance of relying on statistical evidence to establish an accused person’s age in the context of s 236B of the Migration Act 1958 (Cth).
Section 236B of the Migration Act fixes mandatory minimum penalties for certain people smuggling offences. In particular, the section fixes a mandatory minimum penalty of five years imprisonment for the offence created by s 233C of the Migration Act – the aggravated offence of people smuggling. The aggravated offence of people smuggling involves organising or facilitating the bringing or coming to Australia of a group of at least five unlawful citizens. It is the offence with which most Indonesian crew of people smuggling vessels are charged.
The legislature has made it plain that it does not intend the mandatory minimum penalties for people smuggling offences to apply to children. Subsection 236B(2) provides:
This section does not apply if it is established on the balance of probabilities that the person was aged under 18 years when the offence was committed.
This paper is not concerned with the issue of which party bears the onus of establishing that an individual charged is under the age of 18 years. This latter issue is complicated by, among other things, the related issue which arises when a particular court has exclusive jurisdiction to hear and determine a charge for an offence alleged to have been committed by a child (see for example, the Children’s Court of Western Australia Act 1988 (WA)).
Nor is this paper concerned with the debate about the applicability of available statistical information to a population of young Indonesian males who have arrived in Australia after travelling to Australian on a vessel carrying asylum seekers and who say that they are children.
The issue of how informative skeletal maturity, as revealed by a wrist x-ray, is on the issue of whether an individual is over the age of 18 years is considered in Chapter 2 of this report. That issue is not further considered in this paper.
2 Statistics are concerned with populations – not with individuals
It is universally acknowledged that a person’s precise chronological age cannot be determined from a wrist x-ray or from any other biomedical marker. However, statistical information is available as to the approximate age at which young people generally will show a mature x-ray.
Where an individual suspected of people smuggling has been shown by x-ray to have a mature wrist, it has been common in Australian courts for opinion evidence to be given in terms such as:
The probability of the subject of the x-ray being less than 18 years old at 17 November 2010 is approximately 22%.
Chapter 2 of this report reveals the inaccuracy of this suggested probability. However, since it may prove possible to calculate a more accurate statistical probability of a person’s being a particular age, it is useful to understand the import of evidence of this kind. For convenience this issue is here examined on the (false) assumption that the above probability is accurate.
Taken at face value, a statistical probability of this kind has a tendency to mislead. The author of the above statement, in referring to the probability of the individual being less than 18 years of age, sought to identify the statistical probability of the individual being less than 18 years of age; he was not saying anything directly concerning the individual the subject of the x-ray.
This is because statistics are concerned, not with individuals, but with populations. The expert opinion quoted above intended to identify the mathematical probability within a given population of an individual who is less than 18 years of age showing a mature wrist on x-ray. Mathematical probabilities rely on the principle of indifference which is discussed by Ligertwood and Edmond in Australian Evidence: A Principled Approach to the Common Law and the Uniform Acts.[1316]
The true import of the above statement is that, assuming that the subject of the x-ray comes from a population comparable to that from which the relevant statistics have been derived, if the individual the subject of the wrist x-ray, and a sufficiently large number of other individuals from the same population who have x-rays showing the identical degree of skeletal development, are all found to be adults, this finding will be correct in 78 out of every 100 cases – but inevitably incorrect in 22 out of every 100 cases. Put another way, the statistic accepts that approximately 22% of individuals with an x-ray identical to the subject’s x-ray will be less than 18 years of age and says nothing about whether the subject might be one of them. To assess the subject’s age on the basis of this statistical probability alone will be to act to his serious prejudice should he be one of the 22% who matures early.
The above reference to a ‘sufficiently large population’ is intended to reflect the difference between a theoretical and an empirical probability. This difference can be illustrated by reference to the toss of a coin. We take it as a given that, with a sufficiently large population, the toss of a coin will come down heads: tails 50:50. This result, however, may not be achieved in 100 tosses or even 1,000; it is a theoretical probability which might only be replicated empirically in a much larger number of tosses. In a limited sample the empirical outcome might be heads: tails 75:25 or 25:75 or some other ratio. The smaller the sample, the less likely it is that the theoretical ratio will be achieved.
The significance of the failure of statistical evidence to address any particular case was examined by Dant in ‘Gambling on the Truth: The Use of Purely Statistical Evidence as a Basis for Civil Liability’.[1317] In this article the learned author usefully discusses the difference between rational decision-making for gambling purposes and rational decision-making for judicial purposes. She makes the point that, on the basis of the statistic ‘60% of the marbles in the sack are red’, it is rational to bet that a particular marble withdrawn from the sack is red even though the statistic does not assert anything about the particular marble selected.[1318] Long term winnings will be maximised by betting red on every occasion on which a marble is drawn from the sack. By contrast, the rational basis required for fact finding for judicial purposes is logical or evidentiary support in the context of the individual case.[1319] While it might be rational to bet that the marble is red as this will maximise long term winnings, it would not be rational on the basis of the statistic alone to form a belief that the marble is red; any belief that the particular marble is red will be no more than a guess.
The learned authors of Australian Evidence: A Principled Approach to the Common Law and the Uniform Acts (5th edition) analyse the nature of mathematical probabilities in a similar way. They state:
mathematical probabilities describe the nature of entire classes and say nothing about the individual members of that class. To say that an individual hypothesis is proved where its probability exceeds 0.5 is to say that, although there can be no certainty about the occurrence of this individual hypothesis, nevertheless, to act on it as if it had occurred will in the long run produce more correct decisions than incorrect decision. But incorrect decisions there will be.[1320]
Legal writers who have considered this issue not infrequently refer to two theoretical cases to illustrate the unsatisfactory nature of mathematical probabilities as proof of a particular issue – even where the issue in question is only required to be proved on the balance of probabilities. While these theoretical cases tend to be unrealistic, as few cases will arise in which mathematical probabilities constitute the only evidence available and joinder rules would probably obviate some of the difficulties envisaged, the cases usefully illuminate the deficiencies of evidence based on statistics alone.
The first case, found in Cohen’s The Probable and the Provable, postulates that 499 people have paid for admission to a rodeo and 1,000 people are found to be at the event, with ‘A’ being one of those present.[1321] No other testimony is available as to whether A paid for his seat. Plainly enough there is a 0.501 probability that A, and indeed all others present, did not pay, but it would be manifestly unjust to hold A liable. Apart from anything else, there is a 0.499 probability that he did pay. Professor Glanville Williams has suggested that it would remain wrong to give judgment against A even if only 50 of the 1,000 people present had paid, raising the mathematical probability to 0.95.[1322]
The second theoretical case is one initially raised by Professor Glanville Williams himself.[1323] He hypothesises that the Blue Bus Company has far more buses on the road than the Red Bus Company and points out that this constitutes no reason in law to assume that the plaintiff was knocked down by a blue bus rather than a red bus – otherwise the Blue Bus Company would have to pay damages in all cases in which the sole issue is the ownership of the offending bus and it cannot be shown whether the bus was blue or red.
There is judicial support for the above concerns with respect to statistical evidence. In State Government Insurance Commission v Laube[1324] an insurer sued an insured to obtain reimbursement of an amount paid in damages to a pedestrian who had been struck by a car driven by the insured. The insurer claimed that the insured had failed to comply with a term of his policy by driving while so much under the influence of alcohol as to be incapable of exercising effective control of the vehicle. It was proved that the collision occurred at 1:20am and that at 2:35am the insured’s blood alcohol reading was 0.155. Expert evidence was given that, with such an alcohol content, most but not all persons would have their faculties impaired to a degree that would significantly impair their ability to exercise effective control of a vehicle. All three judges held that the insurer had failed to prove its case. King CJ, after referring to the evidence, said:
The most that can be said is that it is statistically more probable than not that any individual with such a blood alcohol level would be incapable of exercising effective control. ... I am clearly of the opinion that the statistical fact that a particular proposition is true of the majority of persons cannot of itself amount to legal proof on the balance of probabilities that the proposition is true of any given individual.[1325]
The fact that most people with a blood alcohol level of 0.15 are incapable of exercising effective control of a motor vehicle does not establish against any individual with that blood alcohol level that the individual is so incapable.[1326]
Criticisms have been made of the above decision, for example by the Hon Mr Justice D H Hodgson in ‘The Scales of Justice: Probability and Proof in Legal Fact-Finding’.[1327] Hodgson J takes the view that:
subject to the requirement of adequate material concerning the particular case, and, in particular, the calling of appropriate evidence by the party bearing the onus of proof, evidence of the kind given in Laube should be enough to enable an inference to be drawn, on the balance of probabilities, if the defendant chooses not to give evidence.[1328]
His Honour’s criticism therefore has limited application to a case in which statistical evidence is the only relevant evidence available to be adduced. Indeed, Hodgson J accepts that ‘mere mathematical probability’ constitutes ‘inadequate material’ on which to base a judicial finding.[1329]
Sir Richard Eggleston has also criticised the decision in State Government Insurance Commission v Laube but it is significant to note that the authorities to which he points were concerned either to estimate a future event (e.g. to predict the life expectancy of a particular person) or involved high probabilities. He additionally points to some statements in the United States which reject mathematical probability as a basis for fact-finding.[1330]
It may be doubted that the approach adopted by King CJ would now attract judicial support where the relevant statistical probabilities are extremely high. It is no longer doubted that statistical evidence based on DNA analysis can be probative in a particular case.[1331] Ordinarily, however, the probabilities generated from DNA evidence are extremely high and other evidence making the defendant a person of interest is available. In R v Galli, for example, where the paternity of a foetus was in issue, it was accepted that the defendant was one of only a very small group of men who had access to the mother of the foetus and one test indicated that he was 172 times more likely to be the father of the foetus that a person taken at random from the population and a second test put that probability at 14 330 times more likely.[1332]
Nonetheless, it has been recognised that, even where statistical evidence of high probabilities is received by a court, it is inappropriate for the decision-maker, whether judge or jury, to approach the issue of chance on a strictly mathematical basis.[1333] In R v Galli, Spigelman CJ referred to the danger that a statistical computation by ‘its very precision and concreteness suggests an exactness which a statistical distribution does not have’.[1334] The Chief Justice recognised the desirability of counterbalancing these features of statistical calculations, observing:
Findings of fact in both civil and criminal cases require common sense judgment and the tribunal of fact is required to reach a level of actual persuasion on the whole of the evidence. This does not involve a mechanical application of probabilities.[1335] (citations omitted)
3 What does it mean to prove a matter of fact on the balance of probabilities?
It is common to identify two broad judicial approaches to what is required in a judicial proceeding to establish a fact on the balance of probabilities. The first approach is that which calls for the decision-maker to ‘feel an actual persuasion of its occurrence or existence before it can be found’.[1336] The second approach is that which looks, not for actual belief, but rather for ‘a more probable inference in favour of what is alleged’.[1337]
The well-known observation of Dixon J in Briginshaw v Briginshaw[1338] is commonly regarded as an example of the ‘actual belief’ approach:
Fortunately ... at common law no third standard of persuasion was definitely developed. Except upon criminal issues to be proved by the prosecution, it is enough that the affirmative of an allegation is made out to the reasonable satisfaction of the tribunal. But reasonable satisfaction is not a state of mind that is attained or established independently of the nature and consequence of the fact or facts to be proved.
The seriousness of an allegation made, the inherent unlikelihood of an occurrence of a given description, or the gravity of the consequences flowing from a particular finding, are considerations which must affect the answer to the question whether the issue has been proved to the reasonable satisfaction of the tribunal.
The statement of Lord Simon in Davies v Taylor that ‘the concept of proof on the balance of probabilities ... can be restated as the burden of showing odds of at least 51 to 49 that such-and-such has taken place or will do so’ is an example of the ‘more probable inference’ approach.[1339] Another example of this approach is found in Malec v J C Hutton Pty Ltd where Deane, Gaudron and McHugh JJ stated:
A common law court determines on the balance of probabilities whether an event has occurred. If the probability of the event having occurred is greater than it not having occurred, the occurrence of the event is treated as certain.[1340]
The different approaches suggested to exist in judicial authority to finding a fact proved on the balance of probabilities are probably more apparent than real in their practical application. This is because in every civil case the court will be required to assess the totality of the evidence before it for the purpose of assessing which of (ordinarily two) competing potential factual findings appears the more likely to be true. For present purposes, however, it is important to note that neither approach would appear to sanction decision-making on the basis of pure statistical evidence. As the discussion above concerning the nature of statistical evidence illustrates, properly understood, a mere statistic cannot of itself amount to proof on the balance of probabilities in a particular case. This is because it does not speak to the individual case but rather to a population. The only exception to this may be where the statistic conveys a very high probability as in the case of most DNA evidence.
In the words of Ligertwood and Edmond:
To rely on a mathematical probability in a general class by ready (and unwarranted) reliance upon the principle of indifference may produce more right decisions in the long run, but to do so flies in the face of a system of justice which seeks truth in individual cases.[1341]
Assuming that there is a real difference between the two approaches identified above, it may be observed that in enacting the Evidence Act 1995 (Cth) the legislature appears to have expressed a clear preference for at least some elements of the Briginshaw approach.[1342] Section 140 of the Evidence Act provides:
(1) In a civil proceeding, the court must find the case of a party proved if it is satisfied that the case has been proved on the balance of probabilities.
(2) Without limiting the matters that the court may take into account in deciding whether it is so satisfied, it is to take into account:
(a) the nature of the cause of action or defence; and
(b) the nature of the subject-matter of the proceeding; and
(c) the gravity of the matters alleged.
Because of its reference to a ‘civil proceeding’, s 140 must be assumed to have no direct application to a criminal proceeding in which an issue must be proved ‘on the balance of probabilities’. However, it seems reasonable to assume that the reference to ‘on the balance of probabilities’ in s 236B(2) of the Migration Act, which was enacted later than the Evidence Act, was intended to reflect the content of s 140(2) of the Evidence Act.
4 Reliance on statistics concerning age in the context of section 236B of the Migration Act.
As noted above, s 236B of the Migration Act fixes mandatory terms of imprisonment for people smuggling offences provided that the person convicted was over the age of 18 years at the date of the offence. The exclusion of children from the reach of the provision reflects not only a natural reluctance to subject children to mandatory imprisonment in adult facilities but also Australia’s obligations as a party to the Convention on the Rights of the Child.
While the legislature has decided that age need only be established on the balance of probabilities for the purposes of s 236B, the consequences of an inaccurate determination that the individual charged is an adult will be extremely serious. It will almost certainly lead, on conviction, to a child being sentenced to a sentence not intended by the legislature to apply to a child and to a child being imprisoned in an adult facility where he or she will face the dangers necessarily inherent in being so held.
As a consequence, any determination that an individual suspected of aggravated people smuggling is an adult ought not to be based on a mere statistical probability – except perhaps where that probability is exceptionally high. Any other approach runs the risk of disadvantaging in a systematic way individuals who mature early – as the statistics plainly reflect that a significant proportion of individuals will. Moreover, in reaching such a determination, a court would need to give careful consideration to the fact that the determination was being made for the purposes of a criminal proceeding concerning a serious offence in which the liberty of the accused, for a significant period of time, was at stake – and where error could result in a child being convicted as an adult, subjected to a mandatory sentence of imprisonment only intended to apply to adults, and thereafter held in an adult jail.
Catherine Branson
June 2012
[1316] A Ligertwood and G Edmond, Australian Evidence Act: A Principled Approach to the Common Law and the Uniform Acts (5th ed, 2010), paras [1.45]–[1.46].
[1317] M Dant, ‘Gambling on the Truth: The Use of Purely Statistical Evidence as a Basis for Civil Liability’ (1988–1989) 22 Columbia Journal of Law and Social Problems 31.
[1318] M Dant, above, p 38.
[1319] M Dant, above, p 44.
[1320] A Ligertwood and G Edmond, note 1, para [1.35].
[1321] LJ Cohen, The Probable and the Provable (1977), p 75, as cited in Hon Sir R Eggleston, ‘Focussing on the Defendant’ (1987) 61 Alternative Law Journal 58.
[1322] G Williams, ‘The Mathematics of Proof’ (1979) Criminal Law Revue 297, 304, as cited in Hon Sir R Eggleston, ‘Focussing on the Defendant’ (1987) 61 Alternative Law Journal 58.
[1323] G Williams, above.
[1324] State Government Insurance Commission v Laube (1984) 37 SASR 31.
[1325] State Government Insurance Commission v Laube (1984) 37 SASR 31, 33.
[1326] Above, see also TNT Management v Brooks (1979) 23 ALR 345.
[1327] See, for example, Hon Justice D Hodgson ‘The Scales of Justice: Probability and Proof in Legal Fact-Finding’ (1995) 69 Alternative Law Journal 731. See also Hon Sir R Eggleston ‘Focusing on the Defendant’ (1987) 61 Alternative Law Journal 58.
[1328] D Hodgson, above, p 742.
[1329] D Hodgson, above, p 736.
[1330] G Williams, note 7.
[1331] R v Galli [2001] NSWCCA 504. See also Yusuf Aytugrul v The Queen [2012] HCA 15, where the High Court recently dismissed an appeal by the appellant against his conviction of murder based on DNA opinion evidence (and other circumstantial evidence) expressed as an exclusion percentage and accompanied by a frequency ratio and explanation of the two. The Court unanimously held found that the DNA evidence was not expressed to the jury in a prejudicial way that would enliven s 135 of the Evidence Act 1995 (NSW). While the DNA evidence states as an exclusion percentage was high, 99.9 per cent, the evidence given was not said to establish that the DNA profile found in the hair came from the appellant.
[1332] R v Galli [2001] NSWCCA, [30]–[31].
[1333] R v Milat (1996) 87 A Crim R 446, 452 (Hunt CJ).
[1334] R v Galli [2001] NSWCCA 504, [50].
[1335] R v Galli [2001] NSWCCA 504, [55]. See also, Seltsam Pty Ltd v McGuiness [2000] NSWCA 29.
[1336] See for example, Communications, Electrical, Electronic, Energy, Information, Postal, Plumbing & Allied Services Union of Australia v Australian Competition & Consumer Commission (2007) 162 FCR 466, 31.
[1337] See for example, Bradshaw v McEwans Pty Ltd (1951) 217 ALR 1, 5.
[1338] Briginshaw v Briginshaw [1938] HCA 34; (1938) 60 CLR 336, 361–362.
[1339] Davies v Taylor [1974] AC 207, 219; [1972] 3 All ER 836, 844.
[1340] Malec v J C Hutton Pty Ltd (1990) 169 CLR 638, 642–643 (Deane, Gaudron and McHugh JJ).
[1341] A Ligertwood and G Edmond, note 1, para [1.45].
[1342] In Qantas Airways Ltd v Gama [2008] FCAFC 69, Branson J stated (at 139): ‘The correct approach to the standard of proof in a civil proceeding in a federal court is that for which s 140 of the Evidence Act provides. It is an approach which recognises, adopting the language of the High Court in Neat Holdings, that the strength of the evidence necessary to establish a fact in issue on the balance of probabilities will vary according to the nature of what is sought to be proved – and, I would add, the circumstances in which it is sought to be proved.’ French and Jacobsen JJ agreed with Branson J on this point.