Australian Iron & Steel Pty Ltd v Banovic

In Pt III, s. 24 defines the elements of discrimination on the ground of sex. Sub-section (1) reads:

A person discriminates against another person on the ground of his sex if, on the ground of -

(b) a characteristic that appertains generally to persons of his sex; or

(c) a characteristic that is generally imputed to persons of his sex,

he treats him less favourably than in the same circumstances, or in circumstances which are not materially different, he treats or would treat a person of the opposite sex. Sub-section (3) reads: A person discriminates against another person on the ground of his sex if he requires the other person to comply with a requirement or condition -

(a) with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply;

(b) which is not reasonable having regard to the circumstances of the case; and

(c) with which the other person does not or is not able to comply.

Non-compliance or an inability to comply with the "requirement or condition" to which sub-s. (3) refers may not amount to a "characteristic" falling within par. (b) or (c) of sub-s. (1), but the test of "substantially higher proportion" means that the "requirement or condition" is a sex-related criterion. For one person to require another to comply with a sex-related criterion adverse to the sex of the other and with which the other cannot comply is sex discrimination in fact; if par. (b) of s. 24(3) is satisfied, it is sex discrimination in law.

The proportions - one substantially higher than the other - of members of the respective sexes who "comply or are able to comply" are proportions of a group which s. 24 does not identify. The group is not a standard group common to all applications of s. 24. Section 24, it will be remembered, does not prohibit discrimination on the ground of sex. The prohibitions are found in s. 25 and it is to that section that reference must be made to ascertain the relevant group. The group must be identified by reference to the particular activity in which, or the particular act by means of which, one person is said to discriminate against another. In that way, the nature of the criterion is ascertained in the practical context in which it operates. In cases arising under s. 25, the group is constituted by the persons of whom the employer requires compliance with the relevant "requirement or condition" when the employer engages in the particular activity or does the particular act mentioned in that section. That is the group to whom the employer applies the supposed sex-related criterion.

In these proceedings we are concerned with two provisions in s. 25, namely, s. 25(1)(b) and s. 25(2)(c). They read as follows:

(1) It is unlawful for an employer to discriminate against a person on the ground of his sex -

(b) in determining who should be offered employment; or

(2) It is unlawful for an employer to discriminate against an employee on the ground of his sex -

(c) by dismissing him or subjecting him to any other detriment.

When an employer is determining who should be offered employment out of a group of applicants and he requires applicants to comply with what is said to be a sex-related criterion, the relevant comparison is between the proportion of male applicants who comply or are able to comply with the criterion and the proportion of female applicants who comply or who are able to comply with the criterion. In a case where knowledge of a vacancy is disseminated (for example, by advertisement), the proportions would not be of males or females in a known group of applicants but of males and females in the group who might have applied for appointment had appointment not been subject to compliance with the supposed sex-related criterion.

Here it is alleged against the employer that it discriminated against each of the first respondents, its employees, on the ground of sex "by dismissing [her] or subjecting [her] to any other detriment". Finding it necessary to reduce its workforce, the employer retrenched employees by application of the criterion "last on, first off" (or by requiring compliance with a "requirement or condition" of specified "gate seniority"). The employer applied the criterion of "last on, first off" indifferently to all members of its workforce requiring each employee to have been employed before the cut-off date (6 January 1981 on one claim; 30 September 1980 on another) in order to avoid dismissal or the threat of dismissal. The group to which that requirement was directed was the employer's entire workforce. The question which arises is whether the proportion of men employed in that workforce who were employed before those dates was substantially higher than the proportion of women employed in that workforce who were so employed. If "yes", the employer's requirement is to be seen as sex-related and (subject to satisfaction of s. 24(3)(b)) the adoption of the criterion of "last on, first off" in dismissing employees can properly be regarded as discriminatory on the ground of sex. But if the answer is "no", it cannot properly be said that, by dismissing any one of the first respondents, the employer discriminated against that employee on the ground of her sex. In fact, the answer is: no. At the times of the threatened dismissals and actual dismissals, the relevant proportions of the sexes employed in the workforce who complied with the condition of gate seniority were, in the case of both male and female employees, approximately 94 per cent. The statutory test reflects what is manifest in fact: by retrenching its employees according to the criterion "last on, first off", the employer did not discriminate on the ground of sex.

To avoid that conclusion, the respondents seek to reconstitute the group to which reference must be made in ascertaining the proportions of the sexes who complied with the condition that the employee should have been employed before a specified date. McHugh J. demonstrates why it is inappropriate to look to any group other than the group of the employer's workforce. I would add that it is meaningless to look to any group other than the workforce when the relevant acts of alleged discrimination are the dismissals of particular employees and the "requirement or condition" with which a person was required to comply to avoid dismissal was employment before a specified date. The chief argument against this conclusion is that the employer had constituted the workforce at the time by unlawfully discriminating against women on the ground of their sex in determining who should be offered employment (s. 25(1)(b)) and that it is wrong to test the proportions of male and female employees who complied with the requirement of employment before the specified date by reference to a group which had been so constituted. I turn to that argument.

The requirement of gate seniority is not, on its face, discriminatory on the ground of sex: it is, to use the jargon, "facially neutral". It does not attract s. 24(1)(a). When an employer in dismissing employees applies a facially neutral requirement indifferently to all members of his workforce and the respective proportions of men and women within that workforce who comply or are able to comply with the requirement are not substantially higher and lower, the elements of discrimination falling within s. 24(3)(a) do not appear. Although the constitution of the group has been affected by earlier unlawful discrimination, the Act does not operate by reference to a hypothetical workforce reconstituted to eliminate the effects of the earlier unlawful discrimination. The application of s. 24(3)(a) does not depend on hypothetical groups. If the application of the facially neutral requirement to such an hypothetically reconstituted workforce were to produce substantially higher and lower proportions of compliance or ability to comply, it would be necessary to vary or abandon the facially neutral requirement and to dismiss employees of one sex rather than another in order to produce a workforce from which the effects of the earlier discrimination would be removed. But it offends ss. 24(1) and 25 to apply a criterion which is not facially neutral in order to alter the composition by the sexes of a workforce. Dismissals for the purpose of advancing the members of one sex at the expense of members of the other sex amounts to unlawful discrimination by virtue of ss. 24(1)(a) and 25(2)(c). Putting the proposition in another way, the Act does not provide for "reverse discrimination" or "affirmative action" by an employer in dismissing employees from the employer's workforce in order to undo the effect of prior unlawful discrimination in its constitution.

To accept the respondents' argument, the employer would have been bound to dismiss from its workforce men who would not have been dismissed upon application of the facially neutral requirement of gate seniority in order to retain in the workforce women who had been adversely affected in their seniority by discrimination practised by the employer at an earlier time. But if an employer were to dismiss men simply on the ground of their sex - whether in order to undo the effect of earlier discrimination or for any other reason - the employer would be guilty of unlawfully discriminating against men on the ground of sex. Acts of unlawful discrimination are not excused because they are designed to remedy the consequences of earlier unlawful discrimination. The opposite to "discrimination on the ground of sex" is not discrimination against the opposite sex; it is non-discrimination. In the enthusiasm for equality, whether of numbers or proportions, between the sexes in a workforce it is as well to remember that the object of any dismissal is an individual who has no personal obligation to underwrite the perfect fulfilment of the statutory policy.

The victims of unlawful discrimination must find their remedy in the statutory cause of action for the activity in which, or for the act by which, the employer unlawfully discriminated against them, that is, a remedy in damages for the particular activity or act. Those damages might properly include an amount in respect of lost gate seniority. If the statutory ceiling on damages precludes full compensation for lost gate seniority, the shortfall is not to be made up by the creation of a new cause of action whenever, in an action or act mentioned in s. 25, a non-discriminatory criterion is applied indifferently to the workforce. An employer, who has discriminated unlawfully in employing his workforce, is not placed under a duty to engage in further unlawful discrimination in the selection of the employees whom he will dismiss. The employer is not required to place at the top of his dismissal list a person who has innocently benefited or, even more, a person who has not benefited from the employer's earlier unlawful discrimination. Why should employees bear the burden of rectifying the consequences of past discrimination? It offends justice to dismiss an employee in order to rectify the consequences of an illegality on the part of the employer. The Act does not require dismissals in a priority which will rectify those consequences; to the contrary, it prohibits them.

Decisions in other jurisdictions or under different statutory provisions on the legitimacy of reverse discrimination to undo the consequences of past discrimination have no necessary relevance to the construction of the Act. But for those who seek an analogue from another jurisdiction, reference may be made to Johnson v. Transportation Agency [1] decided under Title VII of the Civil Rights Act 1964 U.S.. The conflict in opinions in the Supreme Court of the United States in that case is mirrored in the conflict revealed in the judgments now published.

For these reasons and for the reasons stated by McHugh J., I would allow the appeal.

The first respondents are eight women ironworkers who, on or after 30 September 1980, obtained employment with the appellant, Australian Iron & Steel Pty. Ltd. ("A.I.S."). Some of them were dismissed in 1982 when A.I.S. retrenched employees on the basis of "last on, first off". Those who were not dismissed were identified as persons who were at risk of dismissal in the retrenchment programme. The only issue in this appeal is one of law. It is whether a finding of the Equal Opportunity Tribunal of New South Wales ("the Tribunal") that the "last on, first off" method of retrenchment involved, in the particular circumstances of this case, unlawful sex discrimination under the Anti-Discrimination Act 1977 N.S.W. ("the Act"). An appeal under s. 118 of the Act against that finding was dismissed by the Court of Appeal of the Supreme Court of New South Wales. From that part of the decision of the Court of Appeal, A.I.S. now appeals to this Court. It is important to stress that appeals to the Court of Appeal under s. 118 are limited to questions of law and that the Court's function on this appeal does not encompass a general review of the Tribunal's finding on the merits.

Some account of the recruitment practices of A.I.S. is necessary to an understanding of the claim that, by adopting the "last on, first off" method of retrenchment, A.I.S. engaged in acts of sex discrimination. Until mid-1980 A.I.S. pursued recruitment practices that resulted in women constituting only a very small proportion of the A.I.S. ironworker workforce. Another result of those practices was that the waiting time for employment as an ironworker was considerably longer for women than for men. The practices were changed in mid-1980. Thereafter the number of women employed as ironworkers increased in absolute numbers and also as a proportion of the total ironworker workforce. By way of illustration, between June 1977 and April 1980 A.I.S. took on 4,289 new ironworker employees of whom only 58 were women; between 1 July 1980 and 22 August 1980 it took on 468 new ironworker employees of whom 71 were women. Notwithstanding this change, it was found by the Tribunal that "the waiting time for women as compared with men was still vastly disproportionate being measured in years rather than a few days or weeks".

A downturn occurred in the steel industry in 1981 and A.I.S. ceased recruiting new ironworkers from September of that year. The downturn continued and in October 1982 A.I.S. announced that it would be retrenching ironworkers as from 14 November 1982: retrenchments would be effected on the basis of "last on, first off"; the group of employees likely to be affected was identified as those who commenced work after 30 September 1980. As it happened, A.I.S. retrenched those ironworkers who commenced work after 6 January 1981.

In the proceedings before the Tribunal it was found that the delays in employing women ironworkers constituted unlawful discriminatory conduct. Damages were awarded to women ironworkers who had suffered loss or damage in consequence. By s. 113(b)(i) of the Act damages may not exceed "$40,000 by way of compensation for any loss or damage suffered by reason of the respondent's conduct". The Tribunal further found that the retrenchments (both as announced and as implemented) constituted discriminatory conduct separate and distinct from that involved in the delays in employing women. On the basis of that finding further orders for damages were made in favour of the first respondents or those whom they represent.

In so far as vulnerability to retrenchment under a "last on, first off" policy was the consequence of discriminatory delay in employment, it is arguable that it constituted loss or damage suffered by reason of that delay and was necessarily to be comprehended in any award of damages based on it. However, that point has not been taken and the only question which presently arises is whether the announced retrenchments and the actual retrenchments constituted acts of unlawful discrimination.

Subject to certain exceptions which are not presently relevant, s. 25(2)(c) of the Act renders it unlawful for an employer to discriminate against an employee on the ground of his sex "by dismissing him or subjecting him to any other detriment". It is not in issue that those persons who were employed after 30 September 1980, both those who were and those who were not retrenched, suffered detriment in so far as they were identified as persons who might be retrenched. Thus, the only issue is whether, in the particular circumstances which existed as a result of earlier recruitment practices, A.I.S., by announcing retrenchments and by implementing those retrenchments on the basis of "last on, first off", discriminated against women ironworkers on the ground of their sex.

The Act relevantly specifies what constitutes discrimination on the ground of sex as follows:

(1) A person discriminates against another person on the ground of his sex if, on the ground of -

(b) a characteristic that appertains generally to persons of his sex; or

(c) a characteristic that is generally imputed to persons of his sex,

he treats him less favourably than in the same circumstances, or in circumstances which are not materially different, he treats or would treat a person of the opposite sex.

(3) A person discriminates against another person on the ground of his sex if he requires the other person to comply with a requirement or condition -

(a) with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply;

(b) which is not reasonable having regard to the circumstances of the case; and

(c) with which the other person does not or is not able to comply.

It was argued on behalf of the first respondents that the "last on, first off" method of retrenchment offended both s. 24(1) and s. 24(3) of the Act. The argument in relation to s. 24(1) was put in this way: the first respondents were retrenched or put at risk of retrenchment because they lacked employment seniority; they lacked employment seniority because the recruitment practices of A.I.S., both before and after mid-1980, resulted in the waiting period for employment as an ironworker being longer for women than for men; therefore they were retrenched because they were women.

Section 24(1) is usually understood to refer to direct discrimination or acts involving different treatment of men and women; s. 24(3) is usually understood to refer to indirect discrimination or acts having a disparate impact on men and women. See, for example, the discussion of the similar provisions of the Sex Discrimination Act 1975 U.K. in James v. Eastleigh Borough Council [2] . Doubtless, the presence of s. 24(3) in the Act takes much of the force from any argument that s. 24(1) should be given a broad application similar to that accorded to provisions in the Civil Rights Act 1964 U.S. and in the Canadian Human Rights Act 1976-1977 proscribing discrimination "because of" sex, race or other characteristic thereby made the subject of unlawful discrimination. The American and Canadian provisions have been applied to acts or decisions made by reference to criteria or standards which are apparently non-discriminatory (sometimes referred to as "facially neutral") but which have a discriminatory effect. In the United States that application is the result of treating evidence of discriminatory effect as prima facie proof of discrimination because of sex, race, or other identified characteristic. See Griggs v. Duke Power Co. [3] ; Albemarle Paper Co. v. Moody [4] ; Dothard v. Rawlinson [5] . However, for that to be done it is necessary for the plaintiff to identify the specific employment practice that is challenged: Wards Cove Packing Co. Inc. v. Atonio [6] . The significance of the concept of onus of proof and related considerations in the area of anti-discrimination law in Canada is discussed in Ontario Human Rights Commission v. Simpsons-Sears Ltd. [7] .

1. [1989] 3 W.L.R. 123, at p. 126; [1989] 2 All E.R. 914, at pp. 916-917; [1990] Q.B. 61, at p. 72. 2. (1971) 401 U.S. 424, at p. 431. 3. (1975) 422 U.S. 405, at p. 425. 4. (1977) 433 U.S. 321, at p. 329. 5. (1989) 57 L.W. 4583. 6. [1985] 2 S.C.R. 536, at pp. 558-559.

Section 24(1) and like provisions in the Act directed to other acts of unlawful discrimination operate by reference to the ground which provides the basis for the act or decision in question. Similar provisions in the Sex Discrimination Act 1975 U.K. and in the Equal Opportunity Act 1984 Vict. have occasioned some discussion of the relevance of intent to discriminate. In Department of Health v. Arumugam [8] , Fullagar J. said of the Victorian legislation that "the "ground" of one's discrimination must be conscious to fall within the proscriptions of the Act". On the other hand, in Reg. v. Birmingham City Council; Ex parte Equal Opportunities Commission [9] , Lord Goff stated that "[t]he intention or motive of the defendant to discriminate is not a necessary condition to liability". His Lordship explained that, if it were otherwise, "it would be a good defence for an employer to show that he discriminated against women not because he intended to do so but (for example) because of customer preference, or to save money, or even to avoid controversy".

1. [1988] V.R. 319, at p. 327. 2. [1989] A.C. 1155, at pp. 1193-1194.

It is not difficult to envisage situations in which the ground of an act or decision may be identifiable as one falling within s. 24(1)(a), (b) or (c) notwithstanding that the act or decision is not actuated by a motive to discriminate. One need go no further by way of example than an act or decision - as in the past frequently happened - denying women certain opportunities by reference to the inadequacy of toilet facilities. And in that situation it is possible that "consciousness" may extend only to the inadequacy of toilet facilities without a full appreciation that that consideration is but an aspect of a characteristic that appertains generally or is generally imputed to women. And there may be other situations in which habits of thought and preconceptions may so affect an individual's perception of persons with particular characteristics that genuinely assigned reasons for an act or decision may, in fact, mask the true basis for that act or decision. Thus, in the ascertainment of the true basis of an act or decision it may well be significant that there is some factor, other than the ground assigned, which is common to all who are adversely affected by that act or decision. In certain situations that common factor may well be seen to be the true basis of the act or decision. And that may also be the case where some factor is identified as common to a significant proportion of those adversely affected.

Even if it could be said that a factor common to all or to a significant proportion of those who were adversely affected by the decision of A.I.S. to retrench by the "last on, first off" method was that they were women, a further finding that that was the true basis of the decision would be necessary to render s. 24(1) applicable. See James v. Eastleigh Council [10] . There is no finding to that effect by the Tribunal. And the argument made on behalf of the first respondents does not even go so far as to suggest that the true basis of the decision to retrench by the method of "last on, first off" was in any way related to the fact that some of the persons who would be retrenched were women. Instead, it merely identifies the factor which caused some women ironworkers to be within the group retrenched or the group likely to be retrenched rather than within the group unaffected by the decision. That is not sufficient to render s. 24(1) applicable.

1. [1989] 3 W.L.R., at pp. 128-129; [1989] 2 All E.R., at pp. 918-919; [1990] Q.B., at pp. 74-75.

Section 24(3) operates by reference to "a requirement or condition". There is some awkwardness in relating that expression to s. 25(2)(c) which renders it unlawful for an employer to discriminate against an employee "by dismissing him or subjecting him to any other detriment". However, it is plain enough that it was a requirement or condition of continued employment with A.I.S. that the employee should have commenced work prior to 6 January 1981. And it was a requirement or condition of being employed without risk of retrenchment that employment should have commenced prior to 30 September 1980. Risk of dismissal and, in most cases, dismissal were the consequences of failure to comply.

The critical consideration in the present appeal is whether the condition or requirement that employment should have commenced before the date which attracted dismissal or the date which put an employee at risk of dismissal is one falling within s. 24(3)(a). That paragraph applies only if the condition or requirement is one "with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply".

Section 24(3) requires the ascertainment of those proportions which complying men and women bear to some larger group or base group. The first question which arises is whether s. 24(3)(a) directs that the proportions be ascertained by reference to separate groups of men and women. In the Court of Appeal Street C.J. was of the view that the proportions could be ascertained by reference to a single group, comprised of men and women, being the "aggregate workforce [of A.I.S.] in November 1982". We do not think that that approach is permitted by s. 24(3)(a). The word "proportion" requires more than mere numerical comparison. A calculation made by reference to the aggregate workforce is, in truth, no more than numerical comparison. And, ordinarily, the simple ratio obtained by reference to a single base group comprised of men and women will be influenced by the sexual composition of that base group. Having regard to the purpose of the Act, s. 24(3)(a) is to be seen as requiring a calculation which will reveal whether sex, as distinct from the sexual composition of the group, is a factor influencing the number of complying men as compared with the number of complying women. For that purpose it is necessary that complying men be ascertained as a proportion of other men and that complying women be ascertained as a proportion of other women.

The more difficult question concerns the identification of the groups of men and women which will enable the proportions of complying men and women to be calculated. On that issue s. 24(3)(a) is wholly silent. In the Court of Appeal, Street C.J. held that the identification of a base group was "a matter of fact to be determined having regard to the context in which the requirement or condition is imposed", adding that "[i]t will be the group to which it is directed and which will be divided or segregated by it". On the other hand, Priestley J.A. (with whom Mahoney J.A. agreed), after a consideration of decisions based on the similar provision in the Sex Discrimination Act 1975 U.K., inclined to the view that the relevant groups were the male and female populations of New South Wales.

As previously indicated, s. 24(3)(a) requires an exercise which will ascertain whether sex is significant to compliance with the condition or requirement in question. It may be expected, as was indicated by Street C.J. in the Court of Appeal, that the base groups which are appropriate to that exercise will vary according to the context in which the condition or requirement is imposed. That being so, there is no warrant for reading s. 24(3)(a) as invariably requiring the calculation of proportions by reference to the general male and female populations. Equally, there is no warrant for reading s. 24(3)(a) as excluding that calculation if it will reveal the significance, if any, of sex to compliance. The silence of s. 24(3)(a) as to the method of identifying appropriate base groups should, we think, be taken as requiring the Tribunal to determine for itself the base groups which will reveal whether sex is a significant factor in compliance. That is not to say that a determination of the appropriate base groups is merely a finding of fact which is beyond appellate review by reason of the limitation on appeals in s. 118 of the Act.

The determination of the appropriate base groups will ordinarily involve the making of findings of fact. But it also involves a reasoning process which is not dissimilar from that involved in the process of determining whether or not evidence is relevant to an issue in a trial. A decision to select particular base groups involves a question of law, at least in so far as it is a question of law whether the base groups selected produce the exercise required by s. 24(3)(a), namely, the ascertainment whether sex is significant to compliance. In this respect, it is necessary to inquire, as was pointed out in the recent decision of the United States Supreme Court in Wards Cove Packing Co. [11] , whether the groups are "too broad" or "too narrow". That exercise is, in essence, the same as an inquiry whether irrelevant considerations were taken into account or relevant considerations were not taken into account.

It was argued on behalf of the appellant that the Tribunal erred in not selecting as base groups those groups of men and women who together comprised the workforce of A.I.S. and who were, to adapt the words of Street C.J. in the Court of Appeal, the persons to whom the condition or requirement was directed and who would be divided or segregated by it. Calculations by reference to those base groups reveal that the proportions of men and women retrenched and put at risk of retrenchment were not significantly different.

The comparison of the proportions of complying men and women to the male and female populations to be divided by the condition or requirement in question will reliably reveal the extent of the significance, if any, of sex to compliance only if sex is not a factor influencing the composition of those populations. In the present case the composition of the A.I.S. workforce had been influenced by the discriminatory recruitment practices of A.I.S. Calculations made by reference to those groups of men and women who together comprised the A.I.S. workforce would thus not reliably disclose the significance, if any, of sex to compliance. That being so, the Tribunal did not err in failing to select the groups of men and women comprising the workforce of A.I.S. as the base groups for the purpose of calculating the proportions of complying men and women.

Because the composition of the groups of men and women comprising the A.I.S. workforce was influenced by past discriminatory practices the significance, if any, of sex to compliance had to be ascertained by other means. Presumably a statistician could have ascertained that significance, if any, by adjusting or correcting for the effects of past discriminatory practices. But s. 24(3)(a) is premised on a less sophisticated exercise. It merely requires the selection of relevant base groups which do not themselves incorporate the effect of allegedly discriminatory practices and which can accordingly be used as reference points for ascertaining the effect of those practices. That requirement will not be satisfied if the definition of the selected base groups itself catches up and entrenches the effect of past discriminatory practices. Thus, in the present case where A.I.S.'s selected touchstone was that of gate seniority, the selected base groups will not be unaffected by sex if they are defined in a way which merely reflects the adverse effect upon the gate seniority of female employees of the discriminatory delay between application for employment and commencement of employment.

The approach adopted by the Tribunal was to select base groups within the A.I.S. workforce by reference to date of application for employment. It seems that the date selected was the date of application of the woman last employed on 6 January 1981. That was a permissible approach for it was open to the Tribunal to find that, in so far as seniority of some form or another was to determine whether or not a person was to be retrenched, the selection of that date would reveal those persons from amongst whom the persons to be retrenched or identified as likely to be retrenched would be selected. And when so revealed and grouped according to sex the resultant groups were unaffected by the past discriminatory practices of A.I.S. The selection of those groups allowed for the ascertainment of the significance, if any, of sex to compliance with the condition or requirement for continued employment. The course adopted by the Tribunal was thus permitted by s. 24(3)(a) of the Act. The result of that course was to disclose that the case fell within s. 24(3)(a) since the proportion of male employees able to comply with A.I.S.'s touchstone of gate seniority would, other things being equal, necessarily be higher than that of female employees whose gate seniority, vis-à-vis male employees who applied for employment at about the same time, was reduced by the effect of the past discriminatory practices.

It was further argued on behalf of the appellant that the "last on, first off" method of retrenchment was necessarily to be seen as reasonable, within the terms of s. 24(3)(b), because it resulted in men and women being retrenched in proportion to their numbers in the A.I.S. workforce. That argument assumes that the numerical composition of the workforce is the standard by reference to which reasonableness should be judged. Viewed solely from the perspective of discrimination, composition of the workforce is an appropriate standard only if the composition of the workforce is not, as in the present case, the result of discriminatory practices. However, s. 24(3)(b) is also directed to those considerations identified by the Supreme Court of the United States in Albemarle Paper Co. [12] , as "the employer's legitimate interest in "efficient and trustworthy workmanship" ". Perhaps in the Australian context those interests should be identified as extending to the maintenance of a stable workforce and one not subject to industrial disputation which otherwise might result if established patterns of industrial regulation and representation were put at risk. No attempt was, however, made to justify the "last on, first off" method's exacerbation of the adverse effects of past discriminatory practices as reasonable having regard to those interests. That being so, it was open to the Tribunal to find, as it did, that the condition or requirement was unreasonable in that it operated to keep alive the effects of past discrimination on the ground of sex and it was not established that it served the employer's legitimate interests.

This appeal raises the question whether Australian Iron & Steel Pty. Ltd. ("A.I.S.") discriminated against female ironworkers employed at its steelworks at Port Kembla when, because of a downturn in the steel industry, it was forced to retrench workers during a period extending from the end of 1982 into the first half of 1983. The retrenchments were made upon the principle of "last on, first off" or, as it has been called, the "reverse gate seniority" principle. The reason why it is said that retrenchment upon this basis discriminated against female workers is that for some years before 1980, A.I.S. had discriminated against females in the employment of ironworkers so that at the time of the retrenchment the proportion of females in its workforce at Port Kembla who had been recently employed was higher than it would have been if there had been no discrimination in the past. In other words, women ironworkers tended to be amongst the most junior, as a result of the prior discriminatory practice. The retrenchment programme was first announced on 11 October 1982, when it was said that those selected for retrenchment would be those whose last commencement date with the company was after 30 September 1980, the retrenchment programme to begin as from 14 November 1982. In the event, retrenchments stopped at those who commenced work before 6 January 1981. At about the time the retrenchments were to begin, the total number of ironworkers employed by A.I.S. was 8,208 of whom 7,698 were male and 510 were female. In all, 553 ironworkers were retrenched of whom 521 were male and 32 were female.

The Anti-Discrimination Act 1977 N.S.W. prohibits discrimination on the grounds of sex, race, marital status, physical impairment, intellectual impairment and homosexuality. Part III of the Act deals with sex discrimination. Section 25 makes it unlawful for an employer to discriminate against an employee on the ground of sex in such matters as the terms on which he offers employment and the provision of access to opportunities for promotion and transfer. The discriminatory dismissal of an employee is covered by s. 25(2) which, so far as is relevant, provides:

It is unlawful for an employer to discriminate against an employee on the ground of his sex -

(c) by dismissing him or subjecting him to any other detriment.

What amounts to sexual discrimination for the purposes of the Act is defined by s. 24:

(1) A person discriminates against another person on the ground of his sex if, on the ground of -

(b) a characteristic that appertains generally to persons of his sex; or

(c) a characteristic that is generally imputed to persons of his sex,

he treats him less favourably than in the same circumstances, or in circumstances which are not materially different, he treats or would treat a person of the opposite sex.

(3) A person discriminates against another person on the ground of his sex if he requires the other person to comply with a requirement or condition -

(a) with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply;

(b) which is not reasonable having regard to the circumstances of the case; and

(c) with which the other person does not or is not able to comply.

The New South Wales Act follows the method adopted in the Sex Discrimination Act 1975 U.K. where both direct discrimination and indirect discrimination are prohibited. In the New South Wales Act direct discrimination is prohibited by s. 24(1); indirect discrimination is prohibited by s. 24(3). A requirement may fall within the latter sub-section although it is "facially neutral", that is, although on its face it is neutral, in that it does not distinguish between the sexes or is not phrased in terms of a characteristic commonly associated with one sex.

The Equal Opportunity Tribunal of New South Wales found that A.I.S. had in its recruitment practice discriminated against 34 female complainants on the ground of sex. Its practice between the time the Anti-Discrimination Act came into force in June 1977 and April 1980 when complaints were first lodged was to employ men almost immediately upon application but to place women applicants on a waiting list for substantial periods of time, often amounting to several years. The finding in relation to each of the 34 complainants was that A.I.S. had discriminated against her by delaying hiring her. The tribunal also held that A.I.S. had, both in threatening to retrench ironworkers in accordance with the principle of "last on, first off" and in retrenching them on that basis, acted in breach of the Act. Damages were awarded for each breach, although damages for the threat of retrenchment and damages for actual retrenchment were treated as being in the alternative.

The tribunal's treatment of retrenchment, or the threat of retrenchment, as an act of discrimination in addition to the act of discrimination in delaying the hiring of a complainant, is important, because s. 113(b)(i) of the Act restricts the amount of damages to which any complainant is entitled for any single act of discrimination to $40,000. Eight of the 34 complainants were each awarded damages which in total amounted to more than $40,000 and it is against only those complainants that A.I.S. has appealed to this Court, recognizing that damages for retrenchment or the threat of retrenchment would have been an available head of consequential damage under the complaints based upon delay in hiring. In other words, A.I.S. has not sought to appeal where the complainant was able to rely upon direct discrimination in the form of delay in hiring, but only where, in order to support an award of damages in excess of $40,000, she had to rely upon retrenchment or the threat of retrenchment as a separate act of discrimination, direct or indirect. And in the case of the eight complainants against whom A.I.S. appeals, it does not seek to disturb the finding of the tribunal that it discriminated against them by reason of its delay in hiring them.

The question is, therefore, whether, having discriminated against them by delaying their employment, A.I.S. further discriminated against the first eight respondents to this appeal, within the meaning of s. 24(1) or s. 24(3), by retrenching them upon a "last on, first off" or "reverse gate seniority" principle. The question whether there was discrimination within the meaning of s. 24(1) may, I think, be disposed of shortly.

Were it not for s. 24(3), it may have been possible, in order to give effect to the intention of the Anti-Discrimination Act, to have given s. 24(1) a broad construction embracing indirect discrimination. Nevertheless some difficulty would have been encountered because the sub-section expressly extends the grounds upon which a person may discriminate upon the ground of sex to include, in addition to sex itself, a characteristic that appertains generally to persons of the complainant's sex or a characteristic that is generally imputed to persons of the complainant's sex. In other words, the sub-section says that a person will discriminate on the ground of sex if he discriminates, not only on the express ground of sex, but also upon other specified grounds. It is hard to avoid the conclusion that the extension of the grounds upon which sexual discrimination may take place, to include matters other than sex itself, marks the limits of the sub-section. Be that as it may, when s. 24(1) is read with s. 24(3), it is clear to my mind that s. 24(1) cannot be read so as to cover indirect discrimination, for to do so would be to render s. 24(3) superfluous. That being so, discrimination will not amount to discrimination on the ground of sex within the meaning of s. 24(1) merely because it is upon a ground which produces an effect equivalent to the effect of discrimination upon the ground of sex, unless the ground upon which the discrimination is based is a characteristic that appertains generally to persons of the complainant's sex or is a characteristic that is generally imputed to persons of the complainant's sex.

I do not think that this construction of s. 24(1) requires it to be applied subjectively. The mere assertion of a ground which is not sex will not take discrimination out of the sub-section if the true basis for the action in question is in fact sex. Thus, in the present case if it could have been shown that the "last on, first off" principle was selected as the ground for retrenchment, not as an equitable means of shedding redundant employees, but as a means of shedding female employees more quickly, s. 24(1) would have applied. The true ground would then have been sex and any discrimination would have been on that ground. But that was not shown and in my view s. 24(1) had no application. Cf. James v. Eastleigh Borough Council [13] .

1. [1989] 3 W.L.R. 123, at p. 128; [1989] 2 All E.R. 914, at pp. 918-919; [1990] Q.B., at p. 74.

Section 24(3), which defines indirect discrimination, has a much wider application and covers discrimination which is revealed by the different impact upon the sexes of a requirement or condition. The starting point with s. 24(3) must be the identification of the requirement or condition. Upon principle and having regard to the objects of the Act, it is clear that the words "requirement or condition" should be construed broadly so as to cover any form of qualification or prerequisite demanded by an employer of his employees: Clarke v. Eley (I.M.I.) Kynoch Ltd. [14] . Nevertheless, it is necessary in each particular instance to formulate the actual requirement or condition with some precision.

It is not, I think, enough in the present case simply to see the requirement or condition for continued employment as being contained within the principle of "last on, first off". That principle was applied within defined limits and it is necessary to incorporate them in the requirement. Thus, it was accepted upon both sides in this Court that the requirement was that, for an ironworker to remain in employment once retrenchments had begun in November 1982, he or she must have commenced employment before 6 January 1981.

The requirement having been identified, s. 24(3) then demands that a comparison be made between the compliance rates for each sex, in order to determine whether "a substantially higher proportion of persons of the opposite sex to the sex of the [complainant] comply or are able to comply". There is more than one possible method of making the comparison.

One method is to make a bold comparison, within the aggregate workforce of A.I.S. at the time of the retrenchments, between the raw figures for the number of men who complied and the raw figures for the number of women who complied. Given that 7,177 men were employed before 6 January 1981 while only 478 women had commenced employment before that date, it is self-evident upon that comparison that the requirement was satisfied by many more men than women.

The problem with that form of comparison is that the result may merely be a reflection of the fact that the workforce was sexually imbalanced. Indeed, where the sexes are not evenly balanced in a workforce, the application of the "last on, first off" principle will almost always result in the retrenchment of a higher proportion of one sex. Where, as in this case, the men employed outnumbered the women by a ratio of fifteen to one, it was only to be expected that the number of men who complied with any condition, however genuinely neutral and non-discriminatory, would greatly outnumber the number of women who could comply. Upon this approach, the fact that the sexes in a workforce are unequal is itself a significant factor in determining whether a requirement imposed upon that workforce amounts to discrimination within the meaning of s. 24(3), regardless of whether or not that inequality is the result of a prior discriminatory practice. Such an approach could only be justified by treating s. 24(3) as being aimed generally at discouraging workforces in which the sexes are unequally represented and there is, in my view, no basis for interpreting the sub-section in such a far-reaching manner. Obviously, the reach of the sub-section was intended to be far less ambitious and to extend only to discriminatory requirements or conditions imposed upon a workforce, whether the sexes in the workforce happen to be unequal or not. The sub-section was not intended to embrace requirements which are truly non-discriminatory and it must, therefore, require something more than a direct comparison between the number of men who comply and the number of women who comply with a requirement imposed by an employer.

The English legislation makes it clear that what is to be compared for the purpose of establishing that a requirement is discriminatory are not the raw figures of those of each sex who comply but the proportion of men who comply and the proportion of women who comply. Each proportion upon this method is made up of only one sex so that any difference in the compliance rates between the two sexes is revealed. Section 1(1)(b) of the English Act specifically refers to the need to compare two proportions. It provides:

A person discriminates against a woman in any circumstances relevant for the purposes of any provision of this Act if -

(b) he applies to her a requirement or condition which he applies or would apply equally to a man but -

(i) which is such that the proportion of women who can comply with it is considerably smaller than the proportion of men who can comply with it

Although the New South Wales Act refers to a proportion only once when it speaks of the need to determine whether a "substantially higher proportion of persons of the opposite sex to the sex of the [complainant] comply or are able to comply", it seems clear that the most appropriate way to test the incidence of discrimination between the sexes is to compare the relevant proportion of men with the relevant proportion of women. Having regard to the similarity between the two pieces of legislation, it seems unlikely that it was intended that the New South Wales Act should require the application of a different test. Moreover, a comparison upon a proper basis between the two proportions has the effect that inequality of the sexes in the workforce will not of itself determine whether the compliance rate by one sex is substantially higher than that of the other and will not, therefore, pre-empt the question whether a requirement which has been imposed upon a workforce is discriminatory.

But a proportion must be a proportion of something, so that it is necessary to determine the appropriate grouping or pool within which to calculate the proportions which are to be compared. The English cases have discussed in detail the problems associated with the determination of an appropriate base group. Two main contrasting approaches can be identified. One seeks to narrow the base group to the particular group of persons to whom the requirement is directed, while the other seeks to establish a broader base beyond the immediate context. The need to select an appropriate base group, which may be limited to the immediate context, was recognized in Kidd v. D.R.G. (U.K.) Ltd. [15] , where Waite J., in delivering the decision of the Employment Appeal Tribunal, held that:

for the purposes of the statutory comparison, like must be compared with like in a context appropriate to the case under review. The consequence is that the particular section of the members of the public upon whose lives the impact of the relevant requirement or condition has to be measured is liable to vary from case to case - ranging from the population as a whole at one end of the scale to employees of a single workplace at the other: and there is liable also to be ample scope for debate in many instances as to which section of the public within that range is the right one to choose for a particular case. 1. [1985] I.C.R. 405, at p. 409.

As that passage indicates, the contextual approach will provide different answers depending upon the circumstances of each case. For example, where a requirement is contained in a published offer of employment, the relevant base group may be made up of those who might be expected to be eligible to take up the employment based upon geographical, educational and other restraints. Similarly, where a requirement is imposed upon existing employees, the relevant group may be the class of employees affected.

In this case, the appellant submitted that the appropriate base group, being that to which the requirement in question was directed, was the aggregate workforce of A.I.S. That base group, so it was submitted, had to be divided into the two sexes to enable the proportions of each sex affected to be calculated and compared. Accordingly, the appellant argued that the relevant proportions to be compared for the purposes of s. 24(3) were the actual number of men who complied with the requirement divided by the number of men to whom the requirement was directed, as against the actual number of women who complied with the requirement divided by the number of women to whom the requirement was directed. On this basis the appellant argued that 93.23 per cent (7,177/7,698) of all male ironworkers in the workforce were not retrenched and 93.73 per cent (478/510) of all female ironworkers in the workforce were not retrenched, so that it could not be said that a substantially higher proportion of men complied or were able to comply with the requirement in question. However, for reasons which will appear later, to adopt the aggregate workforce as the base group in this manner would be to fail to take into account the continuing discriminatory effect of past recruitment practice upon the retrenchment programme.

The broader approach to the determination of a base group was adopted at first instance in Price v. Civil Service Commission [16] . In that case an industrial tribunal dismissed a complaint by a female civil servant that she had been discriminated against by the imposition of an age requirement as an eligibility condition for a job as an executive officer. The complainant submitted that the age requirement imposed (between 17 and 28 years) was one with which fewer women could comply than men because many women between those ages were occupied looking after children. The industrial tribunal adopted as the appropriate base group the whole population of England, divided into males and females, and held that the proportion of women within the specified age group was not considerably smaller than the proportion of men. In allowing the appeal, the Employment Appeal Tribunal considered that the base group might better have been restricted to qualified men and women of the relevant ages. A similar broad approach seems to have been adopted in Steel v. Union of Post Office Workers [17] .

1. [1977] 1 W.L.R. 1417; [1978] 1 All E.R. 1228. 2. [1978] 1 W.L.R. 64; [1978] 2 All E.R. 504.

Adopting the broad approach in this case, it would be possible to take the whole of the population of, say, New South Wales, divided into males and females. Upon such an approach, although the proportions would be small, because those who could comply would necessarily be persons within the A.I.S. workforce, it would be evident that a greater proportion of men could comply with the condition than women, because the division of the sexes in New South Wales is approximately equal and the number of men and women who complied with the condition were unequal, by far the greater number being men.

However, to apply the sub-section in this way may lead to the unacceptable result that a requirement is sexually discriminatory when clearly it is not. For example, if an employer seeks to retrench only red-headed employees, the relevant requirement for continued employment, namely, that a person have hair of a colour other than red, is clearly not grounded upon sex. But, to continue the example, if the workforce consists of 5,000 men and 500 women and 4,500 men and 450 women comply with the requirement, and if the aggregate workforce, divided into the respective sexes, is the base group, it is apparent that the same proportions, 10 per cent of males and 10 per cent of females, are able to comply. However, if the entire population of New South Wales were adopted as the base group, the sexes being approximately equally divided, 4,500 over the New South Wales male population yields a substantially higher proportion of men who can comply with the requirement than 450 women over the female population, and s. 24(3)(a) is satisfied. Moreover, the requirement, being arbitrary, would be likely to be unreasonable and thus satisfy s. 24(3)(b). And it is easy to imagine a red-headed female complainant who would not be able to comply with the requirement so as to fall within s. 24(3)(c). Thus imposing a requirement which may be unfair, but is not sexually discriminatory, would be held to amount to sexual discrimination. The sexually unequal nature of the workforce would demand the conclusion that the more numerous sex in the workforce had been favoured.

I have already adverted to the problem that in selecting an appropriate base group in this case, the group should not be defined so as to mask the effect of the previous discriminatory recruitment practice. Failure to make some adjustment for this would be to avoid the very sort of wrong which the Anti-Discrimination Act was intended to redress. A problem of this nature was considered in Reg. v. Secretary of State for Education; Ex parte Schaffter [18] . In that case, the requirements for eligibility for hardship study grants for lone parents included a requirement that the parent must once have been married. If the base group was made up of single lone parents divided into sexes, the result was that about 20 per cent of lone parents of both sexes had never been married. However, the figures also showed that there were about four times as many female lone parents who failed to comply with the requirement as there were male lone parents who failed to comply. Schiemann J. [19] , rejected as the base group all lone parents in the student population because:

if one reduces the size of the pool under consideration to a very small size there is a very real risk that you have incorporated an act of discrimination into your definition. If I were to adopt [the respondent's] submission in this case and only look at single lone parents, and then compare the proportion of female lone parents which is single with the proportion of male lone parents which is single, I would indeed find no significant difference, but I would have fallen into precisely that trap. In that case, Schiemann J. appears to have assumed that many more women than men were affected by the requirement as the result of a sexist practice in the community whereby unmarried mothers rather than unmarried fathers typically carried the responsibility of caring for their children. To make an adjustment for this practice, he selected as the appropriate base group, a pool comprising all students with dependent children claiming grants, in which a much greater proportion of men than women complied with the condition of once having been married. 1. [1987] I.R.L.R. 53. 2. [1987] I.R.L.R., at p. 56.

One method of making an appropriate adjustment, which in my view is satisfactory in this case, is that submitted by the respondents and adopted by the Equal Opportunity Tribunal, namely, taking a base group consisting of those employees who applied for employment at about the same time as one of the complainants, Mrs. Donka Najdovska. She was treated before the tribunal as the principal complainant. Mrs. Najdovska applied for employment in August 1977, but was not employed until 26 February 1981. It is possible to speak of those who applied for employment, men and women, about August 1977, because, provided a period is selected which throws up a sufficient number of persons, the precise length in this case is not significant. The period must however be sufficiently short to enable it to be said that during that period the women applying for employment had not yet been subjected to discriminatory treatment. In that way a group of men and women who were treated equally by A.I.S. can be identified, because the discriminatory effect of men being taken on almost immediately and women being placed on a waiting list had not then appeared. In selecting the base group in that way that element of discrimination is not built in. It is eliminated because the base group which is selected is in fact a sample of the workforce which is not subject to the bias which would arise, as a result of the discriminatory recruitment practice, if the whole workforce were taken as the base group. The selection of an appropriate sample is a common and statistically acceptable means of calculating the proportions of persons in a larger community who answer a particular description. Of course, to select a group in this way is not to impose an additional requirement of eligibility for continued employment, but rather to choose a pool whose members do not have a characteristic which could bias the result.

Having selected an unbiased sample as the base group, it is then divided into males and females, and the proportions of those males and those females who comply with the requirement that they should have commenced employment before 6 January 1981 can be calculated. The proportion of males would significantly exceed the proportion of females because of the men who applied for employment at about the same time as Mrs. Najdovska in August 1977 (whatever the precise period selected), practically all of them, on the evidence, would have been employed almost immediately upon application. By way of contrast, of the women who applied for employment about August 1977, it is clear that few of them, including Mrs. Najdovska herself, were able to comply with the relevant requirement, it being accepted on both sides that the delay in hiring for women was very considerable. Thus, after selecting a base group by way of an unbiased sample in order to eliminate the effect of the prior discriminatory practice, the requirement that employees should have commenced employment before 6 January 1981 is revealed to be one with which a substantially higher proportion of men than women could comply. It was not argued that, notwithstanding that conclusion, the requirement could nevertheless be justified as being reasonable having regard to the circumstances of the case and it follows that retrenchment upon the basis of "last on, first off" constituted, in this case, discrimination on the ground of sex within the meaning of s. 24(3) of the Anti-Discrimination Act.

I should add that the principle of "last on, first off" is not inherently unfair or unreasonable. All things being equal, it may be a common sense way of selecting employees for retrenchment in the unfortunate event of that being necessary: Clarke v. Eley (I.M.I.) Kynoch Ltd. [20] . The requirement imposed in this case was unreasonable because in the particular circumstances it repeated the discriminatory effect of the prior recruitment practice.

The question in this appeal is whether Australian Iron & Steel Pty. Ltd. ("A.I.S.") discriminated against eight women ironworkers ("the respondents") when it retrenched them on the basis of a "last on, first off" retrenchment policy in circumstances where 6. 27 per cent of female ironworkers employed by A.I.S. were retrenched as compared with 6. 77 per cent of male ironworkers. The Equal Opportunity Tribunal of New South Wales ("the Tribunal") held that the retrenchment of the respondents was in breach of s. 25 of the Anti-Discrimination Act 1977 N.S.W. ("the Act") and awarded them damages. An appeal - which the Act limits to errors of law - against that decision was unanimously dismissed by the New South Wales Court of Appeal.

Between June 1977 (the date of the entry into operation of the Act) and April 1980 (the date of lodgment of the initial complaints of discrimination), A.I.S. filled 4,289 vacancies at its Port Kembla Steelworks. The Tribunal found, and on appeal it has not been disputed, that A.I.S. discriminated against women in filling these vacancies. Only 1.35 per cent of them were filled by women.

When a downturn in the steel industry occurred towards the end of September 1981, A.I.S. stopped employing new ironworkers. On 11 October 1982, A.I.S. announced its intention to reduce the workforce at its Port Kembla works. It announced that retrenchment notices were to be given to 400 workers on 14 November 1982 and that a further 400 workers would probably be retrenched at a later date. Retrenchments were to proceed in accordance with a "last on, first off" policy. When the notice of retrenchments was announced it was expected that the workers selected for retrenchment would be those who commenced employment with the company after 30 September 1980. The workers who were ultimately retrenched, however, were those who commenced work on and after 6 January 1981. As at 12 November 1982, A.I.S. employed 8,208 ironworkers. Of these workers, 7,698 were men and 510 were women. A total of 553 ironworkers - 521 men and 32 women - were retrenched.

Fifty-five sex discrimination complaints by 34 female ironworkers were ultimately made against A.I.S. But one ironworker, Donka Najdovska, was treated in the Tribunal as the principal complainant. On 13 October 1980 she lodged an initial complaint with the Counsellor for Equal Opportunity. She alleged that she had applied for work with A.I.S. on 22 August 1977, but A.I.S. had refused to employ her on the ground of her sex. However, A.I.S. eventually gave her employment. She commenced work at the steelworks on 26 February 1981. She was the last of the original complainants to be given employment by A.I.S. On 9 December 1982, she lodged a second complaint. She claimed that, in announcing its intention to retrench ironworkers on the basis of length of service, A.I.S. had discriminated against her on the ground of her sex. On 23 June 1983, she lodged a third complaint claiming that her employment had been terminated on 20 January 1983 and that she had been dismissed on the ground of sex. On this occasion, the complaint on behalf of Ms. Najdovska was also made on behalf of all other women retrenched by A.I.S. since January 1980 and all women likely to be retrenched in the future by reason of their length of service with A.I.S. In February 1984 the complaints of Ms. Najdovska were referred to the Equal Opportunity Tribunal. In May 1984 the complaints of the remaining 33 complainants were referred to the Tribunal. For reasons of convenience, the Tribunal characterised the complaints as falling into three categories: (1) failure to hire; (2) threat of retrenchment; and (3) retrenchment.

In September 1985 the Tribunal gave a decision upholding all complaints. In October 1986 the Tribunal made awards of damages. In assessing the damages of each complainant, individual calculations were made, where relevant, in respect of delay in hiring, threatened retrenchment, and actual retrenchment. Only a small sum - not more than $500 - was awarded to each complainant for the threat of retrenchment. Section 113(b)(i) of the Act, however, imposes a statutory limit of $40,000 for damages payable in respect of any one claim of discrimination. For those complainants unable to establish an entitlement to an aggregate sum in excess of $40,000, it was unnecessary to consider whether delay in hiring, threat of retrenchment, and actual retrenchment constituted independent acts of discrimination. As Street C.J. observed in the Court of Appeal, the damages for these workers for retrenchment and the threat of retrenchment could be causally linked to the delay in hiring them: Australian Iron & Steel Pty. Ltd. v. Najdovska [21] . Hence, the initial unlawful conduct of delay in hiring was an adequate basis to support all damage sustained up to a limit of $40,000. But for those complainants who could establish damage in excess of $40,000, it was necessary, if they were to achieve full compensation, that delay in hiring and threatened and actual retrenchment should constitute independent unlawful acts of discrimination, capable of founding separate claims for damages. It was for this reason that the Tribunal made findings that the retrenchments and threats of retrenchment as well as the delay in hiring were unlawful acts of discrimination.

In the Court of Appeal, A.I.S. did not seek to challenge the Tribunal's finding of sex discrimination based on the delay of A.I.S. in hiring the women ironworkers: Australian Iron & Steel Pty. Ltd. v. Najdovska [22] . However, A.I.S. did challenge the finding that it had discriminated against female ironworkers in announcing its plans to retrench and in retrenching them. Hence, by the time the matter reached this Court, only the finding of sex discrimination based on the threat of retrenchment and actual retrenchment was in issue.

The relevant provisions of the Act, for present purposes, are ss. 24 and 25 which provide:

(1) A person discriminates against another person on the ground of his sex if, on the ground of -

(b) a characteristic that appertains generally to persons of his sex; or

(c) a characteristic that is generally imputed to persons of his sex,

he treats him less favourably than in the same circumstances, or in circumstances which are not materially different, he treats or would treat a person of the opposite sex.

(3) A person discriminates against another person on the ground of his sex if he requires the other person to comply with a requirement or condition -

(a) with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply;

(b) which is not reasonable having regard to the circumstances of the case; and

(c) with which the other person does not or is not able to comply.

(2) It is unlawful for an employer to discriminate against an employee on the ground of his sex -

(c) by dismissing him or subjecting him to any other detriment.

The Court of Appeal unanimously held that the Tribunal erred in finding that both the threats of retrenchment and the retrenchments of the respondents contravened s. 24(1) of the Act. Street C.J. said that the postulation of less favourable treatment required comparison with men "in the same circumstances" and that there were no men "in the same circumstances" [23] . Priestley J.A., with whose judgment Mahoney J.A. agreed, said that the language of s. 24(1) was not apt "even on the broadest approach, to describe the retrenchment facts" [24] .

1. (1988) 12 N.S.W.L.R., at p. 594. 2. (1988) 12 N.S.W.L.R., at p. 608.

However, all members of the Court thought that it was open to the Tribunal to find that the complainants had proved that the threats of retrenchment and the retrenchments satisfied the terms of s. 24(3). They rejected the contention of A.I.S. that the "last on, first off" policy was not discriminatory because the women ironworkers were not required to comply with a requirement or condition "with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply". A.I.S. had submitted that the requirement or condition was not discriminatory because both the number and the proportion of the total number of male ironworkers retrenched were greater than the respective number and proportion of female ironworkers retrenched.

It is common ground that, by retrenching the female ironworkers, A.I.S. required them "to comply with a requirement or condition" within the meaning of s. 24(3). The condition was formulated in various ways, but in general terms the parties accepted that A.I.S. had required of its employees as a condition of continued employment after retrenchments commenced that he or she had commenced employment with the company before 30 September 1980 or, alternatively, 6 January 1981. Nevertheless, the proposition that a retrenchment criterion of "last on, first off" is a requirement or condition of the kind referred to in s. 24(3) is at first sight difficult to accept. However, s. 24(3) of the Act is based on the provisions of s. 1(1)(b) of the Sex Discrimination Act 1975 U.K.. Courts and tribunals interpreting the phrase "requirement or condition" in that legislation have generally given it a broad interpretation in order to implement the objectives of the legislation. In Clarke v. Eley (I.M.I.) Kynoch Ltd. [25] , the Employment Appeal Tribunal held that, where in a redundancy situation an employer retrenched workers on the basis that part-time workers should go before full-time workers but subject thereto selection should be on the basis "last in, first out", the employer was applying a "requirement or condition". The Employment Appeal Tribunal rejected the submission that "requirement" meant something which calls for or demands something of the person to whom it is applied and that "condition", while including a qualification for holding a position, did not include a qualification for immunity from a disadvantage. Browne-Wilkinson J., giving the judgment of the Tribunal, said [26] :

In our view it is not right to give these words a narrow construction. The purpose of the legislature in introducing the concept of indirect discrimination into the Act of 1975 and the Race Relations Act 1976 was to seek to eliminate those practices which had a disproportionate impact on women or ethnic minorities and were not justifiable for other reasons. The concept was derived from that developed in the law of the United States which held to be unlawful practices which had a disproportionate impact on black workers as opposed to white workers: see Griggs v. Duke Power Co. [27] . If the elimination of such practices is the policy lying behind the Act, although such policy cannot be used to give the words any wider meaning than they naturally bear it is in our view a powerful argument against giving the words a narrower meaning thereby excluding cases which fall within the mischief which the Act was meant to deal with. In Watches of Switzerland Ltd. v. Savell [28] , the Employment Appeal Tribunal said that it agreed with earlier statements "that the statutory words should be given a liberal interpretation in order to implement the object of the legislation". In Home Office v. Holmes [29] , the Employment Appeal Tribunal held that an obligation to work full-time was a requirement or condition of employment within the meaning of the Act. Waite J., delivering the judgment of the Tribunal, said [30] : It appears to us that words like "requirement" and "condition" are plain, clear words of wide import fully capable of including any obligation of service whether for full-time or for part-time, and we see no basis for giving them a restrictive interpretation in the light of the policy underlying the Act InStyles v. Secretary of Department of Foreign Affairs and Trade [31] , Wilcox J., after referring to these cases and the decision of the Court of Appeal in the present case, said: As Browne-Wilkinson J. pointed out, the purpose of a provision such as s. 5(2) is to eliminate practices which amount to indirect discrimination, practices which may at first sight have nothing to do with sex discrimination but which may impose a disadvantage upon persons of one sex which is disproportionate to its incidence upon the other sex. The forms of such disadvantage being almost infinitely various, the court should look at situations realistically, assessing whether, in a particular case, there was insistence upon a particular requirement. His Honour's decision on this point was upheld by the Full Court of the Federal Court. Bowen C.J. and Gummow J. said, in the context of that case - an appointment case - that a requirement or condition meant "a stipulation which must be satisfied if there is to be a practical (and not merely a theoretical) chance of selection" [32] . 1. [1983] I.C.R. 165. 2. [1983] I.C.R., at p. 171. 3. (1971) 401 U.S. 424. 4. [1983] I.R.L.R. 141, at p. 146. 5. [1984] I.C.R. 678. 6. [1984] I.C.R., at p. 683. 7. (1988) 84 A.L.R. 408, at pp. 423-424. 8. (1989) 88 A.L.R. 621, at p. 629.

In my opinion, this course of authority justified the parties in accepting that A.I.S. had required each of its employees to comply with the condition that to remain in employment after retrenchments began he or she must have commenced work before 6 January 1981.

Proportions of Ironworkers who Comply or are Able to Comply

The substantial question argued in the present appeal, however, was whether "a substantially higher proportion of persons of the opposite sex" to the female ironworkers complied or were able to comply with the condition which I have identified. English courts and tribunals have examined - correctly in my opinion - the practical ability of complainants to comply with requirements or conditions and have disregarded purely theoretical possibilities. In Price v. Civil Service Commission [33] the Employment Appeal Tribunal said that it is: "relevant in determining whether women can comply with the condition to take into account the current usual behaviour of women in this respect, as observed in practice, putting on one side behaviour and responses which are unusual or extreme." Moreover, the English cases have tested ability to comply by reference to the time at which the requirement or condition was applied and not by considering whether the complainant at some time in the past could, or might at some time in the future be able to, comply with it. In Clarke v. Eley (I.M.I.) Kynoch Ltd. [34] , where a part-time worker complained that a redundancy scheme was discriminatory in selecting part-time workers before full-time workers for retrenchment, the Employment Appeal Tribunal held that it was irrelevant that the complainant could have transferred in the past to a full-time position. The "only material question" [35] was whether or not at the date of dismissal she could comply with the requirement.

1. [1977] 1 W.L.R. 1417, at p. 1421; [1978] 1 All E.R. 1228, at p. 1231. 2. [1983] I.C.R. 165. 3. [1983] I.C.R., at p. 172.

The present case, however, gives rise to no difficulty in determining which persons complied or were able to comply with the condition: they were all the ironworkers employed by A.I.S. at Port Kembla at the commencement of the retrenchments who had commenced employment before 6 January 1981.

The Tribunal, Street C.J., and Mahoney and Priestley JJ.A. had differing opinions, however, as to which was or were the appropriate pool or pools of persons to be identified for the purpose of comparing the respective proportions of the sexes who could comply with the requirement or condition. The Tribunal confined the pools to the males and females who applied for employment at the same time as the complainants; Street C.J. would have confined the pools to the male and female ironworkers in the A.I.S. workforce; Mahoney and Priestley JJ.A. would have selected the male and female populations of New South Wales as the appropriate pools.

English courts and tribunals have also experienced difficulty in selecting the appropriate pools of persons for the purpose of comparison: Leigh, Of Racial Groups and Non-Groups, Modern Law Review vol. 49 (1986) 235, at p. 239 . The problem was first alluded to by Phillips J. in Price where, speaking on behalf of the Employment Appeal Tribunal, his Lordship expressed doubt [36] whether the Industrial Tribunal had adopted the right approach in taking into account the whole population, male and female respectively, rather than confining the relevant pools to qualified men and qualified women. The genesis of the problem was explained by Waite J., giving the judgment of the Employment Appeal Tribunal, in Kidd v. D.R.G. (U.K.) Ltd. [37] :

The concept of indirect discrimination was one which clearly needed to be framed from the outset with the maximum flexibility if it was fully to encompass the mischief at which the anti-discrimination laws are directed. Flexibility has been achieved, however, at some cost in loss of precision of language. Then there is the further requirement of the Act that for the purposes of the statutory comparison, like must be compared with like in a context appropriate to the case under review. The consequence is that the particular section of the members of the public upon whose lives the impact of the relevant requirement or condition has to be measured is liable to vary from case to case - ranging from the population as a whole at one end of the scale to employees of a single work-place at the other: and there is liable also to be ample scope for debate in many instances as to which section of the public within that range is the right one to choose for a particular case. 1. [1977] 1 W.L.R., at p. 1422; [1978] 1 All E.R., at p. 1232. 2. [1985] I.C.R. 405, at p. 409.

As Mr. Robb, counsel for A.I.S., pointed out, where compliance with a requirement or condition depends upon an attribute which is a characteristic of persons generally, it may be appropriate to consider the capability of all of the members of one sex in the population to comply with the requirement or condition as compared with the capability of all members of the other sex to comply with that requirement or condition. But many requirements or conditions of employment are based on attributes which are restricted to groups and are not shared by the population at large. Mr. Robb was surely right when he submitted that in the case of such requirements or conditions it is meaningless to compare the proportions required by s. 24(3)(a) by reference to the general population. As he pointed out, the inclusion of members of the population who do not have such an attribute is likely to mask the consequences of the application of the requirement or condition to those required to comply with it. If an employer in a remote town in New South Wales advertises for a shop assistant with five years' experience and there are no female but some male shop assistants in that town with that length of experience, it surely cannot be a complete answer to a charge under s. 24(3)(a) of discrimination against a woman shop assistant that New South Wales as a whole has more women than men shop assistants with five years' experience. Whether more men than women comply with the requirement must depend on who are the likely applicants for the advertised position. Moreover, the employer should not be liable at the suit of a woman shop assistant if the town is near the State border and the pool of likely applicants from both sides of the border contains more qualified women than men even if State-wide there are more qualified men than women shop assistants. In such a case it would be male shop assistants who were potentially discriminated against under s. 24(3).

For many occupations, the pool of likely applicants for employment knows no local, regional, State or national boundaries. There is no reason to think that the Parliament of New South Wales intended that the Tribunal should impose artificial State boundaries on a pool of applicants who might be drawn wholly or partly from outside the State. No doubt the act of discrimination must take part in New South Wales to come within the scope of the Act. But that proposition is not inconsistent with the Tribunal, when applying s. 24(3), having regard, in an appropriate case, to persons some or all of whom are drawn from outside the State.

But how is the comparison between proportions to be made? Opinions have differed on the question whether the comparison which the paragraph requires is that of proportions expressed as numbers or proportions expressed as percentages of each sex who comply or are able to comply with the requirement or condition. Those who assert that the comparison is between proportions expressed as numbers treat the total number of those who comply or are able to comply with the requirement or condition as the denominator of each of two fractions with the number of males who comply or are able to comply as the numerator of the first fraction and the number of females who comply or are able to comply as the numerator of the second fraction. Those who assert that the comparison is between proportions expressed as percentages of the respective sexes compare two fractions in which the numerator and denominator are units of each sex. The numerator of each fraction is the number of members of the sex who comply or are able to comply with the requirement or condition; the denominator is the number of members of that sex to whom the requirement or condition is directed. The numerator is then expressed as a percentage of the denominator.

In the present case, Mahoney and Priestley JJ.A. accepted that the numerator and denominator in each fraction must be units of the same sex. But their Honours thought that the correct denominators in each case were the respective female and male populations in New South Wales. Hence, Mahoney and Priestley JJ.A. favoured the percentage approach. Their Honours said "the proportions of the whole male and the whole female populations of New South Wales would be quite small, but in comparison to one another the proportion of males would have to be substantially higher than that of females (very roughly fifteen to one)" [38] . Street C.J., on the other hand, took the view that the comparison was between the number of male ironworkers who avoided retrenchment and the number of female ironworkers who avoided retrenchment. Since the ratio of men to women who avoided retrenchment was 15:1, his Honour held that the terms of s. 24(3)(a) were satisfied. Thus, although using different methods, the members of the Court reached almost identical results. What these results show, however, is that when the male and female populations of a jurisdiction are roughly equal, the approach of Mahoney and Priestley JJ.A. is in reality, though not in form, a numbers approach.

In Styles v. Secretary of Department of Foreign Affairs and Trade, Wilcox J., at first instance, and Bowen C.J. and Gummow J., on appeal, also adopted the numerical comparison approach. Bowen C.J. and Gummow J. said [39] :

Within the universe of discourse created by the terms of the sub-section, one looks at all those who comply or are able to comply and asks whether of that total those of the opposite sex to the aggrieved person may be described as a substantially higher proportion than those of the sex of the aggrieved person. Applying this construction in the present case, the total of those persons equivalent to Grade A2 and above who comply or are able to comply is 20; the proportion of persons of the opposite (male) sex to the aggrieved person who comply or are able to comply is 17/20 (or 85 per cent); and the proportion of persons of the (female) sex of the aggrieved person who comply or are able to comply is 3/20 (or 15 per cent). The former proportion is substantially higher than the latter proportion. On their Honours' approach one compares two fractions in which the denominator of each is the total of men and women who comply or are able to comply with the requirement or condition and the numerators of the respective fractions are the respective numbers of men and women who comply or are able to comply with the requirement or condition. In truth this is simply to compare the number of one sex who comply or are able to comply with the requirement or condition with the number of the opposite sex who comply or are able to comply with the requirement or condition. Pincus J., the other member of the Full Court, rejected the numerical comparison approach which Bowen C.J. and Gummow J. applied. His Honour thought that Mahoney and Priestley JJ.A. were correct in the present case in holding "that there must be two separate denominators - one for each of the two proportions compared". But Pincus J. also said "that the view of Street C.J. in the A.I.S. Case [40] is correct: "the base group is to be constituted by the persons to whom the requirement or condition was directed and who are divided or segregated by it", with the reservation that I would take two separate base groups or pools". 1. (1989) 88 A.L.R., at p. 630. 2. (1988) 12 N.S.W.L.R., at p. 596.

With great respect to those who have taken the contrary view, I am of the opinion that Pincus J. was correct in holding that there must be two base groups or denominators and that one group is constituted by the males, and the second group by the females, to whom the requirement or condition is directed and who are divided or segregated by it. Although s. 24(3)(a) mentions the "proportion of persons" only once, the natural meaning of the paragraph is that the comparison is between the proportions of each sex who comply or are able to comply with the requirement or condition. Hence, the comparison is between the proportions of the sexes, not the numbers of each sex who comply or are able to comply with the requirement or condition. But, if proportions are to mean something more than numbers, the only way the proportions of each sex can sensibly be compared is by comparing fractions in which the units contained in the numerator and denominator of each fraction refer to members of the male and female sexes respectively. On any view of the paragraph, its language requires that the numerators of the fractions must be the numbers of the respective sexes who comply or are able to comply with the requirement or condition. Otherwise it would be impossible to determine the "proportion of persons" of each sex who comply or are able to comply with the requirement or condition. And it follows from the need for the units in each fraction to refer to members of the same sex that the determination of the proportion also requires that the denominator in each fraction be the number of persons of that sex who are required to comply with the requirement or condition and who are divided or segregated by it.

Moreover, the construction which I have placed on s. 24(3)(a) is the one which exposes facially neutral but nevertheless discriminatory practices. It is a construction, therefore, which promotes the very object of s. 24(3), viz., the elimination of indirect discrimination. The construction which Street C.J., Wilcox J., Bowen C.J. and Gummow J. have placed on s. 24(3)(a), however, leads to the result that the conditions of the paragraph are never satisfied when the members of a complainant's sex, who are able to comply with the requirement or condition, are more numerous than the members of the opposite sex. Such a construction, therefore, fails to expose many facially neutral but discriminatory conditions. Mr. Robb gave an illuminating example. Suppose the Police Department employs 1,000 women and 200 men as clerk/typists and a condition of promotion is experience as a clerk/typist in a country police station. If 100 women and 60 men have the relevant experience, then on the approach favoured by Street C.J. in the present case and by various members of the Federal Court, a woman clerk/typist who complained of indirect discrimination would fail. Yet the condition is one with which a substantially higher proportion of men (30 per cent) than women (10 per cent) can comply.

In the present case, however, it is not clear what method of comparison the Tribunal used. The Tribunal said:

Given what Mr. Basten calls the "staggering" increase in the number of women hired from June, 1980 to February, 1981, as a percentage of the workforce, the complainants say that when the company imposed the requirement or condition on each employee that it would retrench on the basis of "last on, first off" it was clearly one with which significantly less women than men could comply. It was one with which the complainants who were retrenched could not comply.

It appears to us to be undeniable that the selection of gate seniority as the criterion, while facially neutral, operated to the disadvantage of women who had to wait years to get jobs as compared with men who were employed immediately or without significant delay. All the complainants were in the position that because of the continuing effect of past discrimination at the point of hiring they had very much less "gate" seniority than men who applied for jobs at or about the same time.

We are satisfied that whether measured against the threatened cut-off date of 30th September, 1980 or the actual cut-off date of 6th January, 1981 the reverse gate seniority principle was a requirement or condition for continuing employment at the steelworks which answers the test in s. 24(3)(a), namely that a substantially higher proportion of men than women complied or were able to comply with it.

This passage does not articulate the precise train of reasoning which the Tribunal followed in reaching its conclusion that a substantially higher proportion of men than women complied or were able to comply with the condition imposed by A.I.S. The submission of Mr. Basten, which the Tribunal recited in the first paragraph of this passage, seemed to require a comparison of the respective numbers of male and female ironworkers who could comply with the condition imposed "on each employee" as at the date of retrenchment. Although the opening words of the second paragraph of the passage might suggest that the Tribunal accepted Mr. Basten's submission, the terms of the second and third paragraphs, read as a whole, make it plain, however, that the Tribunal did not accept his submission. Those paragraphs indicate that, for the purpose of determining whether a substantially higher proportion of male ironworkers than female ironworkers complied or were able to comply with the requirement or condition imposed by A.I.S., the Tribunal compared the complainants as one group and the male ironworkers who applied for employment at the same time as the complainants as the other group. Whether the Tribunal made a number or percentage comparison does not appear. Nor from a practical point of view does it matter. On the approach of the Tribunal, no member of the complainants' group could comply with the requirement or condition, and the Tribunal impliedly found that some "men who applied for jobs at or about the same time" as the complainants could comply. I say impliedly found because the statement that the complainants "had very much less "gate" seniority than men who applied for jobs at or about the same time" implies that some or all of the men, who applied for employment at the time the complainants did, were still in employment with A.I.S. when the retrenchment notices were given. Whether there was any evidence that this group of male ironworkers were still in employment with A.I.S. when the retrenchments were made does not appear. But I shall assume that the Tribunal was entitled to find that, when the notices of retrenchment were given, there were one or more male ironworkers still employed by A.I.S. who had applied for employment at or about the same time as the complainants had applied. If there were, then obviously the proportion of those male ironworkers (probably 100 per cent) who could comply with the requirement or condition was higher than the proportion of the complainants (0 per cent) who could comply with the requirement or condition.

But we know that in fact 93.73 per cent of female ironworkers and 93.23 per cent of male ironworkers employed by A.I.S. could comply with the condition or requirement that A.I.S. imposed on its workforce which was that a worker could continue in employment after the date of retrenchment only if he or she had been employed before 6 January 1981. How then did the Tribunal reach the conclusion "that a substantially higher proportion of men than women complied or were able to comply with it"? In my opinion, the Tribunal reached the conclusion which it did because it erred in law in confining the pool of male and female ironworkers for the purpose of s. 24(3)(a) to those workers who applied for employment at or about the time when the complainants applied.

Section 24(3)(a) states that a person discriminates against another person on the ground of sex "if he requires the other person to comply with a requirement or condition - (a) with which a substantially higher proportion of persons of the opposite sex to the sex of the other person comply or are able to comply ". Now the requirement or condition with which each complainant was required to comply was that she had been employed prior to 6 January 1981. To determine whether that requirement or condition was discriminatory the Tribunal had to determine what proportions of each sex could comply with that, and not some other, condition. And the proportions of the males and females who could comply with the condition were 7,177 out of 7,698 men (93.23 per cent) and 478 out of 510 women (93.73 per cent). In determining the proportions of the persons of the two sexes who complied or were able to comply with the requirement or condition imposed by A.I.S., the Tribunal was not entitled to disregard the statutory command of s. 24(3)(a) and select some only of the persons who complied with the condition. It was not entitled to disregard the 478 women or 7,177 men who complied with the condition. Whether the number of persons to whom a requirement or condition is directed and who are divided or segregated by it is large or small, the respective proportions cannot be determined without first ascertaining the numbers of males and females who comply or are able to comply with the requirement or condition. Yet the Tribunal has disregarded all the 478 women and probably all but a few of the 7,177 men who complied or were able to comply with the requirement or condition which A.I.S. imposed.

Mr. Basten, counsel for the respondents, submitted that the identification and ambit of "the pool" were questions of fact for the Tribunal. However, the relevant "pool" depends on the precise scope of the requirement or condition applied by the alleged discriminator. No doubt because of the terms of many requirements or conditions, the Tribunal has considerable latitude in many cases in identifying and in determining the ambit of the relevant pool. But the "pool" which the Tribunal selected in the present case was not determined in accordance with the terms of s. 24(3)(a). It was not based on the requirement or condition imposed by A.I.S.

The Tribunal's decision could only be upheld in my opinion if the requirement or condition applied by A.I.S. was that to continue in employment with A.I.S. an ironworker must have both commenced employment before 6 January 1981 and applied for employment at or about the time that the complainants applied for employment. If that requirement or condition had been applied by A.I.S., the Tribunal's decision would have been correct. In truth, what the Tribunal has done - no doubt unconsciously - is to amend the condition or requirement which A.I.S. applied by adding to it those words which I have emphasised.

Mr. Basten submitted, however, that the pool selected by the Tribunal was correct because "in determining a pool one should look for a pool which does not have a discriminatory element built in". He submitted that, since the composition of the A.I.S. workforce was the product of the discriminatory recruitment practices of the company, the Tribunal was entitled, if not required, to select base groups which were unaffected by these discriminatory practices. But if Mr. Basten's submission is correct, the Tribunal could not have picked a pool with a more discriminatory element built into it than the one which it did. Ironically, it was because the male ironworkers, in the pool selected by the Tribunal, had obtained their employment by discriminatory recruitment practices that the Tribunal was able to find that a greater proportion of male ironworkers than female ironworkers could comply with the "requirement or condition"!

However, the real answer to the submission of Mr. Basten is that s. 24(3)(a) requires the comparison of the proportions of the members of each sex who comply or are able to comply with the requirement or condition with which the complainant is required to comply. For the purpose of the statutory comparison, it is beside the point that the respective sex groups are the product of discrimination. The statutory command requires the comparison of the proportions of males and females who comply or are able to comply with the specified requirement or condition and those proportions can only be determined by ascertaining the number of females and males who comply or are able to comply with the requirement or condition. If in fact those proportions are the product of prior discriminatory practices, s. 24(3)(a) nevertheless requires their comparison. It is true that one of the purposes of the Act is to eliminate sex discrimination. Elsewhere, I have drawn attention to the need to construe an enactment by reference to its statutory purpose and to give effect to it "in so far as, by addition or omission or clarification, the relevant provision is capable of achieving that purpose or object": Kingston v. Keprose Pty. Ltd. [41] . But I do not think that it is possible to modify the grammatical meaning of s. 24(3)(a) by adding to it such words as "not being proportions which are affected by previous acts of discrimination". It must be remembered that, although one of the objects of the Act is to eliminate sex discrimination, sex discrimination in the Act has the artificial meaning given to it by s. 24. That is to say, to find out for the purposes of the Act what is sex discrimination in any particular case, you must apply the terms of s. 24(1) and s. 24(3). If the conduct of an alleged discriminator falls within one or other of the definitions in s. 24, then it amounts to sex discrimination. Otherwise it does not. It would be illogical to modify the definition for the purpose of giving effect to a supposed statutory purpose when the statutory purpose itself can only be discerned by reference to the terms of that definition as unmodified. In my opinion, it would not be a proper application of the principles of statutory construction to read down the statutory definition of sex discrimination in s. 24(3) by adding words such as "not being proportions which are affected by previous acts of sex discrimination" to the end of s. 24(3)(a). Indeed, if the Tribunal's finding in this case is any guide, a complainant will often be advantaged by the selection of a pool which is affected by previous acts of discrimination.

However, even if s. 24(3)(a) were read as if it concluded with such words as "not being proportions which are affected by previous acts of sex discrimination", it would not support the approach of the Tribunal. It would mean only that the base groups affected by the condition imposed by A.I.S. (that is, male and female ironworkers employed at 6 January 1981) would have to be adjusted to eliminate those male ironworkers who obtained their employment as the result of previous acts of discrimination. A restricted construction of s. 24(3)(a), therefore, would not justify the Tribunal in selecting as the base groups only those male and female ironworkers who applied for employment at the same time as Ms. Najdovska and the other complainants. They contained only a few of the persons who complied with the condition imposed by A.I.S., and it is the proportions of all the persons of the respective sexes who comply or are able to comply with the requirement or condition which must be compared.

Once it was established that 93.73 per cent of female ironworkers and 93.23 per cent of male ironworkers complied or were able to comply with the condition imposed by A.I.S., the Tribunal was bound as a matter of law to hold that the retrenchments and threat of retrenchments by A.I.S. were not in breach of s. 24(3). It is, therefore, unnecessary to determine whether the provisions of s. 24(3)(b) and (c) were satisfied. The appeal of A.I.S. in respect of the finding of sex discrimination within the meaning of s. 24(3) should be allowed.

The respondents sought to justify the awards of damages in their favour by asserting that the Tribunal was correct in finding that the retrenchments and threat of retrenchments of the respondents were a breach of s. 24(1) of the Act.

For the respondents to succeed on this ground it was necessary for them to establish that on the ground of sex A.I.S. treated them less favourably than male ironworkers in the same circumstances by threatening to retrench, and in retrenching, them. However, there was no evidence that it was on the ground of sex that the respondents were retrenched or threatened with retrenchment. On the contrary the respondents were retrenched on the ground that they had not commenced employment before 6 January 1981. Accordingly, the respondents cannot rely on breach of s. 24(1) of the Act to support the findings that the threats of retrenchment and the retrenchments amounted to independent acts of unlawful sex discrimination.

A.I.S. must succeed in its appeal and the damages awarded to the respondents must be reduced. This is hardly a satisfactory result because the amounts of damages which the respondents lose were the direct consequence of the discriminatory conduct of A.I.S. in delaying in hiring them. Unfortunately, the respondents are not entitled to full compensation for this discriminatory conduct of A.I.S. because of the statutory limit of $40,000 for any one act of discrimination. Mr. Basten sought to overcome this injustice by arguing that retrenchment on a "last on, first off" basis was an independent act of discrimination. But for the reasons I have given, the respondents are not entitled to treat the threat of retrenchment and the actual retrenchment as separate, unlawful acts of sex discrimination.

The appeal must be allowed. The order of the Court of Appeal dismissing the appeal of A.I.S. should be set aside. In lieu thereof, it should be ordered that the appeal to that Court against the awards of damages to the present respondents should be allowed and that the damages awarded to each of them should be reduced by the amounts specified in par. 9(a) of the notice of appeal. The respondents should pay the costs of the appeal to this Court. The matter should be remitted to the Court of Appeal to make such order as it thinks fit in respect of the costs of the proceedings in the Court of Appeal.