Australian Journal of Human Rights
Is discrimination on the basis of sex involved if an employer pays a part-time worker a higher hourly rate for overtime work only if that employee worked more hours than the normal full-time standard? "Yes" said the Dutch Equal Treatment Commission (ETC), "no" was the answer given by the European Court of Justice (ECJ), thereby de facto, be it not de jure, overruling the former. The ruling of the European Court poses a blow to efforts to improve the legal position of part-time workers, a group which consists largely of women, through the concept of indirect discrimination.
This contrasting outcome in what were two very similar cases, was due to a different application of the concept of indirect discrimination. In the application by the Dutch Equal Treatment Commission the concept of indirect discrimination is potentially far-reaching, the decision of the European Court of Justice shows that such a far-reaching effect is not self evident or compelling. This contribution explores the possibilities and limits of the concept of indirect discrimination, and the conditions under which it can reach more structural forms of discrimination and help achieve not just formal, but substantive equality. To what extent is it a vehicle for real change? This issue is not only relevant for women, but also for other non-dominant groups in society like ethnic and religious minorities or homosexuals. For my analysis I will draw on experiences with the concept of indirect discrimination in various jurisdictions. With slight variations, it is known and applied in countries like the United States, South Africa, Canada, those of the European Union, and Australia. This article will also pay special attention to the jurisprudence of the Dutch Equal Treatment Commission, a semi-judicial body which is developing quite an impressive body of case law on this subject, covering indirect discrimination on the basis of sex, race, nationality, sexual orientation and religion.
This comparative analysis hopes to contribute to a better understanding of the way the concept works (or fails to work) in different legal settings and how to improve its potential for reaching more subtle, systemic forms of discrimination. In all jurisdictions discussed the latter is much in need of improvement. Australia is no exception. The Report of the Australian Law Reform Commission ‘Equality before the law: justice for women’ refers to indirect discrimination as a key issue in discrimination law. However, it immediately adds that, at least where the Australian Sex Discrimination Act is concerned, the provisions dealing with this form of discrimination have not realised its potential: the definition of indirect discrimination is too complex and it is considered difficult to prove a complaint. Their 1994 analysis still holds true. The following, conceptual analysis will hopefully provide inspiration to amend these shortcomings.
The United States can very probably claim to be the "founding fathers" of the concept of indirect discrimination, or in the American terminology: disparate or adverse impact. The roots of disparate impact analysis can be traced to the famous Griggs case. In this case the US Supreme Court had to decide whether completion of highschool as a condition for promotion constituted discrimination on the basis of race, as blacks would have much greater difficulty in meeting such a requirement than whites. The applicable federal legislation, Title VII of the Civil Rights Act, prohibits discrimination on the basis of race, colour, religion, sex and national origin in employment. It does not, however, explicitly cover both direct and indirect discrimination. The Supreme Court ruled that an employer's facially neutral employment practice, which is not based on race as such, but which can be shown to have a disparate impact on blacks, establishes a prima facie case of race discrimination. To save the practice the employer will have to show the practice is jobrelated or otherwise constitutes a "business necessity". In Griggs the Supreme Court concluded the high school completion requirement not to bear a demonstrable relationship to succesful job performance and thus not to serve a business necessity. This approach was later extended to sex discrimination cases.
In Europe, a very similar concept was developed by the European Court of Justice in the area of sex discrimination. It did not come out of the blue. To start with, the European Court of Justice was already familiar with the notion of indirect discrimination on the basis of nationality. Given the importance of eradicating discrimination on this ground in view of the primary goals of the European Community (free flow of workers, goods and services) it can not be surprising that the European Court of Justice was keen to address all kinds of more subtle and hidden forms of discrimination on the basis of nationality, and not just outright exclusions. Thus it held in the case Sotgiu that "the rules regarding equal treatment forbid not only overt discrimination by reason of nationality, but also covert forms of discrimination which, by application of other criteria of differentiation, lead in fact to the same result". The second influence on the development of the concept of indirect discrimination in the context of sex discrimination cases stems from developments in the United States. In the first indirect sex discrimination case to reach the European Court of Justice, Advocate General Warner drew explicitly on Griggs to convince the Court that paying part-time workers less than full-time workers indeed amounts to indirect discrimination on the basis of sex, in contravention of the principle of equal pay enshrined in article 119 (currently art 141) of the EEC-Treaty.
In the landmark case Bilka the European Court of Justice formulated the test to be applied in indirect discrimination cases, which it has consistently reiterated in subsequent cases. At stake was the exclusion of part-time workers from an occupational pension scheme. Though this practice was in itself sex neutral, in its practical effects it excluded much more women than men, given the overrepresentation of women in part-time work. The European Court of Justice ruled that the right to equal pay:
is infringed by a department store company which excludes part-time employees from its occupational pension scheme, where that exclusion affects a far greater number of women than men, unless the undertaking shows that the exclusion is based on objectively justified factors unrelated to any discrimination on grounds of sex.
The Court subsequently qualified this "objective justification" requirement. The measure or practice has to meet "a real need of the enterprise" and to be "appropriate and necessary" to that end. In later cases, where indirect discrimination by the state was at stake, the "real need of the enterprise" requirement was replaced by the more general notion of "legitimate aim".
Conceptually speaking the likeness between the European and American concepts of indirect discrimination and disparate impact is striking. Differences stem from the different context in which the concepts are operating, but do not affect their basic similarity. The American notion of disparate impact under Title VII of the Civil Rights Act is clearly tailored to the employment context, whereas the European notion of indirect discrimination is formulated in a more general way and can thus rather easily be applied in a broader area and to a broader range of grounds of discrimination. To use the concept in such a general way, I would like to define indirect discrimination rather loosely as any distinction, practice or policy which does not discriminate expressly on any of the commonly distinguished grounds of discrimination like race, sex, religion, sexual orientation, handicap or age as such, but which has a discriminatory or disparate impact on a group protected against such discrimination, and for which no objective justification can be shown to exist.
The way the Dutch Equal Treatment Commission has applied the concept of indirect discrimination provides an interesting example of its far-reaching potential.
The Dutch Equal Treatment Commission is a semi-judicial body. Its main task is supervision of the Equal Treatment Act, which came into force in 1994. To this end it can hear complaints and give its opinion, which is not legally binding. The Equal Treatment Act covers discrimination on the basis of religion, belief, political opinion, race, sex, nationality, sexual orientation and civil status in all major areas of social and economic life like employment, the liberal professions, insurance, banking, housing, welfare, education and healthcare. Apart from this act, the Dutch Equal Treatment Commission supervises a much older equal treatment act which deals specifically with sex discrimination in employment. Since the end of 1996 it can also hear complaints under an act prohibiting discrimination on the basis of a difference in working hours (for instance between part-time and full-time workers), which is closely related to indirect sex discrimination.
With some 100-150 opinions rendered a year, the case law of the Dutch Equal Treatment Commission covers a wide range of cases. The majority still concern sex discrimination, which may be explained by a much longer tradition of bringing this type of cases, which already existed under the act regarding sex discrimination in employment. In the following I will give some examples of the way the concept of indirect discrimination works in the hands of the Dutch Equal Treatment Commission, which will show some of its conceptual strength. The Dutch Equal Treatment Commission closely follows the steps of the test as developed by the European Court of Justice:
- does the practice or measure disproportionately affect a protected group?
- if so, can the perpetrator advance an objective justification, that is, is the practice or meaure appropriate and necessary to achieve a legitimate aim, which is not related to any discriminatory purpose?
Especially the ‘necessary’ element can give the test a real bite. Following the European Court of Justice, the Dutch Equal Treatment Commission often considers in this respect whether the same goal could have been achieved by different, non-discriminatory means.
The most extensive part of the case law of the Dutch Equal Treatment Commission concerns sex discrimination. In applying the concept of indirect discrimination, the Dutch Equal Treatment Commission has shown it can be used to challenge succesfully several more subtle and systemic forms of discrimination against women. In this respect it draws much inspiration from the (legally binding) case law of the European Court of Justice, yet it takes the concept at some points a major step further. A first example concerns discrimination in pay practices. The Dutch Equal Treatment Commission acknowledges the detrimental effects on women of basing salaries on experience in gainful employment only and on the last earned pay. As women far more often than men leave the labour maket for some time to take care of their children, such a policy has a disparate impact on women reentering the labour market. Drawing on the decision of the European Court of Justice in Nimz, the Dutch Equal Treatment Commission does not consider such policies objectively justified. It demands that all relevant experience, including relevant experience from unpaid work, is taken into account and assessed on an individual basis. As far as the last earned pay provides a proper indication of relevant experience it is justified to use this as a starting point, but the employer should periodically reconsider whether, compared to a man doing work of equal value, a lower salary on the basis of lesser experience at the start of the employment is still warranted.
Exclusionary pay practices deriving from a segregated labourmarket have also been quite succesfully challenged through claims of indirect discrimination. Many occupational pensionschemes, for instance, exclude typically "female" jobs like administrative functions or cleaning from coverage. Likewise, other financial arrangements are sometimes reserved for technical, that is mainly male, personell only. As such exclusions affect women much more than men, the Dutch Equal Treatment Commission closely scrutinises the justifications advanced. On the whole, it does not consider such practices objectively justified. Financial considerations of the employer are not accepted in themselves as a justification and even if the scheme is part of a collective labour agreement this cannot save it.
The last example in the area of pay concerns a job evaluation scheme which was widely used in the healthcare sector, covering some 150,000 persons. On the basis of an extensive and thorough examination of the system, the Dutch Equal Treatment Commission concluded it harboured several subtle mechanisms which led to a lower appraisal of (elements of) typically "female" functions like nursing compared to (elements of) typically "male" functions like management. To start with, the different elements of nursing activities were much less extensively identified and described than elements of managing, which were laid down in detail. Besides, descriptions of male functions were often formulated more weighty than female functions. Thus aspects of management in the total score for functions was given far more weight than aspects of care taking. Further, similar elements were valued less in female functions than in male functions. Thus, unattractive elements in the latter were compensated for to a larger extent than in the former. And to give a last remarkable example of gender bias in the system: the regular and sensitive contacts of nursing personell on a ward for premature babies with the babies' parents was valued hardly any higher than the contacts some categories of technical personell could incidentally have with patients.
Another area where the Dutch Equal Treatment Commission has developed a rather extensive body of case law challenging more structural forms of discrimination concerns part-time work. As women work far more often part-time than men due to family responsibilities, any distinction between part-time and full-time workers is potentially discriminatory. Following to a large extent the case law of the European Court of Justice, the Dutch Equal Treatment Commission has rendered a long line of opinions in which it makes clear part-time work is to be valued equally as full-time work. Thus, part time workers may not be excluded from pension-schemes, they may not be excluded from policies to reduce the hours of work, with full pay, for older workers, and they have to receive the same compensation for their health insurance as full-time workers, and not pro rata.
A remarkable line of cases deals with the "right" to a part-time job instead of a full-time one, to facilitate combining paid work and care taking activities. The Dutch Equal Treatment Commission construes the refusal of an employer to make a part-time job available, as a policy which has a disparate impact on women, as women much more than men are in need of and looking for part-time work as they still bear the larger part of caretaking activities. This means the employer has to advance an objective justification for his practice of making a job full-time or nearly full-time available only. To this end the employer has to show the job can only be done by being available for the required number of hours. Similarly, the request of a worker to reduce her number of working hours may not be refused unless the employer can show his refusal is objectively justified, that is, is both appropriate and necessary to achieve a real need of his enterprise. The Dutch Equal Treatment Commission is not inclined to be very lenient towards the employer.
Discrimination on the basis of race, nationality and religion
The line of cases covering discrimination on the basis of race (to be understood in the broad sense of the UN Convention against Racial Discrimination), nationality and religion is much less extensive. Still, it provides several examples of how the concept of indirect discrimination can be used in this area as well.
In the Netherlands, discrimination on the basis of race, nationality and religion are quite often related, due to the growing presence of ethnic minorities from the former Dutch colonies, from Turkey and Marokko (migrant workers), and as asylum seekers. One of the first type of cases on indirect discrimination to reach the Dutch Equal Treatment Commission concerned job requirements demanding a good command of the Dutch language. Such requirements evidently have a disparate impact on ethnic minorities. The Dutch Equal Treatment Commission considers the requirement of proficiency in the Dutch language as justified if it is necessary for performance of the job. Thus, it is not discriminatory to require it for the function of a librarian or for a doorman, but is not allowed for a simple cleaning job.
An employer's dress requirements which may conflict with religious dress have given rise to another range of cases. The Dutch Equal Treatment Commission is again pretty strict. Wearing a turban because of the Sikh religion, or a headscarf in respect of being moslim can be banned for valid safety reasons, but not for being unrepresentative in the eyes of customers or creating irritation with co-workers or clients. A public primary school which refused a muslim woman wearing a headscarf as a trainee, because it took this as a manifestation of an intolerant attitude conflicting with the open character of public education, also violated the prohibition of indirect discrimination in the eyes of the Dutch Equal Treatment Commission. The presence of a tolerant or intolerant attitude has to be assessed on an individual basis and may not be based on ascribed characteristics of persons wearing headscarves.
Interestingly, the concept of indirect discrimination can also be used to support claims for reasonable accommodation of religious requirements like praying at work or taking leave on Friday to vist the mosque. The employer who refuses such accommodation has to come up with an objective justification for his refusal.
Indirect discrimination on the basis of nationality is a rather regular feature in the conditions for selection and acceptance of customers. To reduce the risk of customers not being solvent or being untracable in case of non-payment, customers are often asked to present proof of permanent residence like a passport or some specific other immigration document. If they cannot provide one of these, they may be confronted with additional conditions for acceptance. It is evident that such policies affect foreigners much more than people of Dutch nationality, and so indirect discrimination is at stake unless an objective justification can be advanced. On the whole the Dutch Equal Treatment Commission has acknowledged that the purpose of the policies, that is reduction of risk, is legitimate. Yet, it has also consistently held that firms should look beyond the residence permits shown, as a temporary permit does not always give an adequate indication of the actual legal position of the individual concerned and thus to the presence of any riskfactors. The means chosen should be more narrowly tailored and not be based on overbroad generalisations.
This very quick scan of the opinions of the Dutch Equal Treatment Commission shows some of the conceptual potential of indirect discrimination analysis. By focusing on the results of a certain treatment instead of the nature of the treatment in terms of "the same/not the same" itself, it belongs to the realm of notions of substantive equality. In a formal conception of equality, the question whether people are treated the same irrespective of race, sex, disability, religion or sexual orientation and the like is determining. As long as the treatment is the same, problems of equality or discrimination cannot arise, however unequal the results of such treatment may be. A formal equality approach is simply blind to unequal outcomes stemming from, for instance, social and economic differences. But this is exactly where a substantive approach to equality starts from. Unequal results are the triggering device for formulating a problem in terms of equality and discrimination. It is irrelevant whether these result from the same treatment or not. The concept of indirect discrimination similarly identifies unequal results as problems of equality and discrimination. Unequal results demand further scrutiny to see whether the treatment can be justified, notwithstanding this effect. In this way, it directs attention to the myriad forms of more systemic and subtle forms of discrimination, which are not as visible as express discrimination, but which often have an all the more negative impact on vulnerable, non-dominant groups.
The far-reaching potential of the concept of indirect discrimination was clearly acknowledged by the United States Supreme Court in Washington v Davis, a case concerning racial discrimination. The Court had to decide whether disparate impact analysis as developed under Title VII of the Civil Rights Act would be applicable under the constitutional equality clause as well. Given the distribution in the United States of affluence and poverty along racial lines, and the broad scope of the constitutional protection against discrimination, Justice White showed a clear awareness of the tremendous implications which adoption of the disparate impact approach could have:
A rule that a statute designed to serve neutral ends is nevertheless invalid, absent compelling justification, if in practice it benefits or burdens one race more than another, would be far-reaching and would raise serious questions about, and perhaps invalidate, a whole range of tax, welfare, public service, regulatory, and licensing statutes that may be more burdensome to the poor and to the average black than to the more affluent white.
The Supreme Court was apparantly not willing to take this risk of having to invalidate all kinds of federal and state statutes because of their disparate impact on blacks. To block this, the Court in Washington v Davis ruled that discriminatory intent is required to bring an effective claim of discrimination under the Constitution. It held that to be constitutionally valid " ... the invidious quality of a law claimed to be racially discriminatory must ultimately be traced to a racially discriminatory purpose".
In this way, the Supreme Court reduced the use of disparate impact analysis to specific and limited areas like the employment sphere covered by Title VII.
A requirement of "intent" to bring a succesful claim is a very effective means of curtailing the scope of the notion of indirect discrimination. In fact, it more or less kills it. The decision of the Supreme Court in Feeney provides a good example. In this case a woman challenged a statute of the State of Massachusetts which gave absolute preference to army veterans applying for state civil service jobs. As this group was almost exclusively male (98%) and constituted more than 25% of the entire state population, women were in fact almost completely excluded from the state civil service, except for traditionally female jobs, for which male workers would not apply. A clear example of indirect discrimination, one would say: the disparate effect on women is obvious, and even if the purpose of helping army veterans may be legitimate because of services rendered to the general interest, the rather absolute exclusion of women makes the means chosen hardly appropriate and certainly not "necessary".
The Supreme Court, however, never even reached any balancing of interests as would be required under the objective justification test. It killed the claim of discrimination from the start by holding that the contested provisons did not infringe the constitutional equality clause, as they did not intend to discriminate against women. Their purpose was to reward the services rendered by the army veterans, not to prefer men over women. Not just women were affected, some men were disadvantaged as well. The disproportionately large number of women affected does not make any difference to the Court:
“Discriminatory purpose", however, implies more than intent as volition or intent as awareness of consequences. ... It implies that the decisionmaker, in this case a state legislature, selected or reaffirmed a particular course of action at least in part "because of", not merely "in spite of", its adverse effects upon an identifiable group.
This intent requirement renders the whole concept of indirect discrimination harmless. First of all, the barriers to overcome in proving intent are formidable. How is one to identify the intent of institutional decision makers like legislatures? The lawmaking process in a modern democratic state is complex and involves many actors, all adding their own "intent". Besides, more often than not law has a compromise character and may be the result of competing aims. A motivational approach to intent is a model based on what individual people do, it does not seem very well suited to institutional decision making. Secondly, it is questionable whether intentional discrimination is really what it is all about. If anything, experience with race and sex discrimination has shown the larger problem of exclusion and disadvantage of racial minorities and women to derive from much more subtle and unintentional mechanisms than outright and straightforward express discrimination. By requiring intent, the Supreme Court is sidestepping the more fundamental problems of inequality and discrimination. It means that, however outragious the exclusion may be, the constitution will turn a blind eye as long as the state can be considered not to have had any bad intentions. It also means that similar forms of disadvantage may be condoned by the Constitution in one part of the country but not in another, just because of a different state of mind of the local governments involved.
Given these serious criticisms on the intent requirement, the Supreme Court really must have been worried very much by the potential implications of adopting disparate impact analysis under the Constitution. But how real are the Court's fears? To my mind, the Supreme Court is cracking a nut with a sledgehammer. Though the concept of indirect discrimination may have a sweeping effect indeed, it also has an important mitigating mechanism: the objective justification test. This instrument is of crucial importance. If used carefully and intelligently, it can provide a good tool of balancing all interests concerned. On the one hand it requires all policies or measures having a disparate impact on protected groups to be critically examined. On the other it makes sure not every treatment having such an effect is, by definition, unlawful. If the latter were the case, as is sometimes defended, it would make the concept of indirect discrimination too rigid and thus unworkable, or indeed too threatening.
Disadvantage as a determining factor
Contrary to the United States Supreme Court, the Canadian Supreme Court has less qualms about taking disparate impact as sufficient proof of a prima facie case of discrimination. Under the Canadian Charter of Human Rights intent to discriminate is not required. To quote Justice MacIntyre, writing for the Court in Law Society of British Columbia v Andrews:
I would say then that discrimination may be described as a distinction, whether intentional or not but based on grounds relating to personal characteristics of the individual or group, which has the effect of imposing burdens, obligations, or disadvantages on such individual or group not imposed upon others, or which withholds or limits access to opportunities, benefits, and advantages available to other members of society.
Indeed the Canadian Court seems to take adverse effect discrimination even a step further on the way towards substantive equality. In its judgment in Eldridge it held that the failure of hospitals to provide sign language interpretation for deaf patients where it is necessary for effective communication constitutes prima facie discrimination. Again, adverse effect is determining, nothing else. And as the Government could not demonstrate that their action infringed the right of deaf persons in this respect no more than is reasonable, as they excluded this service alltogether, it could not be saved.
The decision in Eldridge is rather unique because it does not deal with a seemingly neutral differentiation with an adverse effect on a protected group, but with a lack of differentiation. It addresses the need to accommodate differences between groups, to take positive steps to ensure that disadvantaged groups benefit equally from, in this case, healthcare. In this respect the Canadian Court reaches the very heart of substantive equality: results, that is actual enjoyment of all human rights on an equal basis, is determining. To realise this end, differential treatment may be warranted.
At the same time Eldridge raises some difficult questions involved in a substantive equality approach, which derive from the open-ended nature of substantive equality. How far should providing facilities for deaf persons go? Should the hospitals provide sign language interpretation every time any deaf person visits a doctor? Or only if major decisions have to be taken? And if deaf people are entitled to such facilities, what about other groups who have communication problems, for example, those who do not have sufficient command of the official language? In Eldridge the Supreme Court of Canada solves the first problem by limiting the constitutional duty to provide sign language for deaf persons to situations where it is "necessary for effective communication". As to the second problem, it sidesteps an argument of this nature advanced by the government by terming it "speculative".
The potential of the concept of indirect discrimination depends highly on the way it is applied. In this respect it is vulnerable in both parts of the test: establishing disparate or adverse impact on a protected group and, if so, deciding whether an objective justification exists.
Establishing disparate impact on a protected group in fact involves two steps itself. A practice or policy must be shown first of all to have a detrimental effect, and secondly, to affect a protected group disproportionately. Though the first step seems quite easy, in practice it sometimes turns out to be pretty complicated. Especially cases concerning part-time work have proven to pose difficult problems. In fact, the diverging ways in which the Dutch Equal Treatment Commission and the European Court of Justice considered this aspect in the case cited in the introduction was responsible for the very different outcome. The case concerned a German pay practice of paying a bonus for overtime work only if the hours worked surpassed the normal full-time standard. Thus, a part-time worker with an individual labour contract of 20 hours would not receive any extra pay for overtime work unless (s)he worked more than 40 hours. Though it seemed evident to the Dutch Equal Treatment Commission in a very similar case that such a pay practice disadvantages part-time workers, the European Court of Justice had a different view. Instead of taking the individually contracted hours as a starting point for defining overtime work, and comparing the pay of full-time and part-time workers in respect of the hours worked on top of that contract, the Court compared the overall wage received by part-time workers and full-time workers for the same number of hours worked. This comparison leads to an entirely different conclusion: as both part-time and full-time workers receive the same pay for each hour worked (a part-time worker with a labour contract of 20 hours receives the same pay for the 24th hour worked as the full-time worker) no differential treatment exists, and thus no disadvantage, to start with. In this way, the claim of indirect discrimination on the basis of sex was killed right from the beginning. So the European Court of Justice never even reached the stage of having to decide whether an objective justification was present, a justification which the Dutch Equal treatmet Commission thought was absent in a very similar case.
Other cases brought to the Dutch Equal Treatment Commission and to the European Court of Justice increasingly show the difficulties of assessing the disadvantage involved in differing conditions of labour for full-time and part-time workers. Is it a disadvantage for part-time workers if they receive a pro rata compensation only for time spent on further education or union activities, or are full-time workers disadvantaged if they only receive the same compensation as part-time workers? And what about having full-time and part-time workers pay the same amount of money for private use of a company car. Is that a practice which disadvantages full-time workers, as they have less time off work than part-time workers to use it privately in the first place? The answers given will decide whether there is any basis for an indirect discrimination claim to start with.
The second step of establishing a disparate impact claim concerns linking the adverse impact once identified to a group protected by the anti-discrimination clause. The vulnerability of this step lies in the amount of proof required: is statistical evidence required, and if so, how detailed and how well tailored to the specific situation should it be? Or is it sufficient to refer to "facts of common knowledge" to establish such a link, as the Dutch Equal Treatment Commission has accepted in several cases? To give an example: does a job requirement of fluency in Dutch establish a prima facie case of indirect discrimination on the basis of race because it is a fact of common knowledge that ethnic minorities in general have a lesser command of that language and are thus disadvantaged by such a requirement? Or should statistical proof be provided before a claim can be succesful? And if statistics are advanced, which ones are relevant? Should we look at the overall fluency of persons pertaining to ethnic minorities, or only of those in the relevant labour pool? And should we use relative or absolute figures? Evidently, the more sophisticated statistical evidence is required, the more difficult it will be to establish a prima facie claim of indirect discrimination. Not just because information may not be readily available or will have to be provided by the defendant, but also because it can be very costly.
The European Court of Justice seems pretty lenient in this respect. It had no difficulty in presuming criteria like seniority, length of occupational training and flexibility in working at variable hours and at variable places to have a disparate impact on women. The overall division of labour between men and women, which makes women the primary caretaker, was sufficient to establish that women will have more difficulty in meeting such criteria than men.
The question whether statistical evidence is required or not also points to another problem: how much disparity is needed to establish disparate impact? In the Feeney case the disproportionate effect on women was overwhelming, but in other cases this may be quite different. Again, the European Court of Justice seems to be inclined to make things not too hard for the claimant. In Danfoss a 6.85 percent wage differential between male and female workers in the same function group was accepted as providing sufficient evidence of a prima facie indirect discrimination claim. Yet in the case of Seymour-Smith & Perez it required a “considerably smaller percentage”of women than men to be able to fulfill the contested requirement.
The objective justification test lies at the heart of the concept of indirect discrimination. Even if the steps just mentioned can effectively neutralize claims of indirect discrimination, in most cases the question whether an objective justification exists is determining. In this stage very much depends again on the way the test is applied, especially how strictly or leniently. Generally speaking, both in the United States and in the EC the test is formulated in pretty strict terms: "business necessity" in the context of Title VII of the Civil Rights Act and "necessary" under EC provisions both suggest a close judicial scrutiny. Yet, in both the case law of the United States Supreme Court and the European Court of Justice we see a relaxation of the objective justification test.
As far as the European Court of Justice is concerned, this is especially apparent in social security cases. Here, the European Court of Justice increasingly leaves a large margin of appreciation to the member states of the EU. The Molenbroek case provides a good illustration of this development and of the difference it makes whether the objective justification test is applied leniently or strictly. Molenbroek concerned a Dutch old age pension statute, which guarantees everyone turning 65 a fixed monthly benefit. For persons with a dependant spouse under 65 a supplement is available to guarantee every household a certain minimum income, the so called "social minimum". In receiving the supplememt, any income earned by the spouse is deducted, whereas any income by the claimant, like an occupational pension, is not taken into account. Though these provisions do not differentiate expressly on the basis of sex, men receive such supplements far more often than women, given the fact that men are generally older than their spouse and, even if this is not the case, generally will be gainfully employed. Besides, far more women than men who have a spouse turning 65 will also be disadvantaged as they will be financially discouraged to stay in their job, as this will negatively affect the supplement.
Several claims of indirect discrimination on the basis of sex were brought simultaniously to different lower Dutch courts dealing with social security cases. Interestingly, they came to differing conclusions, depending on the strictness with which they applied the "appropriate and necessary" requirement of the objective justification test. One court accepted the purpose of the statute to guarantee every household the social minimum as a legitimate aim of social policy, yet did not consider the means employed to be appropriate and necessary. Though the supplement is necessary to ensure the social minimum in some cases, in many others households receive far more than that because the claimant has additional income. A second court on the other hand considered the provisions to be necessary in a sufficient number of cases to ensure the social minimum, to save the statute. According to this court, the "overinclusion", to put it in American terms did not stand in the way of the provisions regarding the supplement being "necessary".
In its answers to the preliminary questions put to it regarding these cases, the European Court of Justice followed the latter, more lenient approach. In subsequent cases dealing with social security issues, this trend was continued. In several cases the European Court of Justice seems to have applied no more than a "reasonableness" test, requiring the member state to show the provisons to be "not unreasonable" only.
A similar trend is apparant in the case law of the United States Supreme Court dealing with disparate impact. Though the standard of "business necessity" seems tough to meet on paper, the Supreme Court has been more and more lenient towards the employer. Thus in its Wards Cove decision the Court required the employer to show that the challenged practice "serves, in a significant way, the legitimate employment goals of the employer ... The touchstone of this inquiry is a reasoned review of the employer's justification for his use of the challenged practice". This seems a pretty far cry from a requirement of "business necessity".
The examples given above clearly show the strength of the concept of indirect discrimination to be largely dependant on the strictness of the review. At least where the European Court of Justice is concerned, the increasing leniency of the Court in applying the objective justification test in social security cases may very well be due to political pressure by the member states of the EU, which are getting nervous about the major, especially financial, impact its case law on indirect discrimination in this area has. If anything, this is a confirmation of the great potential of the concept of indirect discrimination.
Two other factors which may highly influence the effectiveness of claims of indirect discrimination should be mentioned here. One concerns the conceptualization of what discrimination is in a much wider sense, the other concerns the question of the burden of proof.
Symmetrical and asymmetrical conceptions of discrimination
A crucial aspect of discrimination analysis, be it direct or indirect discrimination, concerns the question whether the non-discrimination principle is conceptualized in a symmetrical or asymmetrical way: is it basically concerned with eradicating irrelevant distinctions, whoever is disadvantaged by such distinctions (symmetrical approach), or is its basic object and purpose to combat more structural patterns of disadvantage affecting members of non-dominant groups (asymmetrical approach)? The latter means the non-discrimination clause is essentially directed at the protection of vulnerable groups like racial minorities, women, homosexuals and the disabled, and only secondarily at protecting ethnic majorities, men, heterosexuals or able people. The consequences of one or the other approach are far reaching indeed. This can best be shown by the different approaches taken by the majority and the dissenters in the American case of Adarand Constructors.
In this case the Supreme Court of the United States was faced with a federal preferential treatment policy concerning federal agency contracts. The policy provided for financial compensation if the prime contractor hired subcontractors certified as small businesses controlled by socially and economically disadvantaged individuals. Racial minorities were presumed to fall into this category. Adarand Constructors was not a certified business. Because of this policy, it missed out on a contract. Adarand claimed the race-based presumption in the preferential treatment policy violated the equality clause of the Constitution.
The Supreme Court agreed. What is important to notice here is the level of scrutiny it applied to reach this decision. The Court decided the long standing question whether "benign" discrimination, intended to improve the position of racial minorities, should be considered in the same way as "negative" discrimination against them. It answered this question in the affirmative: the level of review in both cases should be strict scrutiny. In this respect the Court stressed the personal character of the right to equal protection: it has the same meaning for everyone, as the equality clause "protects persons, not groups". Thus, the Supreme Court clearly took a symmetrical approach to the equality clause. The dissenters vehemently attacked this approach. They stressed the need to differentiate between benign and "negative" discrimination in the level of scrutiny to be applied, as "There is no moral or constitutional equivalent between a policy that is designed to perpetuate a caste system and one that seeks to eradicate racial subordination". In their eyes the majority approach is flawed because it ignores the decisive relevance of the social and historical context of discrimination on the basis of race. A proper, contextual approach means that only classifications which exclude racial minorities should be subject to strict scrutiny, not those meant to be inclusionary. By giving center stage to the (historical) context of race discrimination the dissenters took an asymmetrical approach. The differing results of both approaches are evident: under strict scrutiny hardly any classification will survive.
In Europe, the line of the European Court of Justice is ambiguous. The cases of Kalanke and Marschall, both dealing with preferential treatment of women, show a very formal, symmetrical apporach. However, their cryptical and unclear motivation and the extensive critique they generated suggested that subsequent case law could very well take a different turn. In fact this seems to be what happened: in its recent decision in Badeck the ECJ took a very lenient approach to a German preferential treatment programme.
In Canada and South Africa on the other hand, the constitutional courts have adopted an asymmetrical approach. The amelioration of the position of disadvantaged groups in society is considered to be a major purpose of the equality and non-discrimination clause in both the South African Constitution and the Canadian Charter of Human Rights. The South African Constitutional Court, not very surprisingly given the history of Apartheid, is very explicit in this respect: "The more vulnerable the group adversely affected by the discrimination, the more likely the discrimination will be held to be unfair". How this will translate into the review applied is not yet altogether clear. The case of Walker shows white people may bring a successful claim of (indirect) discrimination as well, even if they cannot be considered to be a socially and economically disadvantaged group. Yet, reading the decision of the Court leaves the impression that the Court is indeed being less strict in its review of the non-discrimination clause than it would have been if black people had been similarly affected.
The importance of starting from an asymmetrical approach may be even more important in indirect discrimination cases than in direct ones, as it should be born in mind that all kinds of ordinary social policy measures directed at improving the position of disadvantaged groups would come under attack if a symmetrical approach were taken. Housing improvement subsidies for inner city dwellers only, for instance, could then be seriously challenged by suburban residents as being indirectly discriminatory on the basis of race, as ethnic minorities are concentrated in city centers. Indeed, a Dutch author terms this type of policy "indirect positive discrimination", with all the onus accompanying the term. This seems outragious. Or rather, it would be outragious if such policies were to meet with as strict a review by the courts as "negative" discrimination. It seems evident equality and non-discimination clauses are not written to jeopardize such policies, which are part and parcel of the social welfare state. An asymmetrical approach to discrimination analysis thus seems the only one which makes sense.
Burden of proof
The other factor to mention here which may influence the actual scope of the concept of indirect discrimination concerns the burden of proof. Typically, in indirect discrimination cases the burden of proof lies with the defendant once the adverse effect of a policy or meaure on a protected group has been established. United States case law shows how important this is. In Griggs, the burden of proof was squarely placed on the defendant, once a prima facie claim of disparate impact discrimination was established. In Wards Cove, however, the Supreme Court took a different turn and shifted the burden of proof back to the claimant. The dissenting justices were outraged. They regarded this approach as turning a blind eye to the meaning and purpose of Title VII, and bitterly wondered "whether the majority still believes that race discrimination ... is a problem in our society, or even remember that it ever was". In reaction to the Wards Cove decision, which was unacceptable to the federal government, Title VII was amended to undo the Supreme Court decision. Still, the decision shows the impact the division of the burden of proof can have.
The European Court of Justice on the other hand seems to be inclined to shift the burden of proof even further to the defendant. In Danfoss it decided an employer bears the risk of employing a pay system which lacks transparency: "where an undertaking applies a system of pay which is totally lacking in transparency, it is for the employer to prove that his practice in the matter of wages is not discriminatory, if a female worker establishes, in relation to a relatively large number of employees, that the average pay for women is less than that for men". Thus, the claimants did not have to prove the disparate impact to be related to a specific employment practice of the employer, like in the United States. In the case at hand it meant the Danish Union did not have to establish a direct link between the use of criteria like seniority, length of training and mobility on the one hand and pay differentials between men and women on the other. Establishing a pay differential of 6.85 percent between men and women in the same function group was deemed sufficient to shift the burden of proof to the employer. Of course, this makes a tremendous difference in bringing a succesful claim.
From the above it will be clear that the question whether the concept of indirect discrimination is or will be a vehicle for change is not easy to answer. If anything, the concept is raising questions of equality and non-discrimination wherever vulnerable groups are disadvantaged, regardless of whether this results from a differential treatment or not. Simply put, it is able to challenge all kinds of dominant standards having a detrimental effect on non-dominant groups as problems of equality. In this respect the concept signifies a big step forward along the way to substantive equality.
At the same time, we should realise the concept is necessarily limited, if only because law in itself is an inherently limited instrument for change. Though this may sound like a platitude, it is necessary to keep it in mind. Thus, for instance, the concept of indirect discrimination may direct attention to the fact that requiring fluency in the majority language has an adverse impact on ethnic minorities and is thus prohibited unless objectively justified, but that in itself does not alter anything in the underlying, educational disadvantage suffered by those groups. To achieve genuine equality, much more is needed.
Within such limits, however, there is quite a lot of space for a concept of indirect discrimination to make a real difference. In the Australian context, as in other jurisdictions, whether indeed it will seems to depend on two factors: a clear and proper definition of the concept, and an interpretation by the courts which does not render its potential bite harmless. The importance of both is shown clearly in the experiences with the Sex Discrimination Act.
According to the Australian Law Reform Commission, the indirect discrimination provision in the SDA is far too complex and difficult to prove. The four-part test demands that a complainant is required to (1) comply with a requirement or condition (2) with which she does not or is not able to comply, (3) to which a substantially higher proportion of men complies or is able to comply, and (4) which is not reasonable having regard to all the circumstances of the case. As the onus of proof is on the complainant as well, the test is very difficult to satisfy. The Australian Law Reform Commission therefore recommends to simplify the indirect discrimination provision and to reverse the burden of proof.
The above analysis supports the importance of implementing those recommendations. In fact, shifting the burden of proof to the alleged discriminator once the disparate impact of a measure or practice on women is shown to exist would be in line with the legal practice elsewhere. Yet the analysis also shows much more is needed. Especially the objective justification test, in Australian terms the reasonableness test, which eventually lies at the heart of the concept of indirect discrimination, is crucial. As we have seen, in Europe the ECJ has formulated strict criteria and it seems very important to include such criteria in the legal definition of indirect discrimination. Unfortunately, the Court itself shows in its case law that strictness in formulation is no watertight guarantee for strictness in its application. At the end of the day, then, the potential of the concept of indirect discrimination depends heavily on the courts. In this respect as well the road ahead is strewn with rocks and potholes. In the United States, the Supreme Court has effectively neutralized the whole notion of disparate impact under constitutional review by requiring "intent" to discriminate. But there are other threats. As we have seen, all the stages of bringing a succesful claim of indirect discrimination are vulnerable in one way or another. Much will depend, then, on how the courts will handle the concept. For even if you have a good four wheel drive, evading the potholes and rocks requires a good and steadfast driver as well.
[∗] Prof.dr. M L P Loenen is a Professor of Women's Legal Studies, Faculty of
Law, Utrecht University.
 ETC, Opinion 1-90-10, Oordelen Commissie Gelijke Behandeling van mannen en vrouwen bij de arbeid 1989-1990 (Opinions of the Equal Treatment Commission regarding men and women in employment 1989-1990), Den Haag 1991; ECJ 19 December 1994, Helmig (cases C-399/92 and others), European Court Reports (1994) p 5727.
 For a large selection of contributions dealing with non-discrimination law in these and other countries see Loenen T and Rodrigues P R (eds) Non-discrimination law: comparative perspectives (Martinus Nijhoff Publishers, London, 1999).
 Australian Law Reform Commission, Equality before the law: Justice for Women, Report No 69 (1994) 1.
 Above, note 3, para 3.23 and 3.24.
 See Pritchard S ‘Approaching its use-by date? National enforcement mecanisms: the case of Australia’ in Loenen and Rodrigues (eds), above, note 2, at 371.
 Griggs v. Duke Power Co,  USSC 46; 401 US 424 (1971).
 See US Supreme Court, Dothard v Rawlinson,  USSC 144; 433 US 321 (1977).
 For an overview of sex discrimination law in the EU see, Ellis E EC Sex EqualityLaw (Clarendon Press, Oxford, 1998).
 ECJ Sotgiu (case 152/73), European Court Reports 1974, p 153.
 ECJ 31 March 1981, Jenkins (case 96/80), European Court Reports 1983, p 911.
 ECJ 13 May 1986, Bilka (case 170/84), European Court Reports 1986, p 1607. However, consistently formulating a test to be applied is not the same as consistently applying the test. I will come back to this point in paragraph 4.2.
 Bilka, above, note 10, para 31.
 Bilka, above, note 10, para 34.
 For an overview of social security cases where indirect discrimination by the state regularly figures see, for example, Steiner J ‘The Principle of Equal Treatment for Men and Women in Social Security’ in Hervey T and O'Keefe D (eds) Sex Equality Law in the European Union (Wiley, Chichester, 1996) pp 111-136.
 This seems mainly due to the much larger amount and variety of cases reaching the ETC than the ECJ, though the Helmig case mentioned in the introduction and other cases suggest that the ECJ is less willing to apply the concept of indirect discrimination as favourably for women as several years ago.
 In this case the ECJ held that the connection between seniority and experience must be established for every individual case and may not be taken for granted as a valid generalisation. Seniority does not imply by definition a better job performance and thus cannot be the basis for a higher salary just in itself. As linking conditions of labour like pay to seniority is likely to have a disparate impact on women, this may amount to indirect discrimination on the basis of sex prohibited by article 119 of the EC Treaty. ECJ 7 February 1991, Nimz (case 184-89), European Court Reports 1991, p 2591.
 All opinions of the ETC are published by year in its Oordelenbundel. See opinions 98-27, 98-28, 98-102, 98-138 and 98-142, Oordelenbundel 1998. They are also available on the internet: <http://www.cgb.nl> .
 See, for example, opinions 97-71 and 97-139, Oordelenbundel 1997; Opinions 98-16 and 98-22, Oordelenbundel 1998. See for the issue of costs being an insufficient justification in themselves also ECJ 17 June 1998, Hill/Stapleton (case C-243/95), Rechtspraak Nemesis, nr. 976 (as yet unpublished in the European Court Reports, but available through <http://www.curia.eu.int> ).
 Opinion 98-55, Oordelenbundel 1998.
 Respectively opinion 95-19, Oordelenbundel 1994-1995, opinion 96-117, Oordelenbundel 1996 and opinion 98-112, Oordelenbundel 1998.
 Opinion 331-91-40, Oordelenbundel 1991.
 Opinion 98-56, Oordelenbundel 1998. Recently legislation has come into force in the Netherlands which gives workers a right to reduce their working hours, unless the employer can advance sufficient economic reasons for refusing it.
 Opinion 98-35, Oordelenbundel 1998, opinion 98-128, Oordelenbundel 1998, and opinion 96-29, Oordelenbundel 1996.
 Opinion 96-85, Oordelenbundel 1996, opinion 96-109, Oordelenbundel 1996; opinion 97-24, Oordelenbundel 1997.
 Opinion 99-18, Oordelenbundel 1999.
 Opinion 99-49, Oordelenbundel 1999.
 Formal refugees, for instance, are issued a temporary permit only, yet they have a right to permanent residence as long as they have the status of refugee.
 See, for example, opinion 95-69, Oordelenbundel 1995 and opinion 97-117, Oordelenbundel 1997.
 See more extensively Loenen T ‘Rethinking sex equality as a human right’ (1994) Netherlands Quarterly on Human Rights p.253 and Loenen T ‘The Equality Clause in the South African Constitution: Some Remarks from a Comparative Perspective’ (1997) South African Journal on Human Rights p.401.
 Washington v Davis,  USSC 107; 426 US 229 (1976), at 248.
 Washington v Davis, above, note 26, at 240. See more extensively: Note, ‘Making the Violation Fit the Remedy: The Intent Standard and Equal Protection Law’ (1982/83) 92 Yale Law Journal 328.
 Personell Administrator of Massachusetts v Feeney,  USSC 113; 442 US 256 (1979).
 Feeney, above, note 30, at 279.
 Compare with Rakoff T ‘Washington v Davis and the Objective Theory of Contracts’ (1994) Harvard Civil Rights - Civil Liberties Law Review 63.
 See more extensively on the problematic nature of requiring intent Binion G ‘"Intent" and Equal Protection: A Reconsideration’ Supreme Court Review 1983 (University of Chicago Press, Chicago/London, 1984) pp 397-457 and literature mentioned there.
 See, for example, More G ‘Equal Treatment of the Sexes in European Community Law’ (1993) 1 (1) Feminist Legal Studies 45, who claims the objective justification "surely destroys the meaning of a disparate impact analysis" (at 70).
 Similarly, the South African Constitutional Court takes disadvantage and not intent as its starting point. The South African Constitution explicitly covers both direct and indirect discrimination in its section 8(2). To the Court, this inclusion "evinces a concern for the consequences rather than the form of the conduct. It recognises that conduct which may appear to be neutral and non-discriminatory may nonetheless result in discrimination, and if it does, it falls within the purview of section 8(2)". Constitutional Court of South Africa 17 February 1998, The City Council of Pretoria v Walker (case CCT 8/97). All decisions of the Court can be found at <http://www.law.wits.ac.za> .
 Canadian Supreme Court, Law Society of British Columbia v. Andrews, (1989) 1 Supreme Court Reports 143.
 For an overview of the developments in the Canadian approach to equality and non-discrimination under the Charter see Vizkelety B ‘Adverse Effect Discrimination in Canada: Crossing the Rubicon From Formal to Substantive Equality’ in Loenen and Rodrigues (eds) above, note 2, p 223-236.
 Supreme Court of Canada, Eldridge v British Columbia (Attorney General),  3 Supreme Court Reports. Available from <http://www.droit.umontreal.ca> .
 For an elaboration of the problems related to applying a substantive equality approach see Loenen T ‘Substantive equality as a right to inclusion: dilemma's and limits in law’ (1995) 3 Rechtsfilosofie en Rechtstheorie 194.
 Eldridge, above, note 38, at 89.
 For a more extensive treatment of some of the problems involved see Sjerps I ‘Effects and Justifications. Or How to Establish a Prima Facie Case of Indirect Sex Discrimination’ in Loenen and Rodrigues (eds) above, note 2, pp 237-247; Browne K R ‘The Use and Abuse of Statistical Evidence in Discrimination Cases’ in Loenen and Rodrigues (eds) above, note 2, pp 411-423. For a comparative study of US and Dutch approaches to proof of discrimination in employment see Leenders M Bewijsrecht en Discriminatie bij de Arbeid (Proof of Discrimination in Labour Cases) (Tjeenk Willink, Zwolle, 1997).
 ECJ 17 October 1989, Danfoss (case 109/88), European Court Reports 1989, p. 3199.
 ECJ 9 September 1999, Seymour-Smith & Perez (case C-167/97), consideration 65. Unpublished yet in the European Court Reports, but available through <http://www.curia.eu.int> .
 Raad van Beroep Roermond 8 November 1989, Rechtspraak Sociale Verzekering 1990, 160.
 Raad van Beroep Haarlem 21 August 1989, Rechtspraak Nemesis 1990, 4.
 ECJ 19 November 1992, Molenbroek (case C-226/91), European Court Reports, p 5943.
 See for an overview Stoter W A R, Rechtvaardigingsgronden voor Ongelijke Behandeling van Vrouwen en Mannen in het Gemeenschapsrecht (Justifications for Unequal Treatment of Men and Women in Community Law) Nederlands Juristen Blad (1996) pp 1811-1815.
 Wards Cove Packing Company v Atonio,  USSC 104; 490 US 642 (1989) at 659.
 Elsewhere I have used the term "sensitive" groups, see Loenen, Rethinking Sex Equality, above, note 27.
 United States Supreme Court, Adarand Constructors v Federico Pena USSC 57; , 515 US 200 (1995).
 In constitutional equality analysis, the Supreme Court uses three levels of scrutiny. Strict scrutiny is the heaviest one and is applied in cases of race discrimination. It requires the classification to be "necessary" (or "narrowly tailored") to achieve a "compelling state interest". Intermediate scrutiny is applied in e.g. sex discrimination cases and requires the classification to be "substantially related" to an "important" government objective. Classifications which are not in any respect (semi)suspect only have to pass a very light test: such classifications should not be unreasonable, that is they should be "fairly related to the object of regulation". See for an overview one of the many casebooks on American constitutional law like J.E. Nowak and R.B. Rotunda, Constitutional Law, West Publishing Co, St. Paul, Minnesota 1995 (fifth edition), p. 595.
 Adarand Constructors, supra note 50, dissenting opinion of Justice Stevens.
 ECJ 17 October 1995, Kalanke (case C-450/93), European Court Reports 1995, p. 3051. Cf. S. Moore, Comment, European Law Review 1996, p. 158-161. ECJ 11 November 1997, Marschall (case C-409/95), available through www.curia.eu.int
 ECJ 28 March 2000, Badeck (case C-158/97). Availabble through www.curia.eu.int
 See K. O'Regan, Addressing the legacy of the past: Equality in the South African Constitution, in: Loenen and Rodrigues (eds), Non-discrimination law, supra note 2, p.13-23 and Vizkelety, Adverse effect discrimination in Canada, supra note 37.
 Walker, supra note 35 at 45, citing from the decision in President of the Republic of South Africa v Hugo, CCT 11/96. The terminology used in the South African Constitutional is peculiar in that it forbids unfair discrimination.
 Besides, the victory of Walker did not help him much: though it was granted the City Council of Pretoria had discriminated indirectly by being selective in suing residents of white suburbs for not paying their service charges and leaving township dwellers who failed to pay in peace, the other complaints of discrimination were not acknowledged, so Walker still had to pay his higher rates.
 Within a symmetrical approach, this is entirely consistent, see B.P. Sloot, Positieve discriminatie. Maatschappelijke ongelijkheid en rechtsontwikkeling in de Verenigde Staten en in Nederland (Positive discrimination. Social inequality and legal development in the United States and the Netherlands), Tjeenk Willink, Zwolle 1986, p. 216-219.
 Supreme Court, Wards Cove Packing Company v Atonio USSC 104; , 490 US 642 (1989).
 Id. at 662.
 Danfoss, see supra note 40. See also ECJ 30 March 2000, Jämställdhetsombudsmannen (case C-236/98), consideration 53: “where there is a prima facie case of discrimination, it is for the employer to show that there are objective reasons for the difference in pay”.
 Equality before the law, supra note 3, paragraph 3.24.
 Id. recommendation 3.2.