Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
The Economics of Comparable Worm .. Analytical, Empincal, and Policy Questions Mark R. Killingsworth In this paper I present an economic analysis of the conceptual foundations of comparable worm and of its likely economic consequences, focusing on comparable worm as a remedy for sex discrimination. The first section defines terms and describes the basic nature of comparable worth policy. The second section discusses the basic features of discriminatory labor markets, Hereby describing the context within which proposed comparable worth policies would operate. The third section presents an economic analy- sis ofthe basic concepts underlying comparable worth, including in particu- lar the notion that unequal pay for jobs of comparable worth is discriminatory. The final section examines the likely economic conse- quences of adopting and enforcing a comparable worth policy and contrasts comparable worth with alternative antidiscrimination policies. WHAT IS COMPARABLE WORTH? The first order of business in any discussion of comparable worth policy is to define clearly just what such a policy would entail. Three topics require special attention: comparability, coverage, and compliance. 1 Technical details and foal demonstrations of the propositions developed heuristically in the text appear in a companion paper (M. Killingsworth, 1984b). For additional discussion, see my testimony before the Joint Economic Committee (M. Killingsworth, 1984a) and a related paper (M. Killingsworth, 1985). 86
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS Comparability 87 Many discussions of comparable worth (see Treiman and Hartmann, 1981 :Ch. 4) provide a reasonably clear statement ofthe general nature ofthe standards that would be used to determine comparability: two jobs would be deemed comparable if they are found to require the same skill, effort, and responsibility and to involve the same working conditions. For example, to gauge skill requirements one might analyze the educational attainment and training of the persons in the two jobs in question; likewise, one might use formal job evaluations to derive measures of the effort and responsibility requirements and working conditions of the two jobs. The worth of any given job would then be computed as a weighted sum of the scores it receives for its working conditions and its skill, effort, and responsibility require- ments. Two jobs would be deemed comparable if one job's worth, calculated in the manner just described, is the same as (or within a few points of) the otherjob's worth. In a superficial sense, the job evaluations that would be used in determin- ing the worth of different jobs would therefore be similar to those currently used by employers. However, there is one crucial difference between the kind of job evaluation that would be used with comparable worth policies and the kind of job evaluation typically used by employers: the latter are typically based explicitly on market considerations. For example, commer- cial job evaluation finns often benchmark wages for key jobs on the basis of labor market surveys and use procedures such as regression analysis of existing salary structures to determine the weights that the marketplace itself gives to the different factors considered (see Abowd, 1984; Schwab, 19801. In contrast, some analysts of comparable worth (e.g., Treiman and Hart- mann, 1981:81-82) question such a market-oriented approach on the grounds that the wage relationships that currently exist are likely to be distorted by discrimination. Some comparable worth proponents advocate the use of bias-free job evaluations, i.e., ones that are derived independently of the existing wage structure and in which the weights given to the different factors considered (skill, effort, responsibility, and working conditions) would be determined on an a priori or to put it less charitably, ad hoc basis. Implementing any procedure of this kind is likely to lead to lengthy and perhaps even bitter arguments over specifics (e.g., about the relative magni- tudes that should be used in weighting the skill requirement and working condition components of any job's total worth). These are important ques- tions of detail. However, a clear statement of the general principles that would be followed in determining comparability is sufficient for purposes of this analysis.
88 KILLINGSWORTH Coverage Discussions of comparable worth usually do not specify clearly which jobs would be covered under a comparable worth requirement. In principle, the notion of comparable worth might be applied to literally all jobs, in which case coverage would be universal. In practice, however, most discus- sions of comparable worth present it as a remedy for sex discrimination. This suggests a simpler, partial rule: two jobs that are found to be comparable (in the sense defined above) would be required to pay the same wage only if women are a greater proportion of the workers in the low-paying job than in the high-paying job. (For example, see Blumrosen, 1979:496-497; Treiman and Hartmann, 1981:82-90; Yale Law Journal, 1981:677-678.) Other questions about coverage have also been largely neglected. For example, would coverage be limited to employers with at least some speci- fied number of employees, as under provisions of fair labor standards laws? Would the same comparable worth standard be applied to all establishments of a given employer, regardless of geographic location or industrial classifi- cation? Such details have not yet been discussed systematically. One coverage question, however, does have an explicit answer. Contrary to the complaints of many critics who have charged that comparable worth would amount to government wage fixing on a national basis, most advo- cates of comparable worth have emphasized that comparable worth would be implemented on an employer-by-employer (or even an establishment-by- establishment) basis. Thus, although comparable worth might require a given employer to pay tool mechanics and secretaries the same wage (so long as the two jobs were found to be comparable at the firm in question), it would not establish a uniform national wage for secretaries (or tool mechan- ics) and would not necessarily require even that any other employer pay identical wages to tool mechanics and secretaries. That would depend on whether, at any other such firm, the jobs of tool mechanic and secretary were found to be comparable. Compliance Finally, most discussions of comparable worth are silent about compli- ance procedures, i.e., about how wages would be adjusted if two jobs covered by the comparable worth rule and deemed to be comparable are found to pay different wages. In principle, compliance could be achieved in an infinite variety of ways: the wage of the low-paying job could be raised to equal that of the high-paying job; or the wage of the high-paying job could be reduced to equal that of the low-paying job; or one could split the difference, raising the wage of the low-paying job and reducing the wage of the high- paying job until they are equal; and so forth. In practice, however, discus- sions of comparable worth that offer specifics on this question usually opt for
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 89 wage increases in the low-paying job as either the only, or else the preferred, means of compliance. Having defined the terms, I now proceed to substantive issues. I begin by describing the nature of discriminatory labor markets, by examining the context in which comparable worth strategies would operate. I then present an economic analysis of the conceptual foundations of comparable worth and of its likely economic effects. EMPLOYER DISCR~NATION: ANALYTICAL AND EMPIRICAL QUESTIONS In order to understand the essential questions about discrimination under- lying the comparable worth issue, it is useful to start by considering an imaginary economy with just two jobs, A and B. Employers' demands for workers to fill these two jobs depend on technology and on the demand for output. If the demand for output falls, employers' derived demands for workers in each job will likewise fall. If labor costs increase, the price of output will have to rise, thereby reducing the demand for output and, thus, in a roundabout way, reducing firs' derived demands for workers as well. Finally, if the wage in one job rises relative to the wage in the otherjob, then, even if the demand for output remains unchanged, employers will try to substitute away from the job whose wage has risen and toward the job whose wage has remained unchanged. Of course, particularly if the two jobs are quite different, the scope for such substitution may be quite limited.2 (Much of the discussion below assumes that the scope for such substitution is small, though perhaps not completely nonexistent.) In this imaginary economy, each individual worker is equally able to do eitherjob equally well, but individuals are not all equally willing to do either job. In general, the supply of labor to the two jobs depends on both relative wages and individuals' preferences. At given wages, some individuals would prefer to do job A whereas others would prefer to do B. Indeed, persons with a particularly strong preference for B might want to do it even if the A wage was much higher than the B wage. Other individuals those whose preference for B is weakercould be induced to shift from B to A if the wage in A were to increase relative to the wage in B.3 Thus, the higher the 2 For example, it may be quite difficult to substitute the kind of work done by clerical workers for the kind of work done by craft workers (or vice versa) to any substantial degree. Likewise it may be quite difficult to substitute the kind of work done by nurses for the kind of work done by tree trimmers (or vice versa) to any substantial degree. 3 Note that any shifting from B to A (or vice versa) that is stimulated by changes in relative wages might take a considerable time to work itself out; nothing in the discussion in the text assumes that such shifting would occur immediately.
go KIl :LINGSWORTH A wage relative to the B wage, the greater the total supply of lab,orto job A but even at very high relative wages, some workers would still prefer (want to work in) job B. The actual relative wage (the wage in A relative to the wage in B) of course depends on demand as well as on supply. In general, there is no reason to expect that the two jobs would pay the same wage. For example, suppose that most workers regard job B as less attractive than job A and that, at the moment, both A and B pay the same wage. The number of workers who are willing to do job B may be too small to fill all the job B vacancies that employers want to fill at existing wage levels, and the number of workers who are willing to do job A may exceed the number of job A vacancies available. Under these conditions the B wage will tend to rise and the A wage will tend to fall. Eventually, B will pay more than A. In this example, there is of course a pay differential one job pays more than the other but there need not be any sex differentials. If employers do not favor men over women in employment in either of the jobs, and if men's and women's preferences between the two jobs are not systematically differ- ent, then (1) the relative representation of women in the better-paid job, B. will be the same as the relative representation of women in the lower-paid job, A; and (2) men and women in the same job will receive the same pay. Next, however, suppose that employers discriminate in favor of men in the sense that, at given wages, they would prefer to employ a man rather than a woman in the better-paidjob B. In a situation of this kind, employers act as if men in job B produced not only actual output steel, trimmed trees, and so on but also provided an intangible that may be called maleness. This being the case, women will tend to be excluded fromjob B. and the women who are in job B will be paid less than the men in job B. Relative to a nondiscrimina- tory environment, employers' demands for male workers in job B are higher and their demands for female workers for job B are lower. Employment of women in job B is therefore lower and employment of men in job B is therefore higher than they would be in a nondiscriminatory world. The women who remain in job B will receive a lower wage than the men in job B (because the women do not produce maleness and so are less valuable to discriminately employers than are men). Thus, there are now both pay differentials and sex differentials: 1. The women who are able to remain in job B are able to do so only because their wage falls below that of men in job B. meaning that the wage for women falls relative to the wage for men among B workers. 2. Discrimination in favor of men in job B acts to raise the B wage relative to the A wage among men. 3. The (relative) decline in the wage for women in job B. noted in item 1 above, means that the B wage falls relative to the A wage among women.
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 91 4. Women are now overrepresented in the low-paying job A and under- represented in the high-paying job B. In evaluating these propositions it is important to clear away some of the confusion that plagues many discussions of discrimination. First, although it is often believed that discrimination is profitable for employers, discrimina- tion will in fact reduce profits in the kind of labor market considered here. This is because a firm that employs a more expensive male worker in job B in preference to an equally productive but less expensive female worker is incurring greater costs than the minimum required to achieve a given level of production. (For elaboration of this point, see Ashenfelter and Pencavel, 1976.) In other words, if men and women are equally productive (as I have assumed) but nevertheless receive different wages in the same job, then an employer who is interested only in pecuniary profits will want his job B work force to consist entirely of women: otherwise, costs will be higher than the minimum possible. For example, when asked recently why most of the senior staff of his consulting firm is female, economist Alan Greenspan replied: "I always valued men and women equally, and I found that because others did not, good women economists were cheaper than men. Hiring women does two things: It gives us better qualify work forless money, and it raises the market value of women."4 Ironically, some commentators stand this logic on its head and insist that employers cannot ignore "market realities." In their view, if employer X wants to hire a man, then employer X will have no choice but to match or exceed the offers made to that man by employers Y. Z. etc. which quite possibly will exceed the offers made to equally productive women. (For a recent example of this argument, see Bodner, 1983.) Unfortunately, this appeal to market realities reveals a somewhat limited understanding ofthem: Why is it necessary to hire a higher-paid man, instead of a lower-paid but equally productive woman? For example, if the chairman of an academic department at a university hires or retains higher-salaried men in preference to lower-salaried but comparably productive women on the grounds that market realities have forced him or her to offer higher salaries to the men, then he or she is raising the department's costs but not its quality. The appropriate solution is that of Greenspan: to avoid hiring or retaining a higher-paid man, wheneverit is possible to employ a comparably productive woman forless. Finally, despite widespread assertions to the contrary, there is in fact no reason to believe that competition in the marketplace will eventually drive 4 See L`ewin ( ~ 983), who, oddly enough, refers to Greenspan's employment practices as " gender bias."
92 KILLINGSWORTH discriminatory employers out of business even though, as just noted, such employers earn lower-than-normal profits. s As Goldberg (1982) has pointed out, even though discriminatory firms earn lower-than-normal pecuniary profits, they derive a kind of "psychic income," to the extent that they employ members of the group they prefer (e.g., men). Thus their total profits (meaning both pecuniary and psychic rewards for ownership) are more than sufficient to make them want to stay in business. Indeed, the monetary equivalent of the total profits of a discriminatory employer will exceed the pecuniary profits that could be earned by a nondiscriminatory employer using the same amount of capital. Thus, the greatest pecuniary payment that a nondiscriminatory employer would be willing to offer to take over the discriminatory employer's business is less than the minimum sum that the discriminatory employer would be willing to accept to sell the business. By the same token, because discriminatory firms earn lower pecu- niaIy profits than nondiscriminatory firms, the maximum pecuniary pay- ment that a discriminatory employer would be able to offer to take over the business of a nondiscriminatory employer is below the (pecuniary) profits earned by the nondiscriminatory employer. Thus, the nondiscriminatory employer would be unwilling to sell out to the discriminatory employer. In sum, even in a competitive economy, discriminatory and nondiscriminatory firms can coexist, even in the long run (see Goldberg, 1982, for details). With these caveats in mind, now consider the four characteristics of discriminatory labor markets presented above. These are simply predictions about how the world would look in the presence of employer discrimina- tion predictions that, moreover, are derived from a very simple model of the economy. Is there any evidence to suggest that the real-world economy behaves in this way? The available empirical evidence is certainly not conclusive, but it is suggestive. First, a number of studies have shown that there are indeed male-female pay differences within given jobs or job categories, as pre- dicted by item 1 above, even when other things (e.g. personal, human capital, and job characteristics) are held constant (for examples of such studies, see Ashenfelter and Pencavel, 1976; O'Neill, 1983; Roos, 19811. Second, several empirical studies have found a strong negative association, other things being equal, between female representation and average pay (of men and women combined) within occupations, as predicted by item 4 above (see Treiman and Hartmann, l981:Ch. 2, for a review of these stud- 5 The notion that competition and discrimination are incompatible in the long run is dear to the hearts of both reactionaries (who believe that the economy is competitive and do not believe in the existence of discrimination) and radicals (who do not believe that the economy is competitive and do believe in the existence of discrimination).
AlVALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 93 ies). Finally, evidence in a number of studies suggests that, other things being equal, the reduction in pay associated with being in a mostly female occupation rather than in a mostly male occupation is greater for men than for women, as implied by items 2 and 3 above (for example, see O'Neill, 1983; Roos, 19811. Inevitably, such studies are subject to methodological criticism, and fur- therresearch on these issues is eminently desirable. (For discussion of some of the methodological defects in studies of this kind and suggestions for further research, see Bloom and Killingsworth, 1982.) Despite these cave- ats, however, it is noteworthy that the available evidence seems at least generally consistent with all four of the predictions derived from the model of discrimination presented here. THE CONCEPTUAL BASIS OF COMPARABLE WORTH: AN ECONOMIC ANALYSIS From the standpoint of economic analysis, the concept of comparable worth has two fundamental flaws. First, contrary to what many of its propo- nents assume, there is nothing inherently discriminator in unequal pay for jobs of "comparable worth," and there is no reason why a nondiscriminatory labor market would necessarily entail equal pay for jobs of "comparable worth." Second, although proponents of comparable worth are correct in implicating employer discrimination as an important demand-side factor responsible for male-female pay differentials, they typically pay insufficient attention to the role of employee choices, which is an important supply-side factor that also contributes to such pay differentials . Should Comparable Worth Necessarily Mean Equal Pay? The first difficulty with comparable worth is its central premise: that unequal pay for jobs deemed to be of comparable worth is inherently dis- criminato~y or, equivalently, that jobs deemed to be of comparable worth should receive the same wage. Unfortunately, this notion betrays a funda- mental misunderstanding of the way in which labor markets (even nondis- criminatory ones) operate and of how employer discrimination harms women. Unequal pay for jobs of comparable worth is not inherently dis- criminatory. Even in a sex-neutral labor market one in which sex is entirely irrelevant to market outcomes there is no reason why individuals in jobs of comparable worth should necessarily receive the same wage. The basic reason for this is simple. Individual tastes and preferences differ; comparability is in the eye of the beholder. Suppose that a given individual considers jobs A and B to be equally acceptable- i.e., would have no preference for one over the other if both jobs paid the same wage. Is there
94 KI=INGSWORTH any reason to suppose that all other individuals would feel the same way? Would it be at all surprising if, at given wages, at least some individuals preferred A to B. while at the same time still other persons preferred B to A? Obviously not. Thus, even if the two jobs are found to be comparable according to a formal job evaluation scheme, there is no reason to suppose that all individuals will in fact view them as comparable. There is likewise no reason to suppose that supplies and demands for the two jobs would be equal if the two jobs paid the same wage. Hence, there is no reason to suppose that jobs that are comparable in the eyes of a given individual orjob evaluation firm would in fact pay the same wage even in a sex-neutral labor market. The proponents and opponents of comparable worth have spent much time pondering whether it is possible to compare apples and oranges; the debate on this question, however, misses the essential point. On one hand, it is of course possible to compare apples and oranges (and secretaries and tool mechanics): we do it all the time, using the measuring rod of money. On the other hand, unless all individuals have exactly the same preferences, there is no substance to the notion that a nutritional evaluation of apples and oranges, or a job evaluation of secretarial and mechanic work, will have anything useful to say about the prices that these quantities could or should command in the marketplace. For example, suppose apples and oranges do indeed have exactly the same nutritional content, and suppose further that all individuals are color-blind. Does it follow that apples and oranges will sell for the same price? Not unless all individuals regard apples and oranges as identical. Otherwise if tastes are heterogeneous, so that some individuals like the taste of apples much better than that of oranges the prices of apples and oranges will depend on both supply (e.g., production technology in the apple and orange industnes) and demand factors (including the distribution or variation in individual tastes). In this case, it will be impossible to predict, a pnori, whether apples will fetch a higher price than oranges (or vice versa), even if all individuals are color-blind (so that no one's behavior is influenced by the fact that oranges are usually orange-colored). To many this point will be obvious.6 Of course, individual tastes and 6 But not, however, to Bergmann (in this volume). In her twojob model of the labor market, literally all suppliers of labor have identical preferences. That is, Bergmann tacitly assumes that, at equal wage rates, no one would prefer one job over another. This is the sole case in which one would indeed expect to observe equal pay for jobs of comparable worth in the absence of discrimination. However, in analytical terms it is a rather special case, offering little insight into the operation of real-world labor markets. In contrast, the model in this paper allows explicitly for the heterogeneity of testes (see M. Killingsworth, 1984b, for details).
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 95 preferences differ, so perhaps some illustration may be worthwhile. Follow- ing is one example (from Sharon Smith: see Gold, 1983:43-441: an employer asks us to evaluate the comparability of the jobs of Spanish- English translator and French-English translator. A priori, it would seem difficult to argue that either of these two jobs requires more skill, effort, or responsibility than the other, and it would be surprising if, at a given firm, the working conditions for the two jobs were appreciably different. Presumably, then, most job evaluation schemes would conclude that these two jobs are comparable in terms of their working conditions and skill, effort, and responsibility requirements; and so, presumably, a comparable worth policy would require an employer to pay the same wage to persons in each of the two jobs. But would pay for these two jobs necessarily be equal, even in a sex- neutral labor Harlot? Perhaps, but now suppose we learn that the employer in question is located in Miami. Would it be reasonable to expect that Spanish-English translators in Miami would get the same pay as French- English translators in Miami? Almost certainly not. Would it even be possi- ble to predict which job would receive the higher wage? Again, almost certainly not. Tn~e, one would expect that in Miami the supply of qualified Spanish-English translators would be greater than the supply of qualified French-English translators. Other things being equal, that would mean that the latterjob would pay more then the former. However, other things are not necessarily equal, even in a sex-neutral labor market. In particular, Miami's demand for Spanish-English translators might well be greater than its demand for French-English translators. Other things being equal, that would mean that the latterjob would pay less than the former.7 As this example indicates, even in a labor market in which sex is irrele- vant, wages are determined by market supplies and demands rather than by comparability which may have little or nothing to do with wage deterrnina- tion. The reason is that supplies and demands summarize the tastes and preferences of all individuals, whereas comparability merely indicates the tastes and preferences of one individual or of a single entity (e.g., a job evaluation firm). Whether the individual who determines comparability is I 'homme moyen sensuel (or lafemme moyenne sensuelle) is beside the point: 7 A difficulty with the translators example is that one must assume that the two translation jobs would in fact be found comparable. Although plausible, this assumption is still only an assumption. However, the real world provides illustrations of the same point. For example, Gold (1983:48~9) reports the results of a 1976 job evaluation undertaken for the state of Washington. The job of park ranger received 181 evaluation points, whereas the job of homemaker received 182 evaluation points. Is there any reason to believe that these two supposedly comparable jobs would necessarily receive the same pay, even if sex were altogether irrelevant to all labor market outcomes?
96 All If INGSWORTH most distributions have observations on either side of the median, and mar- ket demands and supplies depend on those observations as well as on the median. Note that this result holds whether or not demand is affected by employer discrimination: it is just as pertinent in a discriminatory as in a nondiscriminatory labor market. In sum, comparable jobs need not receive the same wage. Similarly, there is no reason to believe that noncomparable jobs would necessarily receive different wages. An example of this derives from Adam Smith (1776~. In discussing wage differentials among jobs, Adam Smith expressed his view that the butcher's trade is a "brutal and odious business" and noted that it paid more than other trades with similar skill requirements. As modern writers have observed, however, this may have been due to the cultural homogeneity of eighteenth-centu~y Britain (i.e., essentially identical prefer- ences for all individuals). A pay differential between butchering and other trades need not arise, despite the supposedly brutal and odious nature of butchers' work, if preferences are heterogeneous. As Rees (1976:340) has pointed out, if enough individuals have no strong feelings about or actually enjoy butchering, "it would then clearly be possible to fill all positions for butchers without any compensating wage differential." Similarly, consider police work. Police work is generally regarded as arduous and dangerous, and it would hardly be surprising if, in recognition of this, a job evaluation were to award more job evaluation points to police work then to clerical work. Under a comparable worth standard, then, police work might well be entitled to a higher rate of pay than many kinds of clerical work. But would such a pay differential really be necessary? If enough individuals think of police work as exciting and regard clerical work as dull, it would be possible to fill all police jobs without paying any wage premium at all. (Nor is this merely a hypothetical possibility: consider the lengthy waiting lists of qualified applicants seeking positions in many municipal police departments around the country.) In such a case comparable worth would serve to justify and protect the practice of paying premium wages for police worka job still held mostly by men despite the absence of any rational basis for such a differential. The reasoning in the police and butchers examples is essentially the same as the reasoning in the translators case: statements about the comparability or noncomparability of jobs are simply expressions of a given individual's or job evaluation firm's feelings about those jobs; wages, however, depend on supplies and demands, i.e., on all individuals' tastes and preferences. Job evaluation points will accurately reflect all individuals' tastes and prefer- ences among jobs only if all individuals have identical tastes and prefer- ences. In that case, a priori (or bias-free) job evaluations of the kind envisaged by advocates of comparable worth will indeed provide a useful guide to wage differentials among jobs. Otherwise, however, such evalua-
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 97 lions may provide little or no meaningful information about market supplies to or wage differentials amongjobs.8 When tastes are heterogeneous, jobs do not have any intrinsic absolute worth; indeed, they do not even have any intrinsic relative worm. According to one proponent of comparable worth, "The comparable worth strategy seeks to adjust market wages, via job evaluation, to where they would be in the absence of discrimination" (Hart- mann, 1984:211. Although that may well be the objective of comparable worm advocates, it is simply wishful thinking to imagine that job evaluation can in fact determine what the wage structure would be in the absence of discrimination. What If Labor Supply Is Sex-Related? The arguments just presented are sufficient to dispose of a strict compara- ble worth policy of requiring equal pay for comparable jobs. However, it does not dispose of the basic problem sex discrimination that is a major motivation for most proposals for comparable worth. In particular, although there is no reason why comparable jobs would pay the same wages if sex were irrelevant to labor market outcomes, there is also no reason why a sex- neutral labor market would exhibit the negative empirical association, noted at the outset, between overall average pay and the proportion female in different jobs. Indeed, if the labor market were truly sex-neutral, then, even though one would not expect to find equal pay for jobs deemed to be of comparable worth, one would expect to find the same male-female employ- ment ratio in all jobs, other things (skills and so forth) being equal. 8 The skeptic may ask: If the usefulness of job evaluations is limited or nonexistent, why do firms undertake them? One important reason has to do with a basic difference (noted above) between the bias-free job evaluations advocated by comparable worth proponents and the market-oriented evaluations typically used by firms. Bias-free evaluations would be undertaken on an a priori basis without reference to marketplace supplies and demands. In contrast, the market-oriented job evalua- tions usually used by employers are in general based squarely on market considerations (e.g., benchmarking wages for key jobs on the basis of wage surveys). As such, market-oriented job evaluations may be regarded as attempts to gauge the shortages or surpluses that would develop if wages were set only on the basis of skill, effort, responsibility, and working conditions (as in a bias- free job evaluation), and to adjust wages accordingly. It should be noted that there is nothing inherently sacred about the concept of wage determination by forces of supply and demand. (For example, as emphasized above, employer discriminationa force that affects demandwill cer- tainly distort wages and wage differentials among jobs.) However, establishing the existence of employer discrimination does not establish the validity of the concept of comparable worth, and establishing the existence of unequal pay forjobs of comparable worth is not sufficient to establish the existence of employer discrimination. Indeed, as noted earlier in this section, there is no reason to expect equal pay for jobs of comparable worth even in a completely nondiscriminatory labor market, which is presumably the ideal or reference point that most comparable worth advocates would adopt.
98 KIl :LINGSWORTH Thus, even though job comparability and a strict comparable worth policy have no basis in analysis or logic, one might still argue for a modified comparable worth policy. Under such a policy, which I will call COO-1 for short, employers would be required to pay the same wage to workers in two jobs that are found to be comparable if (but only if) women are a greater proportion of the workers in the lower-paying job than in the higher-paying job (for example, see Blumrosen, 19791. More generally, under COO-1, the concept of comparable worth might be redefined as requiring only that wages in different jobs be equal to what they would be if gender were truly irrelevant to labor market behavior. Thus, COO-1 would not necessarily require equal pay for jobs of comparable worth (because it would leave undisturbed any pay differences among comparable jobs to the extent that such differences would occur in a sex-neutral labor market). It would, however, be concerned with sex-related pay differences among jobs which are a major motivation for much of the discussion of comparable worth. It might be quite difficult, in practice, to determine the pay rates in different jobs that would obtain if sex as such were irrelevant to labor market out- comes. Leaving such questions of practicality aside, however, what can one say about the merits of such a modified comparable worth policy? Like a strict policy of requiring equal pay for jobs of comparable worth, COO-1 is open to serious question. Contrary to what many proponents of comparable worth appear to assume, the wage structure that would prevail if sex were irrelevant to all labor market activity is not necessarily the same as the wage structure that would prevail in the absence of employer discrimina- tion. In the real-world labor market, decisions of workers (labor supply) as well as decisions of employers (labor demand) are sex-related. Even if sex did not affect labor demand (i.e., even if employers did not discriminate), sex differences in labor supply decisions in preferences among jobs would still lead to pay differences among jobs, even jobs of comparable worth. To see this, return to the simple twojob economy described above. In that discussion I implicitly assumed that there was no systematic sex difference in job preferences i.e., that at any given set of wage rates the proportion of women who desire to do either job is the same as the proportion of men. In that case, in the absence of employer discrimination, both jobs would have the same sex composition: the mare: female employment ratios in both jobs would be the same. Now suppose instead, however, that women preferjob A to a greater extent than do men in the sense that, at any given set of wage rates, the proportion of women who desire to do job A exceeds the propor- tion of men. In this case the total supply of labor to job A will be greater and the total supply of labor to job B smaller than if women's preferences for A were the same (on average) as those of men. Relative to a setting in which
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS . 99 women's average job preferences are the same as those of men, the wage in job A relative to that in job B will therefore be lower; and women will be overrepresented in job A (i.e., the mare: female employment ratio will be higher in job B than in job A). Some proponents of comparable worth recognize that overrepresentation of women in some jobs and wage differentials among jobs may to some extent be a consequence of women's own choices rather than of employer discrimination. The National Research Council (NRC) report on compara- ble worth (Treiman and Hartmann, 1981:53) which, it should be noted, endorsed the general concept provides an admirable summary of many of the reasons why women may be overrepresented in low-wage jobs: First, women may be socialized to believe that some types of jobs are appropriate and that others are inappropriate for women; socialization may be so effective for some women that it never even occurs to them to consider other types of jobs. Second, women may have pursued courses of study they thought particularly appropriate to women and in consequence may not have the education or training that would suit them for other available jobs. Third, women may lack information about other available jobs, their pay rates, working conditions, and access to them. Fourth, women may be aware of alterna- tives, but because of actual or expected family obligations may structure their labor force participation in particular ways [that reduce their earnings possibilities]. For example, they may be unwilling to invest a great deal of time, effort, or money in preparing forjobs because they do not expect to remain in the labor force after marriage or after childbear- ing.... Fifth, women may be aware of alternative types of jobs but believe them to be unavailable or unpleasant because of discrimination; their labor market preparation and behavior may be affected in many ways by this perception: the course of study they take; the time, money, and effort invested in training; their willingness to accept promotion, etc. Many of these factors might well be cited as forms of societal discrimina- tion; only one of them can readily be characterized as a form of clear-cut employer discrimination. Perhaps most important, factors such as these suggest that choices made prior to labor market entry, as distinct from employer discrimination, can have an important effect on sex differences in labor market outcomes. For example, Polachek (1978) investigated the college majors of young men and women. As one might expect, his data showed considerable gender differences in college majors: business and engineering were particularly popular among the young men; education, home economics, and medically related fields such as nursing were particu- larly popular among the young women. What is striking and rather unex- pected is that these differences persist even after statistical adjustment for such factors as Scholastic Aptitude Test (SAT) verbal and mathematical scores and a set of attitudinal variables (measuring such things as individ- uals' assessments of the lifetime importance, to them, of steady work,
100 KILLINGS WORTH money, and friends). That is, young men and young women tend to choose different college majors even when other things (abilities and attitudes) are the same. A second example of the same idea derives from Lemons v. City and County of Denver (620 F.2d 228 [lOth Cir.] cert. denied, 449 U.S. 888 [19801), a comparable worth case in which nurses maintained that their predominantly female job paid less than predominantly male jobs (e.g., gardener-florist or tree trimmer) that were allegedly comparable to the job of nurse. Although the nurses maintained that Denver was discriminating against them, they did not allege that Denver had excluded them from the comparable but higher-paid and predominantly male jobs. Indeed, asked if she had ever considered switching from nursing to tree trimming, one of the Lemons plaintiffs said: "I wouldn't work purely for money, not outside nursing. That's whatI like to do" (see V. Killingsworth, 1981:171. In the examples of choice of college major and nursing, the basic point is the same: individual choice as well as (or instead of) employer discrimina- tion can lead to a concentration of women in low-wage jobs. The conse- quences of this for comparable worth policies are rarely appreciated, however. Two implications are particularly important for evaluation of pro- posals for comparable worm. The first is that systematic differences in sex preferences rather than employer discrimination may account for at least some of the negative empirical association, noted above, between overall average pay (for both sexes combined) and the proportion female in different jobs. In other words, employer discrimination of the kind discussed above will lead to overrepre- sentation of women in low-paying jobs, but evidence of overrepresentation of women in low-paying jobs does not necessarily constitute evidence of employer discrimination. It may instead simply be evidence of systematic sex differences in job preferences, as in the example just discussed. Second, and more generally, the wage structure that would obtain in the absence of employer discrimination is not necessarily the same as the wage structure that would obtain if sex were entirely irrelevant to labor market behavior (meaning both supply and demand). The two wage structures would be identical only if sex were irrelevant to labor supply as well as labor demand. This is ignored by both a strict comparable worm policy (which would require equal pay for jobs of comparable worth) and COO-1 (which would require only that workers be paid what they would receive if sex as such were irrelevant to labor market outcomes, even if that meant wage differentials among jobs deemed to be of comparable worth). In other words, neither kind of policy recognizes the possibility that even in the absence of employer discrimination sex differences in preferences and
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 101 therefore in labor supply decisions might lead to a concentration of women in low-paid jobs (e.g., nursing or secretarial work).9 ECONOMIC CONSEQUENCES OF COMPARABLE WORTH AND OTHER REMEDIES FOR DISCRIMINATION Whatever the merits or defects of the conceptual basis for comparable worth, to many the most important question about comparable worth has to do with whether it would be a suitable remedy for discrimination. This section compares comparable worth and other antidiscrimination measures and evaluates the consequences of each. Economic Consequences of Comparable Worth To sum up the argument so far: 1. There is no reason why jobs of comparable worth would or should necessarily receive the same wage, even in a labor market in which sex is irrelevant. 2. To the extent thatjob preferences (i.e., labor supply decisions) are sex- related, it is entirely possible that one would find women concentrated in low-paying jobs even in the absence of employer discrimination (i.e., even if sex were irrelevant to labor demand decisions). 3. There is therefore little or no justification for comparable worth, at least as usually defined. When job comparability provides little or no useful information about the wage structure that would prevail in a labor market in which sex is entirely irrelevant, there is no merit in requiring equal pay for jobs of comparable worth. Moreover, when job preferences are sex-related, there is little merit in a modified comparable worth policy that would, in effect, set wages without reference to such sex differences in job prefer- ences. It should be emphasized immediately that none of this means either (1) that employers do not discriminate on the basis of sex or (2) that employer discrimination does not lead to overrepresentation of women in low-paying 9 Reasoning similar to that given in the text leads to parallel conclusions about the effect of a systematic preference forjob B on the part of men: other things being equal, the job B wage would fall relative to the job A wage, and men would be overrepresented in job B. On balance, then, women's average preference forjob A would have to be stronger then men's average preference for job B if the net effect of these sex differences were to lead to a differential favoring job B and to overrepresentation of women in job A.
102 KIT,[`lNGSWORTH jobs. Indeed, if sex differences in job preferences were the only reason why women are concentrated in low-paying jobs, one would expect to find that men and women earn the same wage within given jobs but the available evidence indicates that that is not the case. This suggests that there might still be some scope for comparable worth, albeit in still more modified form. Under this (further) modification of comparable worth policy, which I will call COO-2, employers would in principle be required to adjust pay in different jobs so as to yield the wage structure that would be observed in the absence of employer discrimination against women. Determining what this wage structure would look like would obviously be difficult. Thus, in practice, COO-2 would probably amount to something much simpler and more straightforward: a policy of requiring wage increases for jobs in which women are overrepresented (for example, see Clauss, 1981 : 85; Yale Law Journal, 1981 :680~. As implied already, the pay adjustments that would ideally be required under COO-2 would almost certainly differ from the changes that would generate the wage structure that would prevail if sex were irrelevant to both labor demand and labor supply decisions. In addition, the pay adjustments required under the ideal COO-2 policy would almost certainly not require equality of pay rates among jobs judged to be comparable in terms of a formal job evaluation. Nevertheless, COO-2 would directly confront one of the most important consequences of employer discrimination: low pay for women. Thus, COO-2 in its ideal form would certainly address a central concern of most proponents of comparable worth, even though it would - have little or nothing to do with the comparability of different jobs. Moreover, in practice, COO-2 as it would actually be implemented has appeal even for those who, on principle, would prefer COO-1. First, it could be argued that sex differences in job preferences are essentially a result of sex discrimination. True, the discrimination that induces girls to consider nurs- ing or teaching as appropriate work and to consider tree trimming or tool mechanics as inappropriate is societal discrimination rather than employer discrimination in the usual sense. However, in this view, the means through which such discrimination is practiced are of secondary importance; what is crucial is the fact that it is, after all, discrimination. Requiring employers to raise wages for low-paying, predominantly female jobs does not attack the source of this discrimination, but it does at least attack some of its effects (Gold, 1983:94-951. A second justification for simply requiring wage increases in female- dominated, low-wage jobs (CW-2, in practice) is that alternatives to such a policy are unsatisfactory. The conventional approach to employer discrimi- nation as such is to break down discriminatory barriers that keep women out of high-paid jobs and to require equal pay for equal work. Eventually such
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 103 measures will completely counteract the effects of employer discrimination: women's relative representation in high-paying jobs will rise, and sex differ- entials within the same job will be erased (for example, see Johnson and Welch, 19761. However, the full impact of such measures may not tee felt for some time to come (Yale Law Journal, 1981:6801. in Moreover, in this view, conventional antidiscrimination measures cannot realistically be expected to do much for older cohorts of women workers; these women entered low-paying jobs some years ago (either because soci- etal discrimination induced them to choose such jobs or because employer discrimination prevented them from entering other jobs), and they are now locked in. Providing access to high-paying jobs can do little for such women; they do not have the training required for such jobs (and at midca- reer it usually makes little sense to get such training), and now, if only because of their initial socialization and the development of habits over time, such women might not even want to be in other jobs, despite the higher pay. A policy of requiring wage increases for predominantly female, low-paying jobs would provide immediate benefits for such women who, in this view, have little to gain from conventional antidiscrimination measures (see Blumrosen, 1979:491~92; Clauss, 1981:81, 91-92; Treiman and Hart- mann, 1981:66-671. Unfortunately, requiring wage increases for predominantly female low- paying jobs is likely to have serious, albeit unintended, adverse side effects, not merely for women as a whole but for older cohorts of women workers in particular. To see why, return to the simple twojob economy discussed above and consider the effects of a wage increase for the predominantly female low- wage job, A, in the short run, i.e., when supplies of labor to the two jobs are essentially fixed. This case is particularly relevant to understanding the effect of comparable worth on older cohorts of women workers, who are locked into job A (in the sense that they could not easily switch to job B and might not even want to do so). The benefits of increasing the wage forjob A are obvious: the A wage will rise both in absolute terms and, of perhaps equal significance, relative to wages (of both men and women) ~ ~ in the high-paying job B. In other words, 10 Although the available evidence is certainly not conclusive, recent research suggests that Contras to what many proponents of comparable worth appear to believeconventional antidis- crimination measures have increased the female: male earnings ratio, female representation in previously predominantly male jobs, and so on (see Beller, 1978, 1980, 1982; Brown, 1982; Leonard, 1984a, 1984b). 1 1 Recall that when employers discriminate, women in job B will be paid less than men in job B. Thus, there are really two job B wages, not one.
104 KILLINGS WORTH not only will the modified comparable worth policy raise wages in low-wage jobs, it will al-so narrow the pay gap between low-wage and high-wage jobs. In addition to these benefits, however, the policy also has costs. Although they are perhaps less obvious than its benefits, they are no less significant. It is particularly important to note that comparable worth imposes costs not only on employers but also on women (and men) in low-wage, predomi- nantly female jobs- the very group the policy is intended to help. First, since the A wage rises, firms' demands for A workers will fall, leading to unemployment for some workers now in job A- who are dispro- portionately female. Second, the increase in the A wage raises labor costs and therefore pnces, so consumer demand will fall. As consumer demand falls, employers' output will contract, leading to decreases in the demand for job B (and thus to decreases in the demands for both male and female workers in job B). 12 In turn, the decline in the demand for job B will lead to unemployment and/or lower wages for both men and women initially in job B. In particular, the wage of women in job B will fall by more than the wage of men in job B. thereby widening the sex differential in pay in job B. 13 In sum, increasing the A wage in accordance with a modified comparable worth policy will obviously raise the A wage and reduce the differential between the A wage and the B wage (of either men or women). However, the policy will also have some unfortunate side effects: 1. The number of workers in job A will fall and unemployment of work- ers formerly in job A will rise. 2. Wages for both men and women in job B will fall. 3. The wage differential between men and women in job B will widen. 4. Employment and output will fall and consumer prices will rise. Thus, in the short run, some women (and men) workers in job A will gain, but everyone else all men and all women in job B and all other women (and men) in job A will lose. The notion that raising wages in low-paid, pre- dominantly female jobs will help older cohorts of women who are locked into those jobs is at best half true: such a policy will certainly benefit some of these women but, by reducing the total demand for such jobs, will necessar- ily harm the rest of them. Now consider the effects of increasing the wage in job A in the long run, 12 This decrease in demand for workers to fill job B will be offset by an increase in the demand for such workers, to the extent that it is possible to substitute job B for job A. In the nature of the case, however, the scope for such substitution is likely to be rather small; see n. 2 above. 13 In intuitive terms, this is because the comparable worth policy forces an increase in the A wage, so that the portion of the discrimination against women that used to show up as a reduced A wage will now show up as a wider male-female differential within job B.
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 105 i. e., allow for the fact that, given sufficient time, supplies of labor to the two jobs will adjust to the changed wages prevailing for those jobs. As in the short-run case, He comparable worth policy will raise the A wage both absolutely and relative to the B wage (of men or women). 14 Again, then, the comparable worth policy raises wages in the low-wage, predominantly female jobs and narrows the pay gap between low-wage and high-wage jobs . - However, in the long run as in the short-run, the policy of raising pay for job A will also have several adverse side effects. First, as in the short-run case, firms' demands for workers forjob A will fall as the A wage rises. This will reduce employment of workers in job A, leading to unemployment for some individuals who would otherwise be in job A. (Since women are overrepresented in job A, this unemployment will hit women harder than men.) Second, the increase in the A wage relative to the wage for both men and women in job B attracts workers toward job A and away from job B. is This reduces employment of both men and women in job B. In the absence of any restraint on the A wage, this increase in the supply of labor to job A would drive the A wage back to its original level. However, the comparable worm policy prevents the A wage from falling; instead, the increased supply to job A turns into more unemployment. Finally, since total employment in job A declines and employment of both men and women in job B also declines, production drops. The drop in production results in an increase in the price level. Are these predictions about the effects of comparable worth policies sup- ported by any empirical evidence? The United States has not implemented comparable worth strategies to any considerable extent, but Australia's experience with its policy of enforc- ing equal pay for work of equal value is illuminating. Under the policy, 14 Thus, the policy-induced increase in the A wage will necessarily reduce the B wage of both men and women relative to the A wage. However, it is not possible to be certain about the impact of the increase in the A wage on the levels of the B wage for either men or women. On one hand, as in the short-run case, the increase in the A wage raises labor costs and thus prices. Other things being equal, this will reduce firms' demands forjob B and will therefore reduce B wages for both men and women. On the other hand, in the long run (but not in the short run), the rise in the A wage relative to the B wage (for both men and women) attracts labor away from job B and toward job A (see text below). Other things being equal, this will reduce the supplies of both men and women to job B and will therefore raise the B wage (of both men and women). The net impact of these two forces on wages in job B is indeterminate a priori: it depends on whether the supply effects are greater or less than the demand effects. All one can say forsure is that B wages for both men and women will fall or, if they rise, will rise by a smaller proportion than the A wage. 15 That is, some workers now in job B will tly to switch to job A, and some individuals who are now choosing a job will drop any plans they might have made to train forjob B and will seek training for and employment in job A instead.
106 KI~INGSWORTH which began in 1972, Australia's federal and state wage tribunals have set the same rate of pay (regardless of the gender of the majority of incumbents) for all jobs judged to be comparable in terms of skill, effort, responsibility, and working conditions. The tribunals fix minimum rates (not actual levels) of pay and have considerable latitude in determining comparability (Gre- go~y and Duncan, 1981:4081. These potential loopholes notwithstanding, Aust~lia's comparable worth policy apparently had a substantial effect on the aggregate female: male earnings ratio: that ratio (for full-time nonmana- gerial adult workers in the private sector) rose from .607 in 1971 to .766 in 1977 (Gregory and Duncan, 1981:4091. For purposes of this discussion, however, the most interesting aspect of Australia's experience is that the policy's side effects appear to have been generally adverse. Gregory and Duncan (1981 :418) found that increases in women's wages attributable to the policy reduced the rate of growth of female employment (below the rate that would otherwise have prevailed), relative to the rate of growth of men's employment, in (1) manufacturing, (2) services, and (3) overall (i.e., in all industries combined). Comparable worth had a negligible effect on the relative employment growth rate of women only in the public authority and community services sector. Their overall estimates imply that, as of 1977, the cumulative effects ofthe policy served to reduce the rate of growth of women's employment, relative to that of men, by almost one-third. Moreover, these estimates may understate the actual effect of the compa- rable worth policy. First, among the things Gregory and Duncan controlled for in deriving their results was the male unemployment rate (which, as indicated above, may also be affected by comparable worth policies). Sec- ond, due to data limitations, Gregory and Duncan analyzed the impact of comparable worm on the growth in the number of women employed without reference to its effect on the hours worked by women workers. Recent research suggests that the impact of comparable worth on hours worked was also negative and that focusing solely on numbers of employed women may therefore understate the full effect of Me comparable worth policy on employment opportunities for women (Gregory and Duncan, 1983; McGavin, l9X3a,b; Snape, 19801. Gregory and Duncan also analyzed the impact of the policy on female joblessness. Their estimates imply that the policy's cumulative impact as of 1977 was an increase in the female unemployment rate of about 0.5 of a percentage point. (The actual female unemployment rate in August 1976 was 6.2 percent.) In sum, the Gregory and Duncan study indicates that Australia's "equal pay for work of equal value" policy adversely affected both the rate of
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 107 relative employment growth for women and the female unemployment rate, as implied by the analysis in this section. 16 The Gregory and Duncan findings that Australia's comparable worth policy had no appreciable effect on women's relative employment in the public authority and community service sector highlight two potential exceptions to the general propositions presented above: the government sector and nursing. Much of the impetus for comparable worth seems to have come from public-sector employees, notably the American Federation of State, County, and Municipal Employees (AFSCME) (see Bureau of National Affairs, 19811. This may well be no accident. In contrast with the private sector, the government sector will have little or no difficulty (at least in the short run) in maintaining the demand for its "output" at existing levels, despite policy-induced increases in its labor costs. To cover the increased labor costs, government can simply compel the rest of the economy to pay higher taxes, keeping its real revenues unchanged. (The private sector will try to coyer policy-induced increases in labor costs by raising its prices, but it cannot compel the rest of the economy to go on purchasing the same amount of its output at the higher prices.) At least in the short run, then, declining demand for public-sector output (and thus declining derived demand for public-sector workers) is not likely 16 Without implying that they necessarily share the opinions expressed in this paper, I would like to thank both Gregory and Duncan for several very helpful discussions of the findings and implica- tions of their study. It is worth noting that numerous advocates of comparable worth in the United States are under the erroneous impression that the Gregory-Duncan study shows that comparable worth had little or no adverse effect on women's employment growth or unemployment rate (for details, see M. Killingsworth, 1984a). The reason for this seems to be that Gregory and Duncan presented two kinds of results: a set of simple descriptive statistics showing the raw or unadjusted time series of women's relative employment growth and relative unemployment rates and a set of regression analyses aimed at isolating the effect of comparable worth, other things being equal, on these measures of women's economic status. The simple time series show that relative employment grew and the relative unemployment rate fell, both before and after introduction of comparable worth. This is apparently the basis of claims by comparable worth advocates in the United States that Australia's policy had no adverse effects. However, by their nature, simple time series data do not abstract from changes in employment or unemployment rates that are attributable to forces other than comparable worth. In contrast, the regression results presented by Gregory and Duncan and discussed in the text explicitly abstract from changes in employment or unemployment rates attrib- utable to forces other then comparable worth (e.g., business cycle fluctuations and secular trends). As noted in the text, those results indicate that, other things being equal, the comparable worth policy reduced women's relative employment growth and raised women's unemployment rate. In other words, the results indicate that, if other factors had remained the same, women's relative employment growth would have been greater and the women's unemployment rate would have been lower had Australia not adopted comparable worth.
108 KIT;LINGSWORTH to be substantial. However, as experience with Proposition 13 indicates, whether this will be equally true in the long run is an open question. Note also that the absence of an appreciable short-run effect on women's relative employment in the government sector has nothing to do with the reasons for, or the extent of, the policy-induced increases in wages for mainly female jobs in that sector. That is, the short-run effects of the policy would be negligible regardless of whether wages in such jobs are raised to a level that is (still) below, equal to, or in excess of wages in "comparable," predomi- nantly male jobs. Nursing is another possible exception to the generally gloomy outlook regarding the likely effects of comparable worth (and it may well be no accident that nurses, like public employees, have been in the forefront of agitation for comparable worth; see Bureau of National Affairs, 19811. However, as I will explain shortly, the reasons may have little or nothing to do with the alleged comparability of nursing and predominantly male jobs. Alternatives to Comparable Worth Rejecting comparable worth certainly does not dispose of the problems of discrimination that have motivated many of the proponents of comparable worth. If comparable worth is not the answer to these problems, what is? This section considers two alternatives: conventional antidiscrimination measures (e.g., equal pay and equal access provisions of Title VII of the Civil Rights Act) and novel expansion or application of antitrust laws. C o n v e n t i ~ n a l A r l t i d i s c r i m i n a t i o n M e a s u r e s The essential difference between comparable worm and conventional antidiscrimination measures is very simple. Conventional antidiscrimina- tion measures make it costly for employers to treat equally qualified men and women differently. Comparable worth makes it costly for employers to employ low-wage, predominantly female labor. As shown above, compara- ble worth solves the problem of the artificially wide gap between low-wage, predominantly female jobs and high-wage, predominantly male jobs only to create other problems, including, in particular, reduced opportunities for women in both low-wage and high-wage jobs. In contrast, conventional antidiscrimination measures compel discriminatory employers to provide greater opportunities for women workers (by making it costly for such employers to deny equal pay or equal access to women with the same skills and qualifications as men). To the extent that conventional antidiscrimination measures are actually implemented, the wages of women rise for three distinct reasons. Two of
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 109 these are obvious. Equal pay requirements raise the wages of women within given jobs, and equal access requirements make it possible for qualified women to switch from low-wage to high-wage jobs. A third reason is more subtle: to the extent that wages respond to changes in supplies and demands (a matter to which I will return presently), the fact that some women are able to leave low-wage jobs for high-wage jobs will reduce supply to the low- wage jobs, leading to wage increases for low-wage jobs. Nor do conventional antidiscrimination measures lead to the kinds of adverse side effects that are fostered by comparable worth. True, each type of antidiscrimination remedy makes certain forms of behavior more costly. But, whereas comparable worth makes it more costly to employ low-wage, predominantly female labor, conventional antidiscrimination measures make it more costly to treat equally qualified men and women differently. In particular, such measures in effect force employers to incur costs (e.g., back pay) whenever they do not treat equally qualified men and women as equals . Depending on how stringent these requirements are and how vigorously they are enforced, the pecuniary costs they impose on employers will offset the nonpecunia~y or intangible factor of maleness that makes discriminatory employers want to treat equally qualified men and women differently. In consequence, stringent conventional antidiscrimination measures, strin- gently applied, will reverse all the effects of employer discrimination described earlier (Johnson and Welch, 19761. To be sure, some comparable worth proponents see comparable worth as a complement to, not a substitute for, conventional antidiscrimination mea- sures such as Title VII. Their reasoning seems to be as follows: conventional antidiscrimination measures will indeed eventually raise wages and employ- ment of women to the levels that would prevail in the absence of discrimina- tion. However, such measures have worked (and will continue to work) very slowly. Comparable worth offers the prospect of substantial and rapid increases in pay, bringing women's wages much more quickly to the levels that would prevail in the absence of discrimination. To the extent that it leads to wage increases, comparable worth will certainly also mean reductions in women's employment. However, such job losses will eventually be over- come as conventional antidiscrimination measures continue to open up job opportunities for women. In sum, conventional antidiscrimination measures will ultimately move the labor market to nondiscriminatory employment and wage levels for women whether or not comparable worth is adopted. How- ever, introducing comparable worth will provide many women with nondis- criminatory wage levels much more quickly than would be the case if comparable worth is not added to conventional antidiscrimination measures. Unfortunately, the premises underlying this reasoning are shaky. In effect, comparable worth asks women to choose between (1) immediate
110 KILLINGS WORTH wage increases and accompanying immediate employment reductions via comparable worth, followed by eventual employment increases (attained via conventional measures exclusively) the comparable worth strategy and (2) slower but steadier gains in both wages and employment vie conven- tional measures alone. It is not obvious that (1) is the better of the two alternatives. Second, as noted earlier, job evaluation cannot be expected to identify the wage structure that would exist in the absence of discrimination; the wage increases that would be required under comparable worth would be quite unlikely to be the same as the wage changes that would bring about a nondiscriminatory wage structure. Partly because of this, comparable worth is likely to be resisted-not only by discriminatory employers but also (and with good reason) by truly nondiscriminatory firms. Thus, the notion that comparable worth may provide quick wage increases may be seriously mistaken; if anything, employer resistance to comparable worth will be stronger than employer resistance to conventional antidiscrimination mea- sures. Finally, comparable worth diverts resources and energy that could otherwise have been devoted to implementation of conventional measures. That is particularly ironic because, once the adverse employment effects of comparable worth take place, a redoubling of conventional antidiscrimina- tion efforts will be necessary to counteract them. Antitrust Laws and the Problem of Employer Cartels As noted earlier, conventional antidiscrimination measures can be expected to raise pay in low-wage, predominantly female jobs (as well as providing higher pay in given jobs through equal pay provisions and permit- ting qualified women to leave low-wage jobs for high-wage jobs) if pay in such low-wage jobs responds to the changes in demand and, in particular, supply that will be set in motion by antidiscrimination measures. However, as many supporters of comparable worth have been quick to point out, that does not always happen; shortages in low-wage jobs, particularly in pre- dominantly female low-wage jobs, do not always lead to higher wages for such jobs. Many comparable worth advocates point to nursing as a particularly dra- matic example of the failure of wages to respond to supply and demand. Although wages in nursing are said to be low, hospitals and other employers of nurses are said to suffer from severe shortages. Yet these shortages are alleged not to have led to wage increases for nurses; about all that has happened is a step-up in recruiting efforts, either offoreign nurses or vie one- time-only inducements (a year's country club membership, a few months' paid rent, and so on) for first-time domestic recruits. Unfortunately, most comparable worth advocates have simply pointed out the seeming paradox
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 111 inherent in situations of this kind without asking how and why such situa- tions could have arisen or what can be done about them. The paradox begins to make sense, however, once one considers the possibility that markets for some kinds of predominantly female jobs (e.g., nursing) have been cartelizedthat, for example, hospitals and other large employers of nurses in major metropolitan areas have agreed not to compete with each other by offering higher wages to attract nurses. In effect, such cartelization amounts to a set of informal or formal areawide wage-fixing agreements. If this accurately describes the labor market for nurses, then it explains not only the alleged low pay of nurses, but also the alleged short- ages of nurses, the failure of nurses' wages to rise, and the almost exclusive reliance on nonwage forms of competition for new recruits. With wages held at an artificially low level, it is not surprising that individual hospitals would like more nurses than they are able to attract (i.e., face shortages); that individual hospitals do not raise pay in an attempt to attract more nurses; or that competition in the nursing market takes the form of foreign recruitment, one-time-only sign-tip bonuses, and so forth, rasher then higher wages just as competition in air travel centered on nonprice matters (seating, food, etc.) when airfares were regulated. Is there any evidence (as opposed to mere conjecture) that markets for nurses and other predominantly female jobs have in fact been cartelized? The nursing labor market is literally a textbook example of a cartelized (or, in economic jargon, "monopsonized") labor market (Devine, 1969:542; Ehrenberg and Smith, 1982:65-66; Hurd, 1973; Link and Landon, 19751. According to one witness at recent congressional hearings, hospital adminis- trators in Denver have colluded to fix wages (U.S. Congress, House, 1983:701. Similarly, another witness testified that employers of clerical workers in cities such as Boston and San Francisco have formed organiza- tions, euphemistically known as consortia or study groups, whose true pur- pose is to engage in wage fixing in much the same way that producer cartels engage in collusive price fixing (see U.S. Congress, House, 1983:~8, 961. To the extent that labor markets are indeed cartelized and I should emphasize that this is something about which, in general, there is very little hard evidence then forcing wage increases in such markets need not have any adverse impact on employment. However, none of this has anything to do with whether the jobs in question are comparable to predominantly male jobs, be they pharmacist, tree trimmer, or parking lot attendant. Indeed, raising wages in cartelized, predominantly female jobs to a level above the one that would prevail in the absence of cartelization will reduce employ- ment in such jobs (relative to the level that would prevail without carteliza- tion), whether the higher wage level exceeds oris below the level prevailing in supposedly comparable predominantly male jobs.
112 KIl I INGSWORTH What the possibility of labor market wage fixing cartelization does suggest is the advisability of a remedy that is quite different from both conventional antidiscrimination measures and comparable worth: enforce- ment and, if need be, amendment of the antitrust laws to ensure that employ- ers cannot collude to depress wages. The existence and importance of cartelization merits serious study; application (or extension) of the antitrust laws to attack discriminatory wage fixing deserves serious consideration. To sum up: both the short-~un and long-run effects of comparable worth seem, at best, less than completely satisfactory, particularly when compared with available alternative means for addressing problems of discrimination. Roughly speaking, comparable worth would provide benefits for some women (and also men) in low-wage jobs at the expense of everyone else (other women in low-wage jobs, and both men and women in high-wage jobs). It would also increase prices and reduce output. Suppose that a public official were to propose that employers be required to pay a tax for each female worker in their employ, with revenues from the tax being used for general purposes (e.g., reduction of the budget deficit). Women would obviously suffer severe costs in terms of reduced employ- ment opportunities, and the associated benefits would be quite indirect. Such a proposal would rightly be attacked as preposterous. Comparable worth is not terribly different, however; it amounts to requir- ing employers to pay a tax for each female worker in their employ and to using the resulting tax revenues to make payments to those women who manage to keep theirjobs despite the impact of the tax on women's employ- ment. 17 In this case, both the costs and benefits are obvious, but they are 17 Note that the Equal Pay Act is also in effect a tax on female labor, and so in this respect is similar to comparable worth; faced with the prospect of having to pay the same wage to men and women doing the same job, a discriminatory employer is likely to employ only men (who produce the intangible known as maleness as well as output per se) in preference to women (who produce output but not maleness). Indeed, Nancy Barrett (1979:55)an economist well known for her advocacy of effective antidiscrimination measureshas criticized the Equal Pay Act on precisely these grounds, arguing that the act was intended to help men compete with lower-paid women. While advocating comparable worth, Barrett has also warned that requiring pay increases for low- paid, predominantly female jobs pursuant to comparable worth will impose losses on firms and that "we can't expect firms to swallow those losses; that's crazy" (1984:32). The essential difference between the Equal Pay Act and Title VII is that the latter requires not only equal pay for equal work but also equal treatment of comparably qualified men and women with respect to hiring, assign- ment, promotion, and the like. Thus, under Title VII, a firm cannot avoid paying women the same wage as men are paid in a given job by simply not employing any women in that jobas would be possible under the Equal Pay Act. The essential difference between comparable worth and Title VII is that comparable worth merely makes it more expensive for firms to employ low-wage, predomi- nantly female labor, whereas Title VII makes it more expensive for firms to engage in unequal treatment of any kind (with respect to pay, hiring, promotion, and so forth) of equally qualified men and women.
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS 113 sustained or received by separate groups of women. No attempt would be made to compensate the losers from the gains of the winners; indeed, it is far from obvious that the aggregate gains of the winners would be sufficient to provide full compensation for the losers, even if such compensation were feasible. All In all, therefore, comparable worth is unlikely to appeal to large numbers of personsbut then, as has been emphasized repeatedly in this paper, there's no accounting for tastes. ACKNOWLEDGMENT Without implying that they necessarily share the opinions expressed herein, I thank Heidi Hartmann, Lawrence Kahn, and Cordelia Reimers for helpful comments. REFERENCES Abowd, J. 1984 Testimony on House Bill 1646 before the Illinois State Legislature, May 2. Graduate School of Business, University of Chicago. Ashenfelter, O., and J. Pencavel 1976 Estimating the effects on cost and price of the elimination of sex discrimination: The case of telephone rates. Pp. 111-124 in P. Wallace, ea., Equal Employment Opportunity and the AT&TCase. Cambridge, Mass.: MIT Press. Barrett, N. 1979 Womeninthejob market: Occupations, earnings, end career opportunities. Pp.31-61 in R.E. Smith, ea., The Subtle Revolution. Washington, D.C.: Urban Institute. 1984 Poverty, welfare and comparable worth. Pp. 25-32 in P. Schlafly, ea., Equal Pay for Unequal Work. Washington, D.C.: Eagle Forum Education and Legal Defense Fund. Belter, A. 1978 Title VII and the male/female earnings gap: An economic analysis. Harvard Women's LawJournal 1:157-173. 1980 The effect of economic conditions on the success of equal employment opportunity laws: An application to the sex differential in earnings. Review of Economics and Statistics 62:379-387. 1982 Occupational segregation by sex: Determinants and changes. Journal of Human Resources 17:371-392. Bloom, D., and M. Killingsworth 1982 Pay discrimination research and litigation: The use of regression. Industrial Relations 21:318-339. Blumrosen, R. 1979 Wage discrimination, job segregation and Title VII of the Civil Rights Act of 1964. Journal of Law Reform 12:399-502. Bodner, G. 1983 Should statistics alone be allowed to prove discrimination? Chronicle of Higher Educa- tion June 22:56. Brown, C. 1982 The federal attack on labor market discrimination: The mouse that roared? Research in Moor Economics 5:33-68.
114 KILLINGSWORTH Bureau of National Affairs 1981 The Comparable Worth Issue: A BRA Special Report. Washington, D.C.: Bureau of National Affairs. Clauss, C. 1981 Thelegalissues. Pp.8~86and91-94inEqualEmploymentAdvisoryCouncil, Compa- rable Worth: A Symposium on the Issues and Alternatives. Washington, D.C.: Equal Employment Advisory Council. Devine, E. 1969 Manpower shortages in local government employment. American Economic Review 59(May):538-545. Ehrenberg, R., and R. Smith 1982 Modern Labor Economics. Glenview, Ill.: Scott, Foresman and Company. Gold, M. 1983 Goldberg, M. 1982 Discrimination, nepotism, and long-run wage differentials. Quarterly Journal of Eco- nomics 97:307-320. Gregory, R., and R. Duncan 1981 The relevance of segmented labor market theories: The Australian experience of the achievement of equal pay for women. Journal of Post Keynesian Economics 3 :403-428. 1983 Equal pay for women: A reply. Australian Economic Papers 22 :60-64. Hartmann, H. 1984 The case for comparable worth. Pp. 11-24 in P. Schlafly, ea., Equal Pay for Unequal Work. Washington, D.C.: Eagle Forum Education and Legal Defense Fund. Hurd, R. 1973 Equilibrium vacancies in a labor market dominated by non-profit firms: The "shortage" of nurses. ReviewofEconomicsandStatistics55:234-240. Johnson, G., and F. Welch 1976 The labor market implications of an economywide affirmative action program. Indus- tnal and Labor Relations Review 29:508-522. Killingsworth, M. 1984a Testimony on comparable worth before the Joint Economic Committee, U.S. Congress, April 10. Pp.84- 127 in U.S. Congress, Joint Economic Committee, Women in the Work Force: Pay Equity. S. Hrg. 98-1050. Washington, D.C.: U.S. Government Printing Office. 1984b Heterogeneous Preferences, Compensating Wage Differentials and Comparable Worth. Unpublished manuscript, Department of Economics, Rutgers University, June. 1985 Economic analysis of comparable worth and its consequences. In Industrial Relations Research Association, Proceedings of the Thirty-Sixth Annual Meeting. Madison, Wis.: Industrial Relations Research Association. Killingsworth, V. 1981 What's a job worth? Atlantic 247 (February): 10, 17. Leonard, J. 1984a The Impact of Affirmative Action on Employment. Working Paper No. 1310, National Bureau of Economic Research. 1 984b What Promises Are Worth: The Impact of Affirmative Action Goals. Working Paper No. 1346, National Bureau of Economic Research. A Dialogue on Comparable Worth. Ithaca, N.Y.: Industrial and Labor Relations Press. Lewin, T. 1983 The quiet allure of Alan Greenspan. New York Times, June 5 (Section 3): 1, 12.
ANALYTICAL, EMPIRICAL, AND POLICY QUESTIONS Link, C., and J. Landon 115 1975 Monopsony and union power in the market for nurses. Southern Economic Journal 41 :498-508. McGavin, P. 1983a Equal pay for women: A re-assessment of the Australian experience. Australian Eco- nomic Papers 22:48-59. 1983b Equal pay for women: A postscript. Australian Economic Papers 22:65-67. O'Neill, J. 1983 The Determinants and Wage Effects of Occupational Segregation. Unpublished manu- script, Urban Institute, Washington, D.C. Polachek, S. 1978 Sex differences in college major. Industrial and Labor Relations Review 34:498-508. Rees, A. 1976 Compensating Wage Differentials. Pp. 336-349 in A. Skinner and T. Wilson, eds., Essays on Adam Smith. Oxford: Oxford University Press. Roos, P. 1981 Sex stratification in the workplace: Male-female differences in economic returns to occupation. Social Science Research 10: 195-224. Schwab, D. 1980 Job evaluation and pay setting: Concepts and practices. Pp. 49-78 in R. Livernash, ea., Comparable Worth: Issues and Alternatives. Washington, D.C.: Equal Employment Advisory Council. Smith, A. 1776 The Wealth of Nations. Reprinted 1937. E. Cannan, ed. New York: Modern Library. Snape, R. 1980 Wage Policy and the Australian Economy in the 1970's and Beyond. Seminar Paper No. 6, Department of Economics, Flinders University of South Australia. Treiman, D., and H. Hartmann, eds. 1981 Women, Work, and Wages: Equal Payfor Jobs of Equal Value. Committee on Occupa- tional Classification and Analysis. Washington, D.C.: National Academy Press. U.S. Congress, House 1983 Pay Equity: Equal Payfor Work of Comparable Value. Parts I and II, Serial No. 97-53. Committee on Post Office and Civil Service, Subcommittees on Human Resources, Civil Service and Compensation and Employee Benefits. Washington, D.C.: U.S. Government Printing Office. Yale Law Journal 1981 Equal pay, comparable work, and job evaluation. Yale Law Journal 90: 657-680.