You are on page 1of 16
AN EMPIRICAL STUDY OF LABOR MARKET SEGMENTATION PAUL OSTERMAN NE area of current interest in labor O economics is the study of struc tural interrelationships among jobs in the economy, an interest clearly ex- pressed in the growing literature on dual labor market theory and labor market segmentation. This paper pre- sents the results of an empirical test of some recent ideas of Piore! and Gordon ? about the segmentation of the labor force. The paper first develops a hypothesis that the quality of individual This study tests a refined version of the dual labor market theory which hypothesizes that the labor force is segmented in three groups: secondary jobs as usually defined, plus a pri- mary sector consisting of upper tier jobs. in which workers enjoy a high degree of auton- omy and personal participation in the work process, and lower tier jobs providing little au- tonomy and participation, Using data from the 1967 Survey of Economie Opporuunity, the thor shows that the determinants of annual earnings differ substantially among the three segments. Human capital characteristics explain the variance in individuals’ earnings very well in the upper tier and moderately well in the lower ter of the primary sector, but in the sec- ondary sector only the amount of time worked proved to be a significant determinant of earn: ings. Paul Osterman is 2 graduate student in the Department of Economies and the Department of Urban Studies and Planning, Massachusetts Institute of Technology. He would like to thank Peter Doeringer, Bennett Harrison, Lisa Peattie, Michael Piore, and Arthur Solomon for their assistance. Partial support for this study was provided by Grant #MH-23615 from the Center for the Study of Metropolitan Problems, National Institutes of Mental Health, to the Research Center for Economie Planning, New York City, through a subcontract to MLT.— Error 508 jobs forms an important dimension in understanding both worker behavior and the distribution of occupational re- wards, and this hypothesis is then tested with data from the Office of Economic Opportunity's Survey of Economic Op- portunity. Segmentation in the Labor Force Virtually all labor market studies have shown that the labor force is seg- mented in some sense, Readers of this paper, for example, will surely agree that entry into the academic profession is not open to everyone who possesses the modal amount of brains and know!- edge of those already in the profession. ‘The successful aspirant must climb a ladder, receive a certificate, and pass through some rituals. Clark Kerr gener- alizes this argument in his classic article, “The Balkanization of Labor Markets,” by arguing that many jobs exist in in- ternal labor markets characterized by limited ports of entry and internal sort- ing procedures Field studies of geo- graphic labor markets, such as Rey- ‘Michael J. Piore, “Notes for a Theory of Labor Market Stratification” (Department of Economies Working Paper No. 95, Massachu- setts Institute of Technology, 1972). =David M. Gordon, “From Steam Whistles 10 Coffee Breaks,” Dissent, (Winter 1972), pp. 197-210 8Clark Kerr, “The Balkanization of Labor Markets,” in F. Wight Bakke, et al, Labor Mo- bility and Economic Opportunity (Cambridge, Mass: MILT. Press, 1954), pp. 92-110 LABOR MARKET SEGMENTATION 509 nolds's* have confirmed that imperfect knowledge about and unequal access to jobs, mobility, promotion, and most other attributes of work lead to seg- mented labor markets within the larger labor pool. ‘The interesting question today is not whether the labor market is segmented, but rather along what lines. The con- cept of internal labor markets, operat- ing at the level of the individual plant or local area, is valid for some purposes, but it is not sufficiently general in scope to use in analyzing the economy as a whole. For that purpose, a far more use- ful concept of segmentation is that be- tween primary and secondary jobs, as described in the theory of the dual labor market.’ Piore summarizes well the differences between the two seg- ments: he primary market offers jobs which pos sess several of the following traits: high ‘wages, good working conditions, employ: ment stability and job security, equity and due process in the administration of work rules, and chances for advancement. The . . secondary market has jobs which, relative to those in the primary sector, are decidedly less attractive. They tend to involve low wages, poor working conditions, considera- ble variability in employment, harsh and often arbitrary discipline, little opportunity to advance.* Many of the distinctions between the two segments can be explained in terms of different job and technological re. 4Lloyd G. Reynolds, The Structure of Labor Markets (New York: Harper and Row, 1951) sSee Peter Docringer and Michael J. Piore, Internal Labor Markets and Manpower Analysis (Lexington, Mass.: D. C. Heath and Co., 1971) 5 David Gordon, Theories of Poverty and Under. employment (Lexington, Mass: D. GC, Heath and Co., 1972) “Michael J. Piore, “The Dual Labor Market,” in David Gordon, ed., Problems in Political Economy (Lexington, Mass: D, C. Health and. Co., 1971), p. 92, quirements and thus different on-the-job training patterns. Employers have little investment in secondary segment em- ployees and hence little incentive to en- courage stability, and the same is true for these employees.’ Rapid turnover and little cumulative growth in job skills result. In times of expansion, many employers will often avoid hiring sec- ondary workers in primary markets by subcontracting, using overtime, and changing job requirements, Clearly this dual market theory can be quite useful in understanding the problems faced by disadvantaged workers and in explain- ing a number of empirical findings, such as evidence of high frictional unem- ployment among the working poor. Simply segmenting the labor force into two parts, however, leaves a pri- mary sector of enormous variety and poor definition. The most common dis. tinction made to subdivide this mass is between blue-collar and white-collar work, the former usually characterized by manual labor and the latter by men- tal labor, Blau and Duncan, in their study of occupational mobility, reaffirm these distinctions: “The American occu- pational structure appears to be parti- tioned by two semi-permeable class boundaries that limit downward mobil- y between generations as well as within lifetime careers."? The barriers Gary Becker, Human Capital (New York: Columbia University Press, 1964); Walter Oi, “Labor as a Quasi-Fixed Factor,” Journal of Political Economy, Vol. 70 (December 1972), Pp. 588-555. sRobert E. Hall, “Why Is the Unemployment Rate So High at Full Employment?", in Arthur M. Okun and George L. Perry, eds., Brookings Papers on Economie Activity, No. 3. (Washing: ton, D.G. The Brookings Institution, 1970) , pp. 369-410, *Peter Blau and Otis Dudley Duncan, The American Occupational Structure (New ‘York: John Wiley and Sons, 1967), p. 59. Copyright © 2001 All Rights Reserved 510 INDUSTRIAL AND LABOR RELATIONS REVIEW to downward mobility are founded, those authors argue, on the reluctance of children of white-collar families to work with their hands and their subsc- quent willingness, if necessary, to accept lower paying clerical and service white- collar jobs, and also on the unavailabil- ity of farm jobs for the offspring of blue-collar marriages. On further reftection, however, the blue-collar-white-collar distinction loses much of its force, as David Gordon has pointed out. There is substantial earnings overlap between the two groups; in fact, Table 1 shows that the distribution of earnings among nonpro- fessional white-collar workers and blue- collar workers was quite similar in 1959, and these data are supported by more recent studies of the dispersion of earn- ings among occupations.» Blue- and white-collar work are simi- lar in a number of other respects. The machine has invaded the white-collar world, and it is unlikely that nonprofes- sional white-collar jobs are less routine or alienating than are blue-collar jobs. White-collar job markets are also. be- coming increasingly balkanized, with limited ports of entry and fewer chances for mobility.'? Alll of this is not to say that no differ- ences exist. between white-collar and blue-collar jobs; unionization, for exam- ple, remains an important division. However, when viewed as a_ totality, white- and blue-collar jobs seem more similar than not. By dimensions such as income, stability, and creativity, the dis- tinction does not seem justified. WGordon, “From Steam Whistles to Coffee Breaks,” p. 198. s3See, for example, Christopher Jencks, Ine: quality’ (New York:’ Basic Books, 1972), pp. 225-296, 1Gordon, “From Steam Whistles to Coffee Breaks,” p. 199. Table 1. Income Ranking of Office and Factory Occupations, 1959, Income Ranking Percentage Distribution of of Selected Clerical and Sales Blue-Collar Occupations® Occupations Occupations 1-100 13.7% 10.2% 101-150 15.9 20.3, 151-200 25.0 20.3 201-250 18.2 24.6 251-300 18.2 17.8 301-321 9.2 6.8 “In the original source, $21 occupations of all kinds were ranked by the median annual earnings of male workers in those occupations in 1959. For ease of analysis, those occupations were grouped as shown: the 100 that ranked highest in income, the next 50 highest, et. Source: Max Rotzick, “A Ranking of US. Occupations by Earnings,” Monthly Labor Review, (March 1965), as adapted in David M. Gordo, “From Steam Whistles to Coffee Breaks,” Dissent, (Winter 1972), p. 198 Degree of Autonomy How, then, can we classify the many different jobs within the primary sector? It is useful to think in terms of two groups of jobs within that sector: those in which workers enjoy a high degree of autonomy and a large amount of per- sonal participation in the production of the final product or service and, second, those jobs—which make up the great mass of work—which do not involve a great deal of worker autonomy and in which the worker lacks a sense of per- sonal commitment or attachment to the final product or service. These distinctions place greatest em- phasis on affective characteristics of the jobs; nonetheless, they bear a very close relationship to Piore’s ideas of the upper and lower tier in the primary sectors For example, he says that upper tier jobs embody the following ‘Piore, “Notes Toward a Theory of Labor Market Stratification.” LABOR MARKET SEGMENTATION Bll characteristics: higher pay and. status, greater promotional opportunities, mo- bility and turnover patterns similar to the secondary market but associated with advancement, lack of elaborate work rules but presence of internalized codes of behavior, and greater variety and room for individual creativity. This study puts greater emphasis on the last characteristic but no doubt we are talk- ing about much the same group. Simi larly, Gordon, in his three-part classifi- cation of workers, uses his concept of "The New Working Class” in much the same sense. It is not entirely clear, however, whether he is talking about the characteristics of jobs or the ideas of younger, educated workei Social Class Division What justifications are there for mak- ing this division of the primary sector into the two groups? Piore points out that the three divisions (secondary, pri- mary, and upper tier or elite workers) correspond to the sociological ideas of lower class, working class, and middle class and, in fact, that sociological de- scriptions of the attitudes of these groups toward work mesh quite well with our idea of the nature of the work in the three divisions. An example is Gans's description based on his study of Boston's West End: on the lower class: “work, like all other re lationships with the outside world, is transi: tory. Indeed, there can be no identification with work at all. Usually the lower class in- dividual gravitates from one job to another with little interest or hope in keeping a job for any length of time.” on the working class: “work is primarily a means of obtain nome to maximize the pleasures of life within the family circle. ‘The work itself may be skilled or unskilled; jordon, “From Steam Whistles to Coffee Breaks: it can take place in the factory or in the office—the type of collar is not important.” on the middle class: “work is not merely a job that maximizes income, but a seri¢s of related jobs or job advances which provide the breadwinner with higher income, greater responsibility, and, if possible, greater job satisfaction." cation for the proposed seg- ments of the labor force. What is dis- turbing, however, is that the shape of the classes is wrong. Piore and Gordon conceive of the upper tier (the “new working class”) as a smallish elite, yet their corresponding class—the middle class—is vast. Many holders of lower tier primary jobs probably should be classified as middle class as the term is commonly understood. Gans solves this problem by creating a new class, the “professional upper middle class.” It seems that the demarcation lines be- tween classes are unclear and that our ideas of the constituency of the various classes of society are, or should be, shiit- ing in a manner comparable to our ideas about labor segmentation. It is probably better, therefore, not to base our ideas about labor-market segments too closely on older, and perhaps no longer serviceable, ideas about social class. Social Status of Occupations Another rationale for the divisions in this study lies in the data available from the National Opinion Research Corpo- ration (NORC) survey of the social sta- tus of occupations, These data are at the heart of Blau and Duncan's study of occupational mobility and of the widely used Duncan scale for ranking occupa- tions. In their extensive interviews with iHerbert Gans, The Urban Villagers (New York: Free Press, 1962) , pp. 246-248. Copyright © 2001 All Rights Reserved 512 INDUSTRIAL AND LABOR RELATIONS REVIEW workers, Richard Sennett and Jonathan Cobb sought to understand why the NORC rankings were so consistently re- plicated by the people they interviewed. They point out that the top rankings on the scale are dominated not by peo- ple in economically powerful jobs but rather “by those who are involved in various interpretative functions in so- ciety, those who minister to physical, ed- ucational, psychic, and spiritual needs.” ** Why should this be? On the basis of their interviews, Sennett and Cobb argue that these jobs involve au- tonomy and that most people's aspira- tions for themselves and for their chil- dren center very strongly on a wish for au- tonomy over work and control over fate. Since the power of professionals lies in their ability to give or withhold knowledge, they are in positions that by and large are not questioned by others; they are “authorities” themselves, “authorities” unto themselves. It is precisely the endowment of professionals with this inner selfsufficing power that gives them a higher status than others with eco- nomic power. For their autonomy makes them seem market-proof in that they can perform their function no matter what is happening to others around them. Their nurturing power appears as an ability that they bring to people; others need them in a way that they do not need others. It is in this sense that the professional is the only truly independent person in a class society— needed more than he or she needs. Sennett and Cobb also point out that status ratings of jobs considered blue- or white-collar are intermixed, depending on the degree of freedom perceived for each job. Although this definition of autonomy has a somewhat perverse twist—ability to manipulate others and freedom from toRichard Sennett and Jonathan Cobb, “Work- ing Class Lives,” Social’ Policy, Vol. 3, No. 3 (September 1972) , p. 39. sIbid., p. 40. their constraints—the idea hits upon an important point. A rare quality of work is freedom: freedom to set one’s hours, freedom to set and keep personal stand: ards, freedom to shape a product to fit one’s sense of worth. Those jobs which embody freedom must be considered in a class to themselves. Job Alienation One of the most thorough studies of autonomy in American industries was conducted by Robert Blauner.1* He ex- amined alienation in the workplace and identified four aspects of ali : powerlessness, meaninglessness, isolation, and self-estrangement. To some extent all four ideas are included in this study's autonomy criterion, but certainly powerlessness is most central, Blauner’s definition of powerlessness is pertinent: A person is powerless when he is an object controlled and manipulated by other per- sons or by an impersonal system (such as technology, and when he cannot assert himself as a subject to change or modify this domination. Like an object, the power- less person reacts rather than acts... . Free- dom js the state which allows the person to remove himself from these dominating situa: tions that make him simply a reacting object.# Blauner goes on to identify four ele- ments of powerlessness in work: 1) the separation from ownership of the means of production and the finished product; 2) inability to influence general mana- 1 policies; 8) lack of control over conditions of employment; and 4) lack of control over the immediate work processes, Blauner finds that the last two are much more important to today workers than the first two. These four criteria are a good summary of those wsRobert Blauner, Alienation and Freedom (Chicago: University of Chicago Press, 1967) . Ibid. p. 16. LABOR MARKET SEGMENTATION 513 used in this study to determine whether to place occupations in the lower or upper tier of the primary sector. Empirical Analysis The material above sets forth the hy- pothesis that the labor force be fruitfully segmented into three parts. To test this hypothesis, each male worker covered by the 1967 Survey of Economic Opportunity (SEO) was as- signed to one of the three segments on the basis of criteria discussed below; simple frequency distributions were pre- pared for cach segment for « number of demographic, earnings, and labor-force behavior variables; and an_ earnings equation for each segment was specified and tested. Probably the most important and con- troversial aspect of this study centers on the criteria used to determine which oc- cupation goes into which segment, Un- fortunately, the procedure used subjective and obviously open to criti- cism. Perhaps the safest procedure would have been to use the Duncan scale, a widely accepted measure of occupational prestige. Duncan derived his scale by studying the NORC prestige rankings for occupations, fitting a equation of mean education and income of each occupation to the rankings, and then using the weights thus derived to rank all occupations.** This procedure resulted in a scale from zero to ninety six. The difficulty with using the Dun- can scale for this study is that it does not explicitly take the nature of the job into account. The distinguishing charac- teristics of secondary jobs—lack of inter- regression 28Otis Dudley Duncan, “A Sociometric Index for all Occupations,” in Albert Reiss, Occupa tions and Socral Status (New York: Free Press, 1961), pp. 109-138, nal labor markets, little on-the-job train ing, etc—are only indirectly reflected in the Duncan scale, And within the pr two occupations that vary quite widely on the dimension of autonomy, like college professors and engineers, could have similar rankings due to their similar education and in- come characteristics. For these reasons this study employed the census five-digit occupational classi- fications and, instead of the Duncan scale, the author's judgment was used to place each occupation in the proper seg- ment. A complete listing of the compo- sition of each segment can be found in the Appendix to this paper. Classifica- tions were based on the criteria dis- cussed earlier: the secondary sector con- tains occupations characterized by low wages, instability of employment, and similar factors, and within the primary sector each occupation was assigned to the upper or lower tier on the basis of the degree of autonomy and personal participation enjoyed by workers in that occupation. There is a high degree of correspondence between the upper tier described in this study and the top seg- ment of the Duncan scale and also be- tween the secondary segment and the lower tail of the Duncan scale. Examples of occupations which were relatively easy to classify are university faculty, skilled metal workers, and park. ng attendants, University faculty clearly enjoy the prerequisites of the upper tier; substantial autonomy and scope for self-expression, not to mention prestige and relatively high income. Skilled metal workers do not enjoy comparable flexibility of hours, opportunity for ex- pression, or freedom from supervision; they do, however, receive good pay, are generally unionized, and have internal career ladders and opportunities for ad- nary sector, Copyright © 2001 All Rights Reserved 514 INDUSTRIAL AND LABOR RELATIONS REVIEW vancement. Finally, parking attendants are classic secondary workers; the quired skills are minimal, their pay is low, and they have few, if any, oppor- tunities for on-the-job training or pro- motion. Not all occupations were as easy to Classify. Secretaries and typists, for exam- ple, seemed to fit partially into second- ary and partially into lower tier classi cations, In most cases their skills are relatively minimal, internal ladders are poor, and there is certainly little auton- omy, all of which suggests the jobs are secondary. On the other hand, in most urban labor markets there is excess de- mand for secretaries and typists which permits substantial interfirm mobility. It is therefore possible to move from, say, a job in a typing pool in an insur- ance company to the position of execu- tive secretary in a law firm or advertis- ing agency, These considerations, plus the job security resulting from excess demand, led to a lower tier classifica- tion. re- Another difficult. case was engineers As in many of the.upper tier jobs, eng! neers are well educated and receive good pay. However, most engineers work in an “assembly line” environ- ment, handling small, assigned pieces of a larger problem. Because they do not enjoy the same autonomy as do most upper tier workers, they were classified as lower tier. Several other occupations illustrate differences with the Duncan classifica- tion. The Duncan scale classifies coal miners in the lowest group, as might be expected from a prestige rating, Coal miners are generally unionized, however, and those who are, enjoy decent pay and benefits as well as something of an inter- nal labor market. For these reasons they were classified as lower tier for this study ‘At the other end of the spectrum, the Duncan index gives a relatively low classification to clergymen, artists, and musicians—roughly the same as foremen and mail carriers. The present study classifies artists, clergymen, and musi- cians in the upper tier because of their much greater than average opportuni- ties for autonomy.2* Clearly the ranking procedure is a major weakness of this paper, but it could be corrected only by someone with superior judgment or, even better, by the development of a generally agreed upon set of criteria for each labor-force segment.?? “The problem is further complicated because the census occupational classifications group musicians and artists with musie and art teach- cers, This illustrates the need for occupation-in- dustry cells rather than the simple occupational classification used in the SEO. Industry-occupa tion cells were not employed in this study because their use would have entailed substantial addi- tional expense, and also because the task of as- signing oceupations to such cells would have been even more dificult than assigning occupa- tions to the segments used in this study. For an example of the use of such occupation-industry cells, sce David Gordon, “Class, Productivity, and ‘The Ghetto: A Study of Labor Market Stratiica- tion” (PhD. dissertation, Harvard University, 1971). everal efforts are under way to develop more formal and “objective” classification pro- cedures. One study (Les Boden, “Occupational Mobility of Young Men," Ph.D. dissertation, M.LT., 1974) classifies jobs along the dimen- sions of complexity and autonomy, using data from the Dictionary of Occupational Titles, and then analyzes job mobility using the index of complexity and autonomy. Another study [Leo Kadanoff, Bennett Harrison, and Benjamin Chinitz, "A Simulation Model of Urban Labor Markets and Development Policy.” Paper pre sented to the 1974 mectings of the Regional Science Association (October 1974) | estimates transition probabilities among jobs in order to test the three-sector Piore segmentation moclel described above. Finally, 1 myself am engaged in the examination of links among jobs held by young workers (Paul Osterman, “The Youth Labor Market,” Working paper (Cambridge, Mass: MILT , July 1974) ) LABOR MARKET SEGMENTATION 515 One major analytical problem re- An important aspect of this as the examination of the role mains. study education plays in determining wages within each segment. However, there may have been a bias toward including high-education occupations in the upper tier and low-education jobs in the other tiers, even though education was not an explicit criterion, In fact, several high- education jobs do appear in the middle segment, but some bias doubtlessly re- mains, Blau and Duncan faced the same problem, but even more seriously, since education appeared explicitly in the equation determining their rankings and they then proceeded to use educa- tion as an independent variable in ex- plaining occupational outcomes. Their comments on this problem are therefore germany The occupational status scores were derived from aggregate data on all males in each oc- cupation category but applied as scores characterizing individuals... . The first re sponse to critics, then, might be that the status score, interpreted as an estimate of ‘occupational prestige, should legitimately re- flect the fact that one determinate of an oc cupation’s prestige is, in fact, the educa tional level of its incumbents, But because not all persons in an occupation have the same educational attainment, the formula for the status score does not by any means produce a perfect correlation between the estimated prestige of the individual's occu: pation and his educational attainment. On the other hand, in the light of our rather full knowledge of occupational prestige, no acceptable estimate of occupational prestige could fail to show some appreciable correla- tion between an individual's education and the prestige of the occupation in which he is engaged. It could be argued, in other words, that the apparent circularity of the procedure that was followed is simply a realistic reflection of the fact that high prestige occupations do recruit men with superior education whereas low-prestige occupations recruit men with inferior schooling, by and large.2? As an additional check, Blau and Dun- can developed an index which, instead of using education as an independent variable, used a dummy variable for white- or blue-collar work, There was a high correlation between the rankings from the two indexes. All of this does not entirely solve the difficulty implied by the education bias, but when combined with the fact that the bias does not actually result in uni- form rankings by education, it does tend to mitigate the problem. The Sample The data employed in this study came from the Survey of Economic Op- portunity of 1967, a national sample containing both the Current Population Survey and a supplemental sample of poor, predominately black families. The sample was weighted to make it nation- ally representative. In order to elimi- nate as many extraneous influences as possible, this study is limited to urban males, white and black, who were heads of families, who were without health disabilities which impinged upon their work activity, and who were in the labor force sometime during the year (but not necessarily the entire year) . Table 2 presents some basic data on each segment of the sample. These data follow the expected pattern: the second- ary sector is poorer, less well educated, and contains more nonwhites than the other segments and the upper tier pre- sents the opposite pattern. The segment size is also much as expected, though the secondary sector would have been larger in proportion to the others had individuals living alone, people with 2sBlau and Duncan, The American Occupe: tional Structure, pp. 125-126. Copyright © 2001 All Rights Reserved 516 INDUSTRIAL AND LABOR RELATIONS REVIEW Table 2. Character ics of the Labor Force Segments. Entire Sample Secondary Lower Tier Upper Tier Percent of sample 100 5.0 89.8 5.2 Percent white 92 27 92.7 98.4 Mean years of education 113 10.4 Wal 16.5 Mean annual earnings * $7608 $9328, $7437 $12,808 Mean age 41.8 eT) 416 42.7 N 4606, 234 4130 242 © Earnings include wages and income from personally owned business but exclude income from other sources, such as transfer payments, dividends, and interest, Source: Data from 1967 Survey of Economic Opportunity. health defects, and women been in- cluded. Table 3 shows the percentage of per- sons in each segment who worked in various industry groups. The data here also follow expected patterns. Secondary workers, for example, are overrepre- sented in industries such as manufactur- Table 3. Distribution of Workers by Segment and Industry (in percent) Lower Upper Secondary Tier Tuer Mining see ry 0 Construction fa rey Manufacturing (nondurable) ied) 1919) 220 ‘Manufacturing (durable) 15.1 25.3 10.2 ‘Transportation ie) 15°07 01) Communications 0.2 12 14 Utilities 20 23 10 Wholesale 61 46 © 0.0 Retail Ns UG 3.2 Financial 37 3B 4 Business and repair services 410 27° 25 Personal services 63 16 O07 Entertainment and recreation 2 (jh) Professional and related services «W153. 57.3 Public administration 7.9 7.07.2 ‘otal 100% 100% 100% Source: Data from 1967 Survey of Economic Opportunity. ing and retail trade, and they are un- derrepresented in construction. (The large numbers of secondary workers found in “professional and related serv- ices” represent, of course, the messenger, dlerical, and similar jobs in. professional establishments.) Also as expected, the upper tier is highly concentrated in the professional, public administration, and manufacturing categories (all in higher level occupations) . Slightly disturbing is the absence of upper tier workers in communications, but this is probably explained by the combination of the small overall numbers in communica- tions and the relatively small size of the upper tier sample. Earnings Within Each Tier ‘The theory underlying the segmenta- tion used in this study implies that the determinants of earnings differ among the three segments. For example, since the secondary market lacks structure and secondary workers lack skills, it is unlikely that seniority or education plays an important role in determining earnings in that segment, The middle segment, on the other hand, is quite structured and the wage-setting process is undoubtedly affected by plant and in- dustry characteristics as well as by skills and other personal attributes. ‘The LABOR MARKET SEGMENTATION 517 upper tier is akin to the secondary mar- ket in that its wagesetting mechanism is often informal, but one would expect that in the upper tier education and ex- perience would count for a good deal An individual's earnings are in all likelihood the result of a combination of personal characteristics, such as edu- cation and race, and structural charac: teristics of the industry in which he or she works. That is, two individuals with the same personal characteristics could have different incomes if working in firms with different capital-to-labor ratios, market structures, or seasonal employ- ment patterns. The reverse is also true, of course; two individuals with different personal characteristics could have the same earnings because of differences in the firms in which they work. There- fore, both personal and structural var bles were included in the regression analysis designed to answer the follow- ing questions: a. 1) Do the earnings functions vary among the three segments of the labor force? That is, do the segments exist in the sense that earnings are generated in fun- damentally different ways in the three segments? 2) If the earnings functions do vary. do they do so as predicted by the dual Jabor market theory and the additional ideas about the nature of the upper tier that this study has presented: To answer these questions, earnings equations were run for each segment The equations wer InE = B, + BA + Bad? + BEd + 20 BR + BU + BH +E pl + ¢ i-7 where E nual earnings (wages plus any in come from privately owned business) A= age Ed = years of school completed A? = age squared R = race (1 if white, 0 if black) U = weeks unemployed previous year = hours worked previous week fourteen dummy variables forthe industry in which the individual worked (I if in that industry, 0 otherwise) The equations were estimated using ordinary least squares. ‘The semi-log form was chosen after a review of the recent literature on earnings functions. The data do not permit complete speci- fication of the model by including terms for ability, motivation, and family back- ground. These omissions should have the effect of inflating the coefficients for the human capital terms such as educa- tion.” There is no reason, however, to assume that this inflation will vary across the segments. The variables “Age” and “Age Squared” were entered together because the quadratic form provides the best fit to the wellknown shape (rising and then leveling) of age-earnings profiles. The dummy variables for industry were entered in order to control for earnings determinants unique to an in dustry (such as market structure and capital intensity) which might other- wise bias the personal characteristic var- iables of interest in this study. In order to avoid singularity, one industry (dura- ble manufacturing) was omitted from 81See, for example, Barry R. Chiswick and Jacob Mincer, “Time Series Changes in Per: sonal Income Inequality in the United States, with Projections 0 1985,” Journal of Political Economy, Vol. 80, No. $, Part I, (May/June 1972), pp. 34-65, and Zvi Griliches and Wil- liam Mason, “Education, Income and Ability,” Journal of Political Economy, Vol. 80, No. 3, Part IT (May/June 1972), pp. 74-108. eSamuel Bowles, “Schooling and Inequality from Generation to Generation,” Journal of Po litical Economy, Vol. 80, No. 3, Part IL (May/June 1972), pp. 219-251 Copyright © 2001 All Rights Reserved 518 the equation and its coefficient thus en- ters into the constant. (The coefficients for the industry variables are not re- ported here but are available from the author upon request.) Empirical Results ‘The results from these regressions are presented in Table 4. It is apparent from the results that the wage-setting process does differ substan the segments. In the secon market, annual earnings seem to depend only on the amount of time worked, as expressed by the unemployment and weekly hours variables. Experience (age) does not contribute significantly to earnings, nor does education. Even race is not important. This equation thus supports the view that secondary INDUSTRIAL AND LABOR RELATIONS REVIEW employers view workers as substantially interchangeable and that differences among workers, at least as measurable in these data, are not important. In both the lower and upper tiers of the primary sector, the earnings process seems to be substantially more system- atic, with almost all of the variables being significant. Also interesting are the differences in magnitudes between those two segments. For example, a year of education returns more income in the upper than the lower tier. More: over, there is a substantial difference in the explanatory power of the two equa tions, of fact which implies that the human capital model better explains wagesetting practices in the upper tier than in the lower. (That conclusion should be tempered, however, by the Table 4. Results for Earning Equations for Three Segments. (Standard errors in parentheses) Independent Variable Seeondary Lower Tier Upper Tier Age 00056 06644 © 10324 #6 (03357) (00527) 02083) Aget 00005 # — 00069 + — 00098 + (00040) (00006) (-00023) Education 01433» 06170 2 09553 "8 (02175) (00283) (-01037) Race 11972 122655 * "31418 (13388) (03188) (22098) Hours worked previous week 01284 « 0718 — 00696 #* 00725) (00084) (00266) ‘Weeks unemployed —.05284 = — 02705 # = 01340 £01320) (00205) (03188) Constant 7.96659 6.22407 5.35066 R 208 246 44 F 2.65 53.9 10.9 Standard error of estimate 84607 31896 40249 Sample size 234 4130 242 © Regression coefficient significant at .05 level. © Regression coefficient significantly different at .05 level from comparable cocflicients in other equa~ tions (determined by pooling samples, adding segment interaction term to each variable for cach segment, and conducting ¢ tests on the coefficients) The dependent variable is the natural logarithm of annual earnings Source: Data from 1967 Survey of Economic Opportunity. LABOR MARKET SEGMENTATION 519 realization that the sample sizes are quite different). ‘The significance of the racial coeffi- cient in the lower tier and its lack of significance in the upper tier and sec- ondary segment conform to our expecta- tions. Secondary employers do not dis- iguish among workers; they all appear equally unskilled and unstable. Upper tier employment conforms most closely to the human capital model, and thus racial discrimination loses force. It is in the lower tier—the bulk of the blue- and white-collar world—that institu- tional constraints and long-established practices permit racial discrimination to continue. ‘The amount of variance explained by the three equations is well within the normal range for studies of this type. Because the data used in these studies describe individuals and not aggregate economic quantities, a great deal of ran- dom “noise” must be expected, and con- sequently most earnings functions ex- plain at best only 20 to 38 percent of the variance.** As noted above, the fact that the R? for the upper tier is quite high indicates that the human capital model explains earnings differences in that segment very well, whereas the low R? for the secondary segment implies, as the theory leads us to expect, that differences in human capital are sub- stantially less important in determining workers’ earnings in that segment. In order to control more stringently for the possible effects of race, the same equations were run separately for whites and for blacks. This was done to guard against the possibility that the previous equations, by “averaging” the coeffi- cients for such personal characteristics 29For examples of typical earnings equations il their success in explaining variance, see the special supplement of the Journal of Political Economy, Vol. 80, No. 8 Part Il (May/June 1972) Table 5. Results for Earning Equations for Three Segments, Whites Only. (Standard errors in parentheses) Independent Variables ‘Secondary Lower Tier Upper Tier Age 00624 7164» 09645 = (08211) (00556) (.01993) Age? —.00008 = .00076 * 00098 + (00050) (,00007 ) (00022) Education 01076 105547 # -08387 * (02774) (00289) (00808 ) Hours worked previous week <01173 00668 = = 00748 (,00976) (.00087)) (,00263) Weeks unemployed ~ 06513 * = 02468 * ‘ (01888) (,00220) Constant 7.71766 6.37499 9.94599 RE -080 189 434 F 2.8 178.0 35.0 Standard error of estimate op 1e¢ 3024 40782 170 Significant at .05 level. + Insufficient cases 3830 239 ‘The dependent variable is the natural logarithm of annual earnings. Source: Data from 1967 Survey of Economic Opportunity. Copyright © 2001 All Rights Reserved 520 INDUSTRIAL AND LABOR RELATIONS REVIEW as education and age, failed to control for some important differences between the two groups, The results, presented in Tables 5 and 6, are substantially the same as those in Table 4 (although, since race is omitted, the equations in Tables 5 and 6 have less explanatory power), The only real surprise is the significant coefficient for education for blacks in the secondary segment. Vir- tually all recent studies have found that education has little benefit for blacks Table 6. Results of Earnings Equations for Two Segments, Blacks Only.* (Standard errors in parentheses) Independent Variables Secondary Lower Tier Age 03653 04916 * (,02733) (01583) Age? —.00039 > = .60056 * (,00033) (00019) Education 05156 — .04505 * (01460) (00684) Hours worked previous week 00329 00447 (.00576) —_(.00277) Weeks unemployed — .04220 — .02972* (00764) (.00479) Constant 6.91190 6.8421 R 303 2128 F 10.2 20.0 Standard error of estimate 46737-63643, Sample size © 123 703 @ There were insufficient cases to estimate an ‘equation for the upper tier. ° Significant at .05 level. «© In previous regressions I used half of the SEO ‘cases available to me (this was possible because the producers of the data prepared several subfiles with equal means and variances for all variables) For this equation, in order to augment my sample size, I used all the cases. Hence the sample size for this table is not comparable to that used in the other tables. ‘The dependent variable is the natural logarithm of annual earnings. Source: Data from 1967 Survey of Economic Opportunity. with a high school diploma or less,*7 and this particular coefficient should be viewed with suspicion. Conclusion For the reasons noted, the findings of this study strongly support the dual labor market theory. Probably the weak- est link in this conclusion is that it is by no means clear that the occupational classifications in the Appendix are opti- mal; it is possible that another classifi- catory scheme would have yielded dif- ferent but equally significant results, It is very unlikely, however, that the re- sults obtained here were a product of chance, for they are consistent with a theory—that of the dual labor market —which has been proposed and tested elsewhere.** For example, it was shown, as the theory predicts, that the human capital model holds up very well for upper tier workers but has little explan- atory power for workers in the second- ary labor market. This leads to the con- clusion that the basic ideas tested in this study are sound, and that the find- ings did not result from a random cut at the labor market that happened to prove significant. Labor market segmentation can have important implications for public pol- icy, Probably one reason manpower pro- grams in the 1960s did not always suc- ceed was that they were not attuned to See, for example, Bennett Harrison, “Educa tion and Underemployment in the Urban Ghetto," American Economic Review, Vol. 62, So, 5" (December 1972), pp. 769-812. Though dual labor market theory has not been tested as extensively as other views of the labor market, and this paper seeks to help fill, that gap, there have been several other empi cal studies. See, for example, Bennett Harrison, Education, Training, and the Urban Ghetto (Baltimore: Johns Hopkins University Press, 1972) ; and Gordon, "Class, Productivity, and the Urban Ghetto.” LABOR MARKET SEGMENTATION 521 the realities of the labor market, The results presented here show, for exam- ple, that policies designed to augment the human capital of secondary workers are not likely to improve their earnings. Broader aspects of public policy are also affected. For example, in Jencks's con troversial study, Inequality, the analysis of the role of occupations in determin- ing income equality and the role of edu- cation in determining income within occupations proceeds with little ac- knowledgement of labor market dynam- ics.*" The findings presented here suggest that an individual’s income is greatly affected by the segment of the labor market in which the individual works, and movements between segments or the alteration or regulation of segments may need to be important matters of public concern, jencks, Inequality. Copyright © 2001 All Rights Reserved 522, INDUSTRIAL AND LABOR RELATIONS REVIEW Appendix Occupational Composition of Labor Force Segments Upper Tier of Primary Sector Accountants Actors Architects Artists & art teachers Authors Cabinetmakers Chemists Chiropractors Clergymen College professors & instructors Dancers Decorators Dentists Designers Economists Editors & reporters Entertainers Faculty (nec) Lawyers & judges Miscellaneous social scientists Musicians & music teachers Photographers Physicians & surgeons Piano tuners Psychologists Scientists Statisticians & actuaries Veterinaria Lower Tier of Primary Sector Advertising agents, Agents Airplane mechanics ‘Apprentices Asbestos workers Assemblers Athletes Auctioneers ‘Auto mechanics Bakers Bank tellers Barbers Bartenders Bill collectors Blacksmiths. Blasters Boarding & lodging housekeepers Boatmen Boilermakers Bookbinders Bookkeepers Brickmasons Bus drivers Building managers & superintendents Bus & street car conductors Buyers Cab drivers & chauffeurs Carpenters Cashiers Cement finishers Compositors & typesetters Craftsmen Cranemen Gredit men Decorators Deliverymen Demonstrators Draftsmen Dieticians Dressmakers Dyers Electrical technicians Electricians Electrotypers Engineers Engravers Farm managers Farm & home management advisors Farmers File clerks Firemen shermen Floormen Foodpackers Foremen Foresters Forgemen Funeral directors Furnacemen Furriers Glazers Hairdressers Heat treaters Inspectors (me.c.) Jewelers Librarians Library attendants & assistants e firemen Log & lumber sealers Longshoremen Loom fixers Lumbermen Machinery operators Machinists Mail carriers ‘Managers, officials, proprictors (n.e.c.) Manufacturing checkers Manufacturing sewers Marshals Meat cutters Mechanics & repairmen (nec) Medical technicians Metal filers Metal heaters Metal job setters Metal molders Metal rollers Midwives Millers Milliners Millwrights Mine operators Motion picture operators Motormen (nec) Nurses, Office machine operators Officials (lodge, society, union) Oilers & greasers Operators & kindred (n.e.c:) LABOR MARKET SEGMENTATION 528 Opticians Optometrists Osteopaths Painters Paper hangers Pattern & model workers Personnel workers Pharmacists Photo process workers Photo engravers Pilots Plasterers Plumbers Policemen Postal clerks Postmasters Power station operators Printing pressmen Professional, technical, & kindred (ne.c.) Public administration inspectors Public administration officials Public relations men Purchasing agents Radio operators Radio & television mechanics Railroad brakemen Railroad conductors Railroad mechanics Railroad switchmen Real estate agents Receptionists Recreation workers Religious workers Roofers Sailors Salesmen & sales clerks Sawyers Secretaries Appendix (Continued) Sheriff Shipping officers Social & welfare workers Shoemakers & repairmen Sports instructors Stationary engineers Stationary firemen Stenographers Stock clerks & storekeepers Stock salesmen Stonecutters, Structural metal workers Surveyors Surveying chainmen Tailors ‘Teachers, elementary Teachers, high school Teamsters Technicians (ne.c.) Textile spinners Textile weavers ‘Therapists and healers Ticket agents Tin and coppersmiths Tool & die makers Truck drivers ‘Typists Vehicle dispatchers Welders Secondary Sector Attendant, hospital Attendant, personal services Attendant, recreation Auto service & parking attendant Baby sitter, private household Bootblacks Bridge tenders Carpenters’ helpers Charwomen Clerical & kindred (n.e.c) Cooks Counter & fountain workers Elevator operators Express messengers Gardeners Garage workers, car washers Guards, watchmen, doorkeepers Housekeepers & stewards Hucksters & peddlers Janitors Kitchen workers (ne.c) Laborers Laundry operatives Laundresses, private household Maids Manufacturing graders & sorters Messengers & office boys Newsboys Office machine operators Packers & wrappers Porters Private household workers (nec) Service workers (ne.c) Shipping & receiving clerks ‘Telegraph messengers ‘Telegraph operators Telephone operators Textile knitters ‘Transportation baggagemen ‘Truckdrivers’ helpers Ushers Waiters Warehousemen (ne.c) Copyright © 2001 All Rights Reserved

You might also like