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.
6 Comparable Worth, Occupational Labor Markets, and Occupational Earnings: Results from the 1980 Census TOBY L. PARCEL Comparable worth refers to equal pay for work of equal value. Proponents claim that jobs that are disproportionately held by women (e.g., nursing, clerical work) are not rewarded at the levels of men's jobs that, although (different in content, deman(1 sim- ilar levels of skid and are performed under similar conditions (Hartmann, 1985; Rem- ick, 19844. Such a reality means that wom- en's years of schooling, experience, and effort are rewarded at a lower rate than are men's, which leads to pronounced and per- sistent sex differences in earnings. At the same time that this issue has arisen in the policy realm, sociologists have be- come interested in structural explanations of economic inequality. Analyses consistent with "the new structuralism" (Baron and Bielby, 1980) investigate the functioning of economic sectors (e. g., Bibb an(1 Form, 1977; Hocison, 1984; Wallace and Kalleberg, 1981), the configuration of class structure in contemporary U.S. society (e.g., Robin- son and Kelley, 1979; Wright et al., 1982), and the impact of organizational mecha- nisms on worker outcomes (Baron ant! Biel- by, 1984; Bielby and Baron, 1984~. One branch of the literature, that concerning the existence and functioning of occupational labor markets, appears particularly relevant to thinking on comparable worth. In this paper I develop a theory of occupational earnings differences that draws on both ideas regarding comparable worth and sociolog- ical arguments suggesting market influences on wages. I test the theory using (lata from the 1980 U.S. census an(l incorporate (lata from the Dictionary of Occupational Titles (DOT) (U.S. Department of Labor, 1977) as controls for job content. In the early sections of the paper I outline theory and hypotheses used to gui:le the analysis. I then (lescribe the (data used to test the hypotheses and report both (descriptive and analytic findings. The findings are inter- prete(1 as bearing on hypotheses regarding the effects of job content ant! market influ- ences on earnings an(1 on the relationships among those factors. ALTERNATIVE THEORIES AND LITERATURES Although there is little (lispute regarding the extent of the male-female earnings gap, analysts slider regarding its causes. Englan~l 134
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS and Farkas (1986) summarize literature sug- gesting that job segregation by sex explains a good deal ofthe difference (e.g., see Bielby and Baron, 1984) and that human capital variation, particularly in work experience (Corcoran ant! Duncan, 1979), also accounts for some of the difference. Those research- ers who investigate earnings differences within a comparable worth framework argue that "skfl~" or "worth" accounts for only a portion of the gender gap in earnings and that the "femaleness" of occupations also influences sex differences in earnings (En- gland et al., 1982; Englanc] and Norris, 1985; Treiman et al., 19841. Critics of comparable worth argue that advocates ignore the role of the quantity of labor supplied ant] demanded in setting wages across occupational markets. Waldaur (1984) argues that the balance between quantity supplied and demanded influences occupational wage levels. Using data on the number of Ph. D.'s in various academic dis- ciplines, he demonstrates a noticeable pos- itive relationship between the existence of nonacademic job opportunities and remu- neration. His finclings, however, cannot be assumed to apply to the labor force as a whole, nor does he subject his argument to a rigorous multivariate evaluation. There is an obvious, although generally ignored, connection between ideas con- cerning the impact of external market forces on wage setting and sociological notions involving the structure and functioning of occupational labor markets. Following work by institutional economists such as Dunlop (1957) an(l Kerr (1954), Sto~zenberg (1975) argues that skill differentiation produces labor market fragmentation along occupa- tional lines, that is, that the supply of labor is differentiated by occupation. He suggests that occupations constitute distinct markets because the skflIs demanded across occu- pations differ. The fact that individual re- turns on skills would be reduced if workers were to seek employment in occupations for which they were untrained supports this ]35 idea. He also argues that the social orga- nization of labor market processes varies by occupation, with some being unionizer] or governe(1 by professional associations that influence wage structures. Occupational so- cial organization can also be viewed in terms of race and sex composition; the higher the percentage of minorities within an occu- pation, the greater the competition between minority and majority workers, which bids (town the wage rate. Several strands of thought embedded in Sto~zenberg's analysis are particularly useful to this discussion. First, he argues that forces of supply and (demand may vary by occupation, a notion very compatible with treatments by Waldaur and others, but much less compatible with the position of a(lvo- cates of comparable worth who argue against the justice of those forces. Second, his (lis- cussion of skill differs at least in emphasis from that implied by the worth perspective. In the comparable worth perspective, skills are those common (dimensions, such as man- ual dexterity or substantive complexity, that underlie substantively distinct bundles of training and talent. It is reliance on these common (limensions that makes it possible to (levelop evaluations comparing the rel- ative worth of different jobs, such as nursing and computer programming. For Sto~zen- berg, however, the emphasis is on skills as differing packages that render the capability to perform well at nursing different from the capability to work as a computer pro- grammer. It is the notion of skill fragmen- tation that supports variation in demand for labor across occupations, a phenomenon that has no place in a comparable worth conceptualization. The theory used in this paper makes greater use of the worth per- spective on skill. Finally, Sto~zenberg (listinguishes from forces of supply and clemand and from skill a third category of factors that can influence wages: dimensions of occupational social organization. His examples include union- ization and race and sex composition, factors
136 that clo not directly tap the quantity or quality of labor available in an occupation or that might be potentially demancled, but which may nonetheless exert independent influence on wages. Thus in this context, "social organization' refers to nonproduc- tivity characteristics of occupations or of the occupational incumbents themselves. Stuc3- ies within the comparable worth tradition, for example, document the impact that the percentage female has on occupational earn- ings, independent of job content, and att- ribute that influence to discrimination. The contribution of Sto~zenberg's think- ing on occupational market social organi- zation to studies of comparable worth is that it broadens the concern with nonproduc- tivity factors to incorporate other variables. Just as comparable worth advocates object to the fact that the higher the percentage female in an occupation, the Tower its wages (controlling for job content), one can identify other nonproductivity measures that may operate similarly. Percentage black is an obvious cancliclate, and as I indicate below, the marital status of occupational incum- bents can be viewed in the same way. Although the percentages of women and blacks are demographic characteristics, Stol- zenberg also includes unionization as a cli- mension of social organization. He argues that the existence of unions within an oc- cupation may push wage levels up, in(le- pendent of the productivity of workers. Thus, by considering how the social organization, broadly defined, of an occupation may in- fluence wages, one is motivated to consider nonproductivity factors other than per- centage female in the empirical analyses. THEORY AND HYPOTHESES The theory borrows from both the worth and occupational labor market perspectives. First, consistent with the worth perspective, I conceptualize skill in terms of common dimensions un(lerlying substantively (lis- tinct combinations of training and talent. PAY EQUITY: EMPIRICAL INQUIRIES Such an assumption enables one to compare the job content of seemingly disparate oc- cupations, an important first step in gen- erating inferences regarding elects of job content on earnings. Second, consistent with Sto~zenberg, I argue that occupations do indeed] constitute distinct labor markets with potentially differing values for key dimen- sions of market conditions that potentially affect the quantity of labor supplied and demanded. I also argue that independent of these market conditions occupations (lifter in terms of social organization. Dimensions of social organization do not reflect pro- ductivity-relevant characteristics of occu- pational incumbents, but they nonetheless may still influence wage levels. Recognizing the existence of these nonproductivity di- mensions is vital to understanding the bases along which pay inequality clevelops. In this conceptualization, supply and cle- mand refer not to actual quantities of labor supplied and demanded but rather to factors that could be expected to influence levels of supply and (leman(l. Thus, the usage does not strictly follow that of economists, nor does it constitute a completely cleveloped theory. Still, within these limits, notions of supply and (leman(1 appear relevant and useful. First, for example, following Wal- (laur (1984), I argue that (liverse labor mar- ket opportunities for employment in an oc- cupation will raise its earnings level. Occupations for which a narrow range of settings provi(le employment opportunities will have reduce(1 bargaining power with employers compared with occupations for which the skflIs demander] could be utilized in a variety of types of employment settings. Second, the higher the unemployment rate in an occupation the lower the wages; high unemployment could indicate weak quan- tity demanded and/or excess quantity of those skills supplied, thus depressing the earnings of those workers who are employe(l in that occupation. Third, I anticipate a negative relationship between earnings and the size of an oc-
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS cupational pool that has skills and experi- ence but is not in the labor force. If workers in an occupation can be easily replaced, and if the potential exists for drawing trained, experienced workers from a reserve labor pool into the labor force, workers currently employed are in a relatively poor position to bargain with employers. Fourth, I argue that substantial government employment in an occupation will increase earnings because government jobs pay more, at least at the entry level, than many jobs in the private sector ant] may tend to bid up wage rates for nongovernment workers in those oc- cupational markets. At the higher levels of government service, however, earnings may be depressed relative to the private sector; these relationships may become obvious in the multivariate analysis. Similarly, I hypothesize that the produc- tivity-related characteristics of occupational incumbents may influence earnings; these can be interpreted to reflect the quality of labor supply. I view the overall educational and experience levels of workers within occupations as supply factors that should be positively related to earnings levels. Fol- lowing general human capital arguments, investments in schooling and experience should be rewarded with financial returns, independent of other market characteristics and of job content. Turning to the social organization of oc- cupational markets, although several studies find that the percentage female does influ- ence earnings, I extend the investigation of the social organization of markets in three directions. First, racial composition is an important and relatively neglecter] aspect of market social organization. Remick (1984) notes the relative lack of attention to the role that racial minority participation in occupations plays in determining earnings, although traclitionally within sociology and economics the "competition" ant] "crowd- ing" literatures have studied the issue (Berg- mann, 1971; Hodge ant] Hodge, 1965; Sny- der and Hudis, 1976; Taeuber et al., 19664. 137 These analyses, however, have not evalu- ated the impact of racial minority partici- pation on earnings controlling for job con- tent and other market conditions. I hypothesize that the higher the level of black and Hispanic participation, the lower the earnings, an hypothesis (Erectly parallel to that concerning female concentration. In addition, the data enable one to distinguish among several distinct racial minorities. Al- though I hypothesize that both percentage black and percentage Hispanic will be neg- atively associated with earnings, I expect that percentage Asian will be positively as- sociatec] with earnings. Asian minority groups in the Unite(1 States have historically achieve higher levels of schooling and socioeconomic status than black and Hispanic minorities, and I expect these differences to be reflected in this analysis as well. Second, the nexus of family organization and work outcomes is of interest. Certainly the marital status of family members can influence earnings. Traditionally, the as- sumption of marital responsibilities has been viewed differently by employers, depending on whether the employee is mate or female. For men, the stereotype is that married men are more settled than unmarried men- they have assumed responsibility for finan- cially supporting a wife an(l possibly chil- (lren and therefore are likely to be more depen(lable than unmarried men. For wom- en, getting married is viewed as a harbinger of interrupted labor force participation in order to rear children or to move with a husband when he is relocate(l. At a mini- mum, marriec] women may have re(luced bargaining power with their employers if the latter perceive them as likely to accu- mulate less firm-specific experience than their male counterparts. Certainly, these stereotypes (lo not operate with consistency. Women are increasingly remaining in the labor force while their children are young (Bureau of Labor Statistics, 1986), and fam- flies in which both spouses are employe(l do not necessarily sacrifice the wife's work
138 situation to accommodate that of the hus- ban(l. However, until very recently the traditional models were considered nor- mative. I utilize these models to suggest two hypotheses. First, the higher the proportion of men who are married in an occupation, the higher its earnings. Second, the higher the proportion of women who are married in an occupation, the lower its earnings. Of course, a number of competing explanations cannot be evaluated in this analysis. Married women may be willing to sacrifice wages for convenient job location, regular hours of work, and minimal demands for travel and other activities that could interfere with family obligations, or they coulcl actively seek such advantages at the expense of wage increases. Married men may be willing to accept or actively seek such job conditions in ogler to realize promotions and attendant wage gains. In addition, occupations with high proportions of unmarried men or with high proportions of married women may contain a higher number of entry-level or secondary labor market type jobs, and thus the relationships observed may not be di- rectly rooted in the intersection of family and work arrangements. Still, if the pro- portions of married men and women are statistically significant net of job content factors and of female concentration, that may suggest the possibility of discrimination independent of sex composition. I expect sex composition ant! proportion of women married to operate in the same direction, and for each to contribute to an explanation ot earnings. Second, following Stolzenberg (1975), who argues that occupational unionization pos- itively influences wages, as well as the ex- tensive literature reviewed by Freedman and Medoff~1984), I hypothesize that union- ization will positively affect earnings across occupations. Finally, I control for the fact that workers in urban areas earn more than those in nonurban areas. To avoid con- founcting, I control for the extent of urban- PAY EQUITY: EMPIRICAL INQUIRIES ization in an occupation, which I expect to have a positive effect on earnings. PREVIOUS TESTS OF RELATED THEORY This theory has not been fully tested in prior research. Several studies have found that the sex composition of jobs or occu- pations affects pay levels (England and Far- kas, 1986; Hodge and Hodge, 1965; Remick, 1984~. Even recent analyses, however, (licI not systematically evaluate the effect on earnings of market conditions influencing supply or demand, nor did they utilize the notion of occupational market social orga- nization to expand the focus of nonprod- uctivity characteristics beyond percentage female. England et al. (1982) control for a wide variety of skill factors in studying earn- ings differences across occupations. They document that although male- ant] female- dominate(1 occupations differ in the skills required, the (differences do not explain the gender earnings gap because female-clom- inatecl occupations pay less than expected on the basis of their skill clemands. In a similar analysis, England ant] Norris (1985) interpret the effect of percentage female on earnings, with other skill factors controllecl, as "comparable worth (discrimination" and suggest that this explains close to 30 percent of the annual earnings differential between men and women employed full time and year-round. Still, the impact of skill on earnings is evaluated, but market conditions anal social organization are slighted. Treiman et al. (1984) use 1970 census data to evaluate whether returns to job content characteristics vary depending on whether the occupation is male or female clominatec] and to estimate those relation- ships across all occupations. They find that workers in female-clominated occupations obtain returns to motor skills and physical demands but not to complexity of their work, and that independent ofthese "worth" factors, the higher the concentration of fe-
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS male incumbents, the lower the level of occupational earnings. They estimate that there is substantial inequality in pay after the worth factors have been controllecI; per- haps as much as 66 percent of the difference is attributable to some form of cliscrimi- nation. This study is of particular importance because it described relationships between measures of worth and percentage female, thus confronting very directly, although in- completely, the fact that there are rela- tionships among measures of job content, market conditions, and occupational social ~ '' . . Organization. Parcel et al. (1986) tested a theory similar to the one -described in this paper using 1970 census data. They operationalized a much broader range of occupational market characteristics than did the earlier works describer] above. They fount] predictable zero-order relationships between occupa- tional earnings and measures of occupational market conditions, such as reserve labor pool and unemployment rate, as well as with aspects of market social organization, such as percentage nonwhite, percentage of female occupational incumbents married, and percentage of male occupational incum- bents married. The strongest effects on earnings were from years of schooling, ex- perience, ant] percentage female, but most of the remaining effects were maintained in multivariate analysis when job content was controlled. This analysis demonstrated the feasibility of conducting an analogous investigation with data from the 1980 cen- sus, the results of which I report below. METHOD Measurement of lob Content The unit of analysis is the detailed oc- cupation, as defined by the 1980 census. To measure job content, I matched DOT (lata to (lata on 1980 detailed census oc- cupations (provided by Dr. Paula England). 139 These data were created using a tape of individual respondents that hacI been dou- ble-coded using 1970 and 1980 occupations. Data for 503 occupations were produced; these occupations were clerivec! by reducing all industrial, government sector, and pub- lic-private variation in occupational titles to a single title by computing weighted av- erages within occupational categories. I con- ducted a principal components analysis with orthogonal rotation of the 503 occupations to facilitate data reduction ant! to clevelop reliable measures that wouicl serve as good measures of job content. In the solution reported here, eight factors met the eigen- value criterion and five were substantively interpretable; the five factors accounted for 67.7 percent of the common variance. I formed factor-based scales of each inter- pretable factor, where inclividual items con- sisted of Z scores and were reversed in direction as necessary so that all items within a scale ran in a consistent direction. For an item to be incorporated into a scale, it ha(1 to have loaded at an absolute value of.40 or higher on a factor. The first extracted factor, "substantive complexity, contains 17 indicators tapping General Eclucational Development; Specific Vocational Preparation; intelligence; com- plexity of functioning with (lata ancl with people; verbal and numerical aptitudes; preference for dealing with abstract and .. .. .. ... creative versus routine, concrete activities; repetitive or continuous processes; direc- tion, control, and planning of others' activ- ities; sensory or judgmental criteria; reach- ing, handling, fingering, an(l feeling; clerical, spatial, and form perceptions; influencing neonle and talking. The scale has a relia- 1- - 1- - ' - C7 bility of .95. The second factor extracted, "physical dexterity/perceptual ability," con- tains 12 items tapping finger and manual dexterities; motor coordination; complexity of functioning with things; form and spatial perceptions; seeing; reaching, handling, fin- gering, and feeling; setting limits; tolerances or standards; color discrimination; prefer-
140 ence resulting in tangible satisfaction versus prestige; and dealing with measurable or verifiable criteria. This scale has a reliability of .91. The third factor extracted, "physical ac- . . . tivity/working conditions,'' contains nine items tapping climbing; stooping; eye-han(l- foot coordination; hazardous working con- ~litions; lifting, carrying, pulling, an(l push- ing; outside working conditions; noise and vibration; clerical perception; and fumes, odors, dust, and poor ventilation. This scale has a reliability of .89. The fourth factor extracted, "people-things," contains nine items that overlap heavily with those con- tained in the substantive complexity factor. Included items are a preference for activities involving machines as opposed to social welfare; setting limits, tolerances, or stan- ~lards; dealing with people; influencing peo- ple; talking; complexity of functioning with people; measurable or verifiable criteria; seeing; and use of sensory or ju(lgmental criteria. It has a reliability of .92 and runs from things at the "positive" end of the continuum to people at the "negative" end of the continuum. The final factor, "un- cIesirable working conditions," contains three items tapping extreme heat, wet anchor humid working conditions, and extreme cold; it has a reliability of .54. The factor solution of DOT data based on 1980 occupational titles is both similar to en c] different from that reported in Miller et al. (19804. They conducted a similar factor analysis of the DOT data matched to 1970 census data on detailed occupations. The major difference between the solutions ap- pears to revolve around the extent of var- iation in occupational characteristics each set of factors allows. The factor solution reported here provides an additional factor, people-things, on which to (li~erentiate oc- cupations. In ad(lition, the factors (lerived here frequently contain more items than does the Miller et al. solution, owing in part to their more restrictive selection cri- teria for inclusion of items within scales. PAY EQUITY: EMPIRICAL INQUIRIES Measurement of Occupational Labor Market Conditions Measures of occupational labor market conditions were derived from the unpub- lishe`1 1980 Census of Population Subject Report, Occupational Characteristics (PC- 2-7A), provided on tape by the Bureau of the Census from the U.S. Census of Pop- ulation matrices. Potential quantity of labor supplie(1 and (leman(led is measured by the percentage of the experienced civilian labor force (ECLF) that is self-employed and by reserve labor pool the ratio of the number ' not in the labor force who last worked between 1975 and 1979 to the number in the occupation's ECLF. Unemployment rate for each occupation is also included, as is the percentage of government employment. Supply characteristics of occupational in- cumbents (quality of labor supplied) inclu(le mean years of schooling in the ECLF and estimates of mean years of experience. To estimate experience for male occupational incumbents, I subtracted from mean age the mean number of years of schooling plus six for each occupation (see Mincer, 1974, for a discussion of this proxy). For females, I took an analogous figure for each occu- pation and, following Beck et al. (1980), deflated the figure to varying degrees de- pencling on combinations of the proportion of white an(1 nonwhite occupational incum- bents ant] on the proportions of currently married and currently nonmarrie(l occu- pational incumbents. Descriptive analysis revealed that this strategy noticeably low- ere(1 the estimates of experience for female occupational incumbents, although it is best to remember that, particularly for females, the figures should be interpreted as esti- mates of experience. Indicators of market social organization that apply to the ECLF include percent female, percent black, percent Spanish speaking, and percent Asian. To tap the concepts relevant to family and work, I include the percentage of male employe(l
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS persons who are married with spouse pres- ent and the percentage of female employed persons who are married with spouse pres- ent. Data on the percentage of occupational incumbents who are unionized were pro- vicled by Prof. Randall Filer, as taken from Kokkelenberg and Sockell (19851. The union data consist of a 3-year moving average centered on 1980. The percentage of em- ployed persons living in urban areas is also included. Measurement of Annualized Earnings Following Treiman et al. (1984), I use annualized earnings as the dependent vari- able. This measure was calculated by taking a weighted! average of mean male ant] female earnings, where these respective figures have been adjusted to take into account occupation-specific average hours worked per year. To create each sex-specific com- ponent, mean annual earnings was multi- plied by 2080 and (livi(led by the product of the mean hours worked per week (mea- sured during the census survey week) times the mean weeks worke<l per year. The con- stant 2080 reflects the assumption that full- time workers work 52 weeks per year and 40 hours per week. For many occupations, women work fewer hours per week and/or fewer weeks per year than men. Annualizing earnings removes the confounding of time worked by gender since it reflects the level of earnings that they wouIcI obtain in an occupation if all incumbents worker] 40 hours per week for 52 weeks per year. Below, I describe the interrelationships among these variables in preparation for multivariate analysis. Of particular interest are the relationships of each independent variable with the dependent variable, the relationships among the measures of oc- cupational market conditions, and the re- lationships between market conditions and measures of job content. The analytic strat- egy involves multiple regression. ]41 RESULTS Descriptive Analyses Table 6-1 presents the means and stan- dard deviations of the variables used in the analysis, along with zero-order correlations. Looking at all occupations, they contain incumbents who have, on average, 12.7 years of schooling and 17 years of experi- ence. They average 32.3 percent female, 9.21 percent black, 5.51 percent Spanish speaking, and 1.78 percent Asian incum- bents; they are relatively urbanized. An average of 65.3 percent of the male and 52.8 percent of the female incumbents are married. The average rate of self-employ- ment is low, an(1 the pool of reserve ex- periencecI labor averages about 20 percent, as does the rate of government employment. Means for the DOT-basec] composites are 0, allowing for rounding error, since they are sums of Z scores; the standard deviations of Z scores deviate from 1 because these variables are composites and because of rounding error. Inspection of the zero-order correlations is important for several reasons. First, cor- relations between each job content or mar- ket characteristic variable and earnings por- tray these relationships in preparation for multivariate analysis. Male and female an- nualized earnings, male and female expe- rience levels, and male and female years of schooling are omitted from the table because their correlations are very similar to those variables for all occupational incumbents. Hypotheses regarding the zero-order as- sociations of in(lependent variables with an- nualizec] earnings are all supported. Earn- ings is positively associate(l with years of schooling, substantive complexity, self- employment, unionization, and percents government employment, males married, urban, an(l Asian. Negative relationships include those with unemployment rate, re- serve labor pool, and percents female, black, and Spanish speaking. Second, inspection
142 V: - Ct _ V) ._ C~ Ct s~ Ct ._ o o U) o ._ C~ a s~ o s~ _ o o V) o ._ Ct ._ Ct C~ U. a~ 1 C~ ¢ E~ z ~ ,o ~ .= CC ~ C~ C~ - - Ct . s" C~ ~ X o . . ~ C~ CD o o o 1 o C~ _ o ~ CD o ~ C~ - - Ct o D" ;, C) o ~ - o ~ o o o . . 1 ~ _ . . o ,~ C o o . . C~ ~ CO ~r C~ 1 r~ C~ . 1 o ~ 9 C) ~ U) X ~ p:; . o Cc O _ Lr) 0 1~ ~ C~ . . . . ~ C~ ~ CD ~ _ _ C] ~ ~r CD ~ o 8 - o c~ o x o c~ . . ~ 1 o c~ ~ o c~ o o c~ c~ 1 - Lr, o . ~r C5, 1 ~r c~ ~ o ~ . . - - . Ct c o c . . . ~ c~ r~ ~ - ~ - ~ cc . . . X CD - c~ ~ o - c~ ~ c~ ~ c~ ~ c cO g o c~ . - c~ c~ - ~ - . . - . co c~ - ~ cN o ~ 1 1 ~ oo c~ - - ~ CD c~ o 1 c~ o 1 ~; .o L~ ~ c~ c cO c~ c~ o . . ~ 8 oo . . U: - c 5 ~ ._ C) ~ ~ Ct . . oo 0 O O O C~ . . - 1 o 1 O (D C~ O . . 1 O O _ C~ . . 1 1 X C~ C~ . . X CD r~ ~ ~ O . . X 1 o X 1 C~ ~ ~ ~ ._ C~ C) ~ C C~ C~ ~ ~ U: . . C~ _ _ GO X U) C~ X - o C~ o 8, - O C~ 8 0 . . - CD o 1~) 0 X _ O 1 C] C~ C~ 1 CS C~ _ O O . . 1 oo - c~ 1 ~ oo ~ c~ o o . . 1 1 ~ co c~ o . . _ cr, cr, o c~ . . 1 0 c~ CD c~ 1 co c~ o 1 c~ o . c] - - 1 r~ o . 1 u) c~ oo c~ 1 c~ co o . o 1 c~ o CD c~ cs 1 r~ ~ _ ~ ~ o ~ o c~ . . . c~ . - v) bC ,. .5 ~ ~ C) . . C~ ~ ~ _
143 C~ C~ C~ C~ C~ C~ 0 0 0 0 c~ O oo ir: ~ c~ X ~ ~ 0 ~ CD . . . . . . . C~ oo ~ C~ C~ ~ 0 _ _ o o o o o o C~ o o o o o 0 X ~ C~ g o - o 0 o C~ o C~ o ~ C~ 8 ~ ~ o U: o _ o ~ X 0 o ~ C~ C~ - 8 ~ ~ ~ ~ _ C~ X CO 0 ~ ~ ~ ~r _ 0 cc c~ ~ ~r o CO ~ ~ 0 o o ~ ~ o o C~ . . . . . . - o 0 ~ ~ o 0 o C~ _ oo CD 0 0 ~ ~ ir, c~ . . . . . . . 1 1 1 1 1 1 ~ ~ ~r x ~r c~ c~ _ ~ o X _ ~ C~ ~r c~ c~ ~ 0 C~ 0 1 1 1 1 1 ~ r~ co X o) cc 0 CC ~ CD _ C~ ~ _ C~ o C~ ~ _ _ CO 1 1 1 1 1 CD C~ ~ ~ ~ C~ X ~ ~ C~ ~ CD CS U) o CC C~ _ ~ _ . . . . . . . 1 ~ CO et _ ~ ~ ~ ~r 0 ~ ~ ~ ~ oo _ _ o ~ ~ . . . . . . . 1 1 CD C~ ~ CO C~ ~ ~ ~ O CD O CO o C~ ~ C~ _ ~ C~ . . . . . . . 1 1 1 1 1 1 1 CC ~ o C~ CD ~ ~ cD ~ X c~ C~ _ o o o C~ C~ . . . . . . . 1 ~ ~ ~ o ~ ~ ~ Cil ~h CD e~ X O o o o o o o o . . . . . . . 1 1 1 1 _ _ ~ oo C~ oo C~ _ oo ~ (D O ~ oo ~ <.D c~ CN c~ _ . . . . . . . 1 1 1 1 1 1 oo C~ C~ ~ o _ ~ o o o o C~ . . . . . . . 1 1 1 1 C~ ~ X C;, 1= X ~ CC) O X O O 0 ~4 _ . . . . . . . 1 1 1 1 c~ ~ ~ if, cc co io cD ~ oo ~ C~ C~ O C] _ O C~ . . . . . . . 1 1 0 C~ X C~ O X X ~ X O t~ C`) O X C~ ~ C~ C~ CS ~ . . . . . . . 1 1 1 1 1 1 C~ ~ 0 oo CC ~ C5) ~ c~ r~ oo oo ~ CD O ~ O ~ O . . . . . . . 1 1 1 1 ~ ~ 5 ": eC O bC O ~ _ tC .0 =, ~ ,~ . . . . . . . ~r ~ ~ ~ GO ~ O _ _ _ _ _ _ C~ U) _' ~, - C~ o o _ .~ o ~ ,= ,0 U' C) O C~ ~ O O ~ O 40 ·- s~ ~ U, ~ C~ O ~ C~ Ct U, ~ ~ ,9 .Ct ~ C~ :> ~ _ ce O a; =4 ._ 3 ~ a; o o ~ ~: U, a' c~ 5 ~ c~ bt~ _ _ C,) C~2 C~ ~ O O ~ Ct ~ ~ ~ O O C) C;: O O ~ O ~ C) 0 5) · - ~ ~ V O3 ~ ~ o Ct ~ .o _ :, . a; ~o, ~ o ~ "Q . _ ~ Ct o ~ ~ C~ V, o C~ ~ - o ~ V ~o ~ ~ C) Ct C~ o a; · C~ o ~ V, s~ .= ;- ~ ~ a' a; o C~ ~ C~ ~ ~ .o o C~ C~ C~ o U) ~ 3 ~ o ~ . s ° V o o ,, ~ C~ bC O c~ · C) ~Ly s~ ct s~ ~ O 50 c,) s~ O E~ ~ ~ ct . ~ ~ u~ 5 ~ bC O 5 ~ ~ ~ 3 ~ O O ~ .5 v, 5 .= ._ Ct
144 of the correlations among the job content variables suggests negative relationships be- tween substantive complexity and the three remaining indicators, but generally positive relationships among those latter measures, such as between physical dexterity and people-things and between people-things and physical dexterity. Third, correlations among the indicators of labor market conditions suggest relation- ships between the occupational character- istics we are familiar with and more newly developed measures. There are clear re- lationships, for example, between the per- centages of female and racial minority in- cumbents and indicators of market conditions; these relationships suggest the possibility of explaining the often-found negative re- lationships between percent minority and earnings in terms of the occupational labor market conditions analyzed here. Note that the higher the percentages of blacks and Spanish-speaking incumbents, the higher the unemployment and reserve labor pool, and that percent female is negatively as- sociate(l with unemployment but strongly associated with reserve labor pool. The more Asian incumbents, the lower the unem- ployment rate, but there is no relationship with reserve pool. In addition, blacks and Spanish-speaking workers tend to be seg- regated into the same occupations, and those occupations have lower levels of married men. Correlations between years of school- ing and the remaining market variables are often strong; there are negative relation- ships between years of schooling and un- employment rate, unionization, experience, labor reserve, and percents black and Span- ish speaking. Schooling shows positive re- lationships with percents government, Asian, urban, and married mates. As will be seen below, these strong correlations with school- ing influence the results of the multivariate analyses. Finally, consider the relationships be- tween the job content variables and market characteristics. Given the lack of attention PAY EQUITY: EMPIRICAL INQUIRIES to market determinants of occupational earnings, there has been a corresponding lack of understancling regarding the rela- tionships between job content and market variables. Substantive complexity is posi- tively associated with percents government, Asian, urban, and married males and is negatively associated with unemployment rate, unionization, labor reserve, and per- cents black and Spanish speaking. Physical activities often shows relationships of similar strength but opposite in sign. It is positively associated with unemployment rate, union- ization, anc! percents black and Spanish speaking, but negatively associated with percents government, Asian, female, an urban. People-things shows similar corre- lations to those of physical activities. Multivariate Analyses We can now examine these relationships in multivariate terms. Preliminary analyses revealed that multicollinearity between an among sets of variables necessitated exclu- sion of certain variables when others were included. In particular, the .889 correlation between years of schooling and substantive complexity necessitated that these variables not be use(1 in the same equation. Similarly, reserve labor pool is used but unemploy- ment rate is omitted from the multivariate analyses. People-things is also omitted from analyses because of its strong relationship with substantive complexity. Table 6-2 pre- dicts occupational earnings as a function of various combinations of job content ancl/or market characteristic variables. Table 6-2, Equation (1), portrays the im- pact of the job content variables on occu- pational earnings. Substantive complexity (laminates the equation; both the physical (lexterity an(l physical activities composites positively influence earnings, but the effect of physical activities is nearly twice as strong as that of physical dexterity. The second equation investigates the impact of the mar- ket conditions variables on earnings. In this
G _ ~1 1 V, C) C] CS CO X - C~ C~ a, - CO CM X - C~ _ O _ CO C~ C~ CD _ C~ C~ O C~ oo X U: _ (D X _ CD CD O oo C~ O O _ C] O . . . . . . . 1 1 ~ Co X · ~14 ~ ~r C~ C~ o . 1 oo C~ . CM C~ C~ C~ 1 ~ X _ C] O ~ . . . ~ _ 1 C~ C~ o C~ _ O oo ~ ~D C~ ~ 0 O O O . . . X C~ _ . . . oo ~ O _ CO X X O ~ ~ ~r c~i _ _ C~ _ ~ C~ _ O . . o CD U) C~ - O oo ~ _ ~ _ ~ C~ . . . . CO Lr) _ C~ C~ O O . . C~ _ oo C~ ~ u: CS' ~ 11 ~ _ °O C~ 1= C5, ~ .^ O oo o^ 11 C~ C~ CS ~ ~ ~ O C~ O _ _ . . . . 1 1 1 1 _ C~ O CK C~) oo c: . . . . O ~ (D X _ ~ ~ 1 1 1 O ~ C~ O oo a) ~ (D . . . . c~ ~ ~^o o ~ ~ c~ ~ 1 CD ~h CD C~ ~ ~ o . . . C5, ~ _ CD o c~ . ~ ~ c~ ~ - o o . . . - ~ ~ 1 o ~ ~ c~ o ~ c~ o o . . . 1 1 oo o '^O ~ ~ ~ ~ X _ ~ ~ _ ~ CD . . . . . . . o ~ c~ ~ oo ~ O ~A~ ~ _ ~O 11 c~ 1~= . ~ oo . 11 ~Aq ~^c o ~^O ~ ~ o _ a, ~ _ . . . . . . O ~ CD C'] ~^O ~^c ~ o ~ 1 1 1 c~ ~ x i o - ~ u: ~ ~^O x - - ~^O . . . . . ~^c ~ o ~ ~^O co ~^o I c~ ~ ~ ~ ~ ~ ~ c~ o x ~ c~ o o ~ . . . . . ~ ~ ~ cs ~ (D ~ ~ _ . . . . . b~ L^O O ~ ~ - ~ ~ ~ c~ ~ - l vA~ - ,& - o ~ ~ ~ o o · x D . - ro _~ o D O ~ o ~ D 3 _, _, V} U) ~ ~ C~ oo C'1 '^O O O . . ~ C~ C~ X L^O O U: 1 1 CD O ~ ~^o O ~^c . . . ~ _ . - ~ _' X ~ O ~ C~ CC oO _ _ 1 1 , ~ O CO C73 .O ~o U: C~ _' ~ Co C~ C~ oo o - C~ 1 _ C~ ~ O ~ ~ ~ X . . . . ~ _ CC _ C C~ ~ CO C~: 1 1 1 - C~ 1~ _ C~ ~ ~ O O _ ~^O O L^o O O O . . . . . 1 1 1 0 oO O ~A~ ~A~ ~ . . . CD ~ C~ U) 1 1 ~ O ~ ~ ~ ~ ~ ~ ~ bO ~ o ~ ' ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ C~ c; £ :, ~ £ ~ - ~ ~ ~ =. a: o _ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~r CD ~ X ~A~ ~ ~AO _ o . . . . . CO ~ ~ o C'J --' C'] ~ ~ O _ ~ _ ~ LO C~ , CS) ~ _ ~ (D . . . X X _ _ L^O 1 _ _ oo - 1 C~ 11 . ~ ~Ag ~; o 11 . ~ 11 CD C~ 11 1~ _ . ~ CD ~^~ _ . . LO 11 145 V^~ o . Ct C~ o L^o s~ Vj o ~ =- ~ C) £ ° Ct C- ~ 11 .= Z s~ .^ , ~ _ V^~ A~ ~ ^ C;S ct: Ce .o bt .= . - - v~ ~ ._ £ :e ~o ~ =._ o o V, v^~ a; ° v^, _ L4 ,1 o~ .. ._ Ct ~ 5= E~ ~ ~ O ~ U)
146 model, mean years of schooling is predom- inate. The remaining statistically significant predictors have smaller impacts on the de- pendent variable although most of their signs are consistent with the hypotheses. Percent female has a negative impact on earnings, and experience and unionization have positive effects. Percent government employment, however, has a negative effect and is also statistically significant, despite its positive zero-order relationship with the dependent variable. Additional analyses suggest that once years of schooling is con- trolled, the net effect of governmental em- ployment is negative. The remaining market anti social organization characteristics fail to attain statistical significance, but that is not uniformly true in the specifications consid- ered below. Equations (3) and (4) present composites of the job content and market variables. Since substantive complexity and mean years of schooling cannot be included simulta- neously owing to multicollinearity, their effects, and those of the remaining predic- tors common to each model, are included sequentially. Years of schooling serves as a more effective control than floes substantive complexity in the prediction of earnings. Its effect is stronger on the depenclent vari- able, the mocle! in which it is contained explains more variance, and it appears to change the effects of other measures. Com- pare, first, Equations (2) and (31. In Equa- tion (3), years of schooling dominates and percent female, experience, percent gov- ernment, and unionization maintain the ef- fects observed in Equation (21. Physical activities is also statistically significant, al- though its effect is less than half as strong as in Equation (1), in which labor market conditions were not controlled. Comparing Equations (3) and (4), sub- stantive complexity is the most important predictor of earnings in Equation (4), as its "counterpart," years of schooling, was in Equation (31. Note, however, that control- ling for market factors reduces the impor- tance of substantive complexity in compar- PAY EQUITY: EMPIRICAL INQUIRIES ison with Equation (1), and that substantive comnlexitv is not as strong a predictor of earnings as is years of schooling. When substantive complexity is included in the model, the effect of physical dexterity is negative, although small in magnitude. Per- cent female remains as strong a predictor as in Equation (3), but the effects of union- ization, percent government, an(1 experi- ence are diminished, although still impor- tant. Additional labor market characteristics also attain statistical significance in Equation (41. Reserve labor pool exercises a statistically significant negative effect on annualized earnings, and the effect of percent urban is statistically significant and positive. In- terestingly, four additional social organi- zation characteristics of occupational labor markets also attain statistical significance. Percent males married exerts a sizable, pos- itive effect on the depen(lent variable and the effect of percent females married is negative, albeit of smaller impact. Both of these effects are independent of gender composition of occupations. Percent Asian positively affects earnings but percent Span- ish speaking has a negative impact; both of these effects are of moderate importance to the model. The overall explained variance is re(luce(l in this specification as compared with Equation (31. Equation (5) enables us to evaluate the possibility that the pattern of missing data has influenced the findings. The other equa- tions were estimated using pairwise deletion of missing data in order to take advantage of all available information, but Equation (5) estimates the mo(le! shown in Equation (3) with listwise cleletion of missing data. Exclu- sions result in 424 cases available for analysis. Differences between the two equations are generally ones of magnitude, although the effect of percent self-employed is statistically significant for the first time. The effects of years of schooling, percent government, and especially, unionization are reduces! in this mode! as compared with Equation (3), but the effects of experience and percent female
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS are similar across the two models. It appears Mat the pattern of missing data may tent] to increase the effects of certain variables when painvise deletion is used thus, the magnitude of estimates using pal se deletion should be viewed as upper bounds and interpreted with caution. Table 6-3 estimates Equation (3) from Table 6-2 for two additional dependent vari- ables: male and female annualized earnings. For these models, experience and years of schooling are measured on a sex-specific basis. Although years of schooling dominates both models, and neither dependent vari- able is pre(lictec! by any of the remaining job content variables, the ejects of other predictors vary noticeably across the moc3- els. Returns to experience are lower for female earnings than for male earnings. The negative ejects of percents government and female are weaker for female as compared ]47 with male earnings, as are the positive ef- fects of unionization and percent males mar- riecI. Male earnings appear to benefit from reserve labor pool, but female earnings are not affected by it positively or negatively. Female earnings are negatively affected by percents Spanish speaking and self- employed, and positively affected by per- cents Asian an(l urban. These findings sug- gest that although years of schooling dom- inates the equations for annualized earnings presented in both Tables 6-2 and 6-3, a variety of labor market characteristics are important to the prediction of male and female annualized earnings, even when schooling is controlle(l. DISCUSSION Several limitations of the clata influence the analysis. First, the analysis has relies] TABLE 6-3 lob Content and Occupational Labor Market Determinants of Female and Male Annualizer! Earnings Female Annualized Earnings Male Annualized Earnings Independent Variables B B* B B* Substantive complexity Physical dexterity/per- ceptual abilities - 18.14 (12.29) - .048 15.22 (19.68) .024 Physical activities/work- ing conditions 30.16 (22.52) .061 50.13 (37.14) .062 Undesirable working conditions - 21.88 (40.31) - .014 36.01 (65.47) .015 Mean years of schooling 1,597.47 (100.49) .925 2,616.69 (150.75) 1.060 Experience 137.84 (36.66) .136 309.10 (36.55) .271 Reserve labor pool - 9. 73 (7.80) - .042 29.32 (12.80) .077 Percent self-employed - 27.52 (7.65) - .107 - 8.89 (12.85) - .021 Percent unionized 59.10 (5.76) .362 72.19 (9.66) .267 Percent government - 23.76 (4.35) - .191 - 58.26 (6.97) - .282 Percent urban 17.22 (10.20) .062 .68 (17.20) .001 Percent males married 23.91 (9.78) .098 30.68 (17.52) .076 Percent females married - 10.65 (16.54) - .021 - 15.26 (26.83) - .018 Percent female - 7.62 (4.61) - .067 - 35.27 (7.53) - .189 Percent black - 8.86 (17.20) - .018 - 30.05 (27.42) - .038 Percent Spanish speaking - 66.63 (31.42) - .075 - 19.35 (52.14) - .013 Percent Asian 228.33 (63.51) .110 74.10 (104.78) .022 Constant - 12,920.89 (1,889.94) - 22,847.44 (2,740.74) R22 = .754; R2 = .745 R2 = .759; R2 = .750 NOTE: Standard errors in parentheses.
148 on data from the DOT to operationaTize the job content of occupational work activities because there is no other data source from which such measures for detailed occupa- tions can be developed. Cain and Treiman (1981) document difficulties with the DOT that must be taken into account, however. For example, DOT ratings on variables are best viewed comparatively within groups of occupations, rather than across occupations of very (lifferent types. A psychiatrist may have a higher "involvement with people" score than either a social worker or a brick- layer, but a comparison between the psy- chiatrist and the bricklayer on that dimen- sion may be misreading because the construction of the scores was cased on within-establishment, not across-establish- ment, comparisons. Since the psychiatrist and the social worker are more likely to have worked together than either of them with the bricklayer, comparisons across very different occupations are only approxima- tions. Second, the DOT has poor measures of authority relationships, which are very im- portant in evaluations of worth and are likely to vary consi(lerably within and between occupations. The DOT is, however, the only source of data for investigations that com- pare occupations using national samples, and it has been the source of job content measures for all studies with a focus similar to this one. Unfortunately, it does not ap- pear that its limitations will be rectified soon. In addition, measurement of expe- rience, particularly for female occupational incumbents, should be viewed as an esti- mate. Despite the limitations of the data, the inferences produced here are sound and important. This analysis makes three types of contributions to thinking on earnings attainment. The first is a conceptualization of occupational market effects on earnings that enhances thinking within "the new structuraTism" in sociology and also con- tributes to thinking on comparable worth. PAY EQUITY: EMPIRICAL INQUIRIES Drawing on Sto~zenberg's notions of oc- cupational labor markets, the analysis sug- gests a theoretical framework for uncler- stan(ling occupational earnings that highlights three types of factors: job content, char- acteristics of occupational labor market con- ditions influencing potential supply and de- man(l, and occupational labor market social . . Organization. This conceptualization represents a com- prehensive approach to occupational market determinants of earnings, anal it may pro- vide adclitional impetus to further analyses of occupational market differentiation in our economic system. When previous studies of earnings (differences conducted from a comparable worth perspective are viewed from this framework, it becomes clear that they have neglectecl conceptualization of market conditions as compared with skill or job content and that they have faded to consider that "percent female" is but one aspect of occupational market social orga- nization. Although female concentration rightly remains a central element of market social organization, this analysis provicles evidence that racial concentrations, union- ization, and marital status also influence earnings. In addition, given that this study has (remonstrated the usefulness of this ap- proach, additional measures can be (level- oped that are compatible with the per- spective. Researchers could assess, for example, how the number of chfl(lren per occupational incumbent influences earnings and investigate whether this could promote male earnings attainment but hinder female earnings attainment. Within this frame- work, number of children could be viewed as an additional dimension of occupational social organization. Aclditional measures of job content tapping authority position with- in organizations could also be adcled within the framework. Filer's analysis (in this vol- ume) suggests a(lclitional possibilities that might also fit within the perspective. The second contribution of this analysis ~ ~ . 1 ~
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS is the detailed descriptive picture it provides of the relationships among variables tapping job content, market characteristics influ- encing supply an(1 demand, en c] measures of market social organization. For example, these data can be used to summarize the market conditions and job content charac- teristics of occupations with high concen- trations of economic minorities. Female- dominated occupations are low in earnings, experience, percent males married, union- ization, and the job content measure of physical activities. They have high reserve labor pools, are urbanized, and have high black and Asian concentrations. Occupa- tions in which blacks predominate tend to be low paying, have incumbents with lower levels of schooling and higher levels of un- employment and reserve labor pools, fewer males married, lower levels of substantive complexity, and modest levels of union- ization. Occupations in which Hispanics are concentrated are similarly organized. How- ever, they are less urbanized than black- clominated occupations and less female dom- inated. They have a weaker negative rela- tionship with males married and higher rates of unemployment and the job content mea- sure of physical activities. As compared with female-dominated occupations, those in which racial minorities are concentrated are more likely to involve unpleasant working conditions. Another insight gained from these data is that the variables classified as "job con- tent" or as "supply characteristics of oc- cupational incumbents" are sometimes closely related, and those relationships may reflect more general social phenomena. The close relationships among substantive complexi- ty, years of schooling, and earnings consti- tute an extreme and obvious example. The occupations that involve complex work ac- tivities offer greater opportunities for earn- ings an(l either require higher levels of schooling to perform job duties or obtain better eclucated incumbents because em- ployers can afford to select on eclucational 149 attainment. Other important clusters of vari- ables involve the negative relationships among reserve labor pool, complexity, and schooling and the even stronger negative relationships among unemployment, com- plexity, and schooling. These data suggest that those occupations in which incumbents have low levels of schooling an(1 perform work low in complexity are likely to face larger reserve labor pools and, especially, to be subject to greater unemployment than those occupations characterized by more complex work en c] more highly educated incumbents. This descriptive unclerstan(ling of occupational market characteristics will facilitate other studies in which occupational labor market conditions are incorporated in analyses of other dependent variables. Third, the multivariate analyses pre- sented in this chapter have demonstrated the usefulness of the vast majority of the independent variables incluclec] in the mo(l- els. Annualized earnings was significantly predicted by only a subset of the variables included: schooling, experience, unioniza- tion, and percents female and government. Male and female annualized earnings, how- ever, were best understood by considering the fuller range of market and social or- ganization factors. In acIdition to the factors affecting annualizecI earnings, male annu- alized earnings was positively affected by percent males marriecl an(l by the size of the reserve labor pool. Female annualized earnings, in acIdition to being affecter! by the five variables notecl as determining an- nualized earnings, was positively affected by percents Asian, urban, and males mar- ried and negatively a~ecte(l by percents Span- ish speaking and seif-employecT. Social organi- zation an(l other market characteristics appear to be particularly important to predicting female annualizecl earnings. In interpreting the preeminence of years of schooling in predicting earnings, its relationships with the remaining market and social organization characteristics should be kept in mincI. Those occupations with highly educatecl incumbents .
150 have other market characteristics favorable to producing earnings: low unemployment and reserve labor pools; Tow concentrations of females, blacks, and Hispanics; high pro- portions of males married; access to govern- ment employment; and complex work activ- ities. When substantive complexity was substituted for years of schooling in the anal- ysis, additional market and social organization factors were shown to be important to ex- plaining variation in the dependent variable, thus corroborating this idea. Substantively, the findings are generally consistent with the hypotheses, with two exceptions. First, it appears that the effect of government employment turns negative once years of schooling is taken into account, possibly because greater earnings oppor- tunities exist in the private sector for well- educated workers. Second, the existence of self-employment opportunities does not turn out to operate as financial "leverage" for employees across occupations. The strong positive effects of schooling, experience, ant] unionization, however, were expected and consistent across specifications. Other findings were consistent with hy- potheses, although less frequently signifi- cant. The effect of percent males married is a case in point. Although not a statistically significant predictor of annualize(1 earnings, it positively affects both male and female annualized earnings. This finding reflects an additional basis for discrimination in- dependent of percent female, and it may suggest that processes of job allocation, pro- motion, an(l salary increases within estab- lishments work against women when pro- portions of males married within occupations are high. That female annualized earning attainment is also influenced preclictably by concentrations of racial minorities again reinforces the notion that percent female is but one dimension of market social orga- nization that affects female earnings attain- ment. Indeed, the effect of percent female on female earnings is modest, although it is stronger in its effect on male earnings. PAY EQUITY: EMPIRICAL INQUIRIES Because many ofthese measures are new- ly constructed or have not been user] in earnings research previously, questions re- garding their interpretation remain. The effects of percent Asian are a case in point. In the analysis, Asian concentration had a stronger impact on earnings when education was not controlled, which suggests that the higher educational attainment of Asian in- cumbents partly accounts for this social or- ganization effect. The analysis also found, however, that female occupational earnings is greater the higher the percent Asian in the occupation. There may be unmeasured productivity factors that explain this, such as job skills not captured by the DOT vari- ables, or it may be that wage rates in oc- cupations in which Asians are concentrated have historically been higher owing to scarc- ity of labor or custom. Additonal research should attempt to adjuclicate among these possibilities. Future analyses should also consider the nature and extent of changes in these earn- ings processes over time. Parcel and Muel- ler (1989) compare the models discussed here with those generated for the 1970 time period in an effort to unclerstand the stability of these market processes during the 1970- 1980 time period. Additional investigations are needed that use these macro-level char- acteristics in conjunction with micro-level data sets to determine more fully how oc- cupational markets operate as important contexts affecting the socioeconomic stanc3- ing of indiviclual workers. Sociologists have been interested in such mo(lels for some time (e.g., Parcel and Mueller, 1983), and these newly developecl measures will pro- vicle an acl(litional resource with which to pursue such investigations. ACKNOWLEDGMENTS I wish to thank Mohammad Siahpush and Gopal Singh for research assistance, Robert Kaufman and Lauren Krivo for helpful sug- gestions, an(l especially, Richard Haller for
COMPARABLE WORTH, LABOR MARKETS, AND EARNINGS essential contributions to data production. Paula England provided the data matching DOT categories with 1980 U. S. census cocles, and Randall Filer provided the data on unionization. John Priebe at the Bureau of the Census facilitatecI my obtaining the un- publishec] data on occupations from the 1980 census. I retain responsibility for remaining errors. REFERENCES Baron, J. N., and W. Bielby 1980 Bringing the firms back in: Stratification, segmentation, and the organization of work. American Sociological Review 4545~:737-765. 1984 The organization of work in a segmented economy. American Sociological Review 49(4):454-473. 151 eds., New Concepts in Wage Determination. New York: McGraw-Hill. England, P., and G. Farkas 1986 Households, Employment and Gender: A So- cial, Economic and Demographic View. New York: Aldine. England, P., and B. Norris 1985 Comparable worth: A new doctrine of sex discrimination. Social Science Quarterly 6643~:629-643. England, P., M. Chassie, and L. McCormick 1982 Skill demands and earnings in female and male occupations.Sociology and Social Re- search 6642~:147-168. Freedman, R. B., and J. L. Medoff 1984 What Do Unions Do? New York: Basic Books. Hartmann, H. I., ed. 1985 Comparable Worth: New Directions for Re- search. National Research Council, Com- mittee on Women s Employment and Related Social Issues. Washington, D. C.: National Academy Press. Beck, E. M., P. M. Horan, and C. M. Tolbert II Hodge, R. W., and P. Hodge 1980 Industrial segmentation and labor market dis- 1965 crimination. Social Problems 28~2~:113-130. Bergmann, B. R. 1971 The effect on white incomes of discrimination Hodson, R. in employment. Journal of Political Economy 7942):294-313. Bibb, R., and W. Form 1977 The ejects of industrial, occupational, and sex- stratification on wages in blue collar mar- kets. Social Forces 55~4~:974-996. Bielby, W., and J. N. Baron 1984 A woman's place is with other women: Sex segregation within organizations. Pp. 27-55 in B. F. Reskin, ea., Sex Segregation in the Workplace: Trends, Explanations, Remedies. National Research Council, Committee on Women's Employment and Related Social Issues. Washington, D.C.: National Academy Press. Bureau of Labor Statistics 1986 Employment and Earnings 33~1~. Cain, P. S., and D. J. Treiman 1981 The Dictionary of Occupational Titles as a source of occupational data. American Socio- logical Review 46~3~:253-278. Mincer, J. Corcoran, M., and G. J. Duncan 1979 Work history, labor force attachment and earnings differences between the races and sexes. Journal of Human Resources 1441~:3- 20. Dunlop, J. 1957 The task of contemporary wage theory. Pp. 3-30 in G. W. Taylor and F. C. Piersons, 196.~ Occupational assimilation as a competitive process. American Journal of Sociology 7143~:249-264. 1984 Companies, industries and the measurement of economic segmentation. American Socio- logical Review 49~3~:335-348. Kerr, C. 1954 The balkanization of labor markets. Pp. 92- 110 in E. W. Bakke, P. M. Hauser, G. L. Palmer, C. A. Myers, D. Yoder, and C. Kerr, eds., Labor Mobility and Economic Oppor- tunity. New York: McGraw-Hill. Kokkelenberg, E. C., and D. R. Sockell 1985 Union membership in the United States, 1973-1981. Industrial and Labor Relations Review 3844~:497-543. Miller, A. R., D. J. Treiman, P. S. Cain, and P. A. Roos, eds. 1980 Work, Jobs and Occupations: A Critical Re- view of the Dictionary of Occupational Titles. National Research Council, Committee on Occupational Classification and Analysis. Washington, D.C.: National Academy Press. 1974 Schooling, Experience and Earnings. Na- tional Bureau of Economic Research. New York: Columbia University Press. Parcel, T. L., and C. W. Mueller 1983 Ascription and Labor Markets: Race and Sex Differences in Earnings. New York: Academic Press. 1989 Temporal Change in Occupational Earnings
152 Attainment, 1970-1980. American Sociolog- ical Review In press. Parcel, T. L., S. Cuvelier, J. Zorn, and C. W. Mueller 1986 Comparable Worth and Occupational Labor Market Explanations of Occupational Earn- ings Differentials. Paper presented at a meet- ing of the American Sociological Association, New York City. Remick, H. 1984 Preface. Pp. ix-xv in H. Remick, ea., Com- parable Worth and Wage Discrimination. Philadelphia: Temple University Press. Robinson, R. V., and J. Kelley 1979 Class as conceived by Marx and Dahrendorf: Effects on income inequality and politics in the United States and Great Britain. Amer- ican Sociological Review 4441~:38-58. Snyder, D., and P. Hudis 1976 Competition, segregation and minority in- come. American Sociological Review 4142~:209- 234. Stolzenberg, R. M. 1975 Occupations, labor markets, and the process of wage attainment. American Sociological Review 40~5~:645-665. Taeuber, A. F., K. E. Taeuber, and G. C. Cain 1966 Occupational assimilation and the competi- PAY EQUITY: EMPIRICAL INQUIRIES live process: A reanalysis. American Journal of Sociology 7243):273-285. Treiman, D. J., H. I. Hartmann, and P. A. Roos 1984 Assessing pay discrimination using national data. Pp. 137-154 in H. Remick, ea., Com- parable Worth and Wage Discrimination. Philadelphia: Temple University Press. U. S. Department of Labor 1977 Dictionary of Occupational Titles 4th ea., Washington, D.C.: U.S. Government Print- ing Office. Waldaur, C. 1984 The non-comparability of the comparable worth doctrine: An inappropriate standard for de- termining sex discrimination in pay. Popu- lation Research and Policy Review 342~:141- 166. Wallace, M., and A. Kalleberg 1981 Economic organization of firms and labor market consequences: Toward a specification of dual economy theory. Pp. 77-117 in I. Berg, ea., Sociological Perspectives on Labor Markets. New York: Academic Press. Wright, E. O., C. Costello, D. Hachen, and J. Sprague 1982 The American class structure. American So- ciological Review 4746):709-726.