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.
Pay the Man. EFects of Demographic Composition on Prescribed Wage Rates in the California Civil Service JAMES N. BARON and ANDREW E. NEWMAN Debates regarding comparable worth have rekindled interest in whether "women's work" is undervalued relative to commensurate work done by men. Various laboratory ex- periments have demonstrated that men and women ascribe less worth to work per- formed by women (see Deaux, 1985; McArthur, 1985; Major et al., 19841. Other research has consistently reported negative associations between the proportion of wom- en in an occupation and reward levels, even controlling for average occupational skills and vocational requirements (see Parcel et al., 1986; Treiman and Hartmann, 1981~. Unfortunately, there have been fewer stud- ies of how gender composition shapes or- ganizational pay decisions. From both a theoretical and policy stand- point, it would clearly be helpful to know more about pay decisions within actual work organizations, rather than having to infer gender biases in organizational personnel systems from laboratory studies or aggregate research on occupations. Several recent lon- gitudinal studies of organizations have doc- umentec3 persistent negative effects of the percentage female in a job on the pay of incumbents, even controlling for workers' 107 skills and job requirements (Pfeffer and Dav- is-Blake, 1987; Rosenbaum, 19851. These few studies are thus consistent with findings from the laboratory ant] occupational re- search mentioned above. The nature of the relationship between gender composition en c] organizational pay practices, however, is still not well under- stood. As Hartmann et al. (1985:18) note, "pay anal salary levels of different jobs within an enterprise are not automatically deter- mined by the operation of abstract forces such as . . . 'low valuation of women's work'.... tPay] decisions are poorly un- derstoocI, partly because it is clifficult for researchers to obtain access to data on in- dividual firms. This paper tries to help fill that void by examining how the prescribed pay rates of jobs vary as a function of the demographic composition of incumbents in those jobs. The data we analyze have several potential advantages. First, they characterize pre- scribed pay rates for jobs (rather than the average pay of incumbents), thereby pro- vicling a measure of job "worth" that is independent of variations across jobs in in- cumbents' human capital. Second, they en-
108 able us to examine the effects of race and sex composition on pay rates. Third, they permit fairly precise controls for the training and task requirements of positions in as- sessing the net effects of demographic com- position on pay. Fourth, they cover multiple time points, which enables us to examine whether changes in (lemographic compo- sition affect the relative worth of jobs. The (lata describe pay structures in Cal- ifornia state agencies. The California civil service is arguably as rationalized and sub- ject to intense scrutiny and pressures for egalitarian reform as any personnel system in the private sector. State governments may also have greater slack with which to absorb pay equity costs than most Private enterprises do (Killingsworth, 1985~. Our results, therefore, provide a useful bench- mark in assessing the magnitude of pay inequities, because pay and personnel prac- tices in the private sector are probably less egalitarian (in general) than in California's state government. THE SETTING The executive branch of the California state government consisted in 1985 of ap- proximately 120 stancling agencies, boards, and commissions and several dozen tem- porary, special-purpose commissions. Total employment in the 120 regular agencies was approximately 150,000 as of March 1985- over 120,000 full-time civil servants and roughly 30,000 part-time and temporary workers. California state agencies range in size from commissions with fewer than 10 full-time stab members to agencies with over 10,000 employees each, such as the Departments of Transportation and Cor- rections. In March 1985, 90 agencies had 10 or more employees; the mean was ap- proximately 1,350 employees ant] the me- (lian was approximately 275. Each agency consists of a chief executive, appointed by the governor, and (in larger agencies) one or two assistant executives, who are also political appointees. The remaining full-time PAY EQUIP Y.: EMPIRICAL INQUIRIES Stan in each agency consists of civil servants enjoying tenure rights. California's civil service merit system con- sists of formal rules and policies governing every aspect of personnel management. These rules and policies were first set out in the California Civil Service Act, passed in 1913. This act calle(l for the use of merit principles in staking state agencies and created a Civil Service Commission to administer the new merit system en cl ensure compliance with its principles by individual agencies and boards. The commission was replaced in the mid-1930s by the State Personnel Boars] (SPB), which continues to administer the system. The Department of Personnel Ad- ministration (DPA) was created in 1981 to administer the civil service salary system en cl has gra~lually gained jurisdiction over additional personnel matters formerly man- aged by the SPB- most notably, position classification. Approximately 5,000 job titles exist across state agencies (although many are without incumbents at any given time) and are or- ganized into the job classification (or cIass) system administered by the DPA and SPB. The DPA an(l SPB also maintain class spec- ification sheets, which describe the duties and qualifications for each class. Other per- sonne! functions, including testing and re- latec] aspects of selection, have been clel- egated to individual agencies over the past several years. The result has been consid- erable decentralization of the California state civil service system and a gradual transition of the SPB from a centralized personnel authority to a body whose responsibilities are increasingly limited to monitoring and oversight. The decentralization has by no means been complete, however. The SPB contin- ues to administer employment activities in- volving agencies too small to have their own personnel function. Moreover, the SPB is responsible for administering "service-wi(le classes" jobs that are used by many agen- cies statewide (e.g., general clerical cIasses) and that are thus subject to centralized SPB
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE control rather than the discretion of indi- vidual agencies or subunits. The SPB is also responsible for ensuring agency compliance with governmental regulations concerning equal employment opportunity, affirmative action, and pay equity. There is one noteworthy difference be- tween the California civil service and the federal government (and other jurisdictions organized along similar lines). The federal civil service system incorporates General Schedule (GS) grades and subdivisions that link educational and experience require- ments to pay rates in a standardized manner across disparate occupational series. In the California civil service system, however, each job is assigned to a particular pay range, but there is greater flexibility across oc- cupations in how this is done and in the . . . precise minimum anc maximum pay as- signed, even for jobs nominally allocated to the same pay-range category. Consequent- ly, there appears to be greater latitude for the demographic mix of incumbents to affect perceived position worth in the California system than in the federal government and other systems having more uniform criteria for mapping job titles to pay ranges. HYPOTHESES Based on previous research, we hypoth- esize that prescribed pay rates are lower in positions staffed disproportionately by wom- en or ethnic minorities, even after con- trolling for job content and requirements. We also hypothesize that the entry of wom- en or minorities into jobs devalues them, that is, reduces their relative pay. Accord- ingly, we supplement our cross-sectional results with analyses of data at two time points in order to examine whether changes in demographic composition produce changes in a job's standing within the wage hier- archy. Organizational theorists posit that the en- · · ~~ · · 1 ~~ v~ronment Is ~mpr~ntect on organizations at the time of founding and that arrange- ments appropriate to that period persist 109 through structural inertia (Hannan and Freeman, 1984; Stinchcombe, 19651. If this is true for jobs as well, then positions created more recently should be more likely to reflect contemporary concerns with pay eq- uity, whereas older jobs should be more likely to be devalued if held by women and minorities. Therefore, we report separate cross-sectional results for newly created ver- sus "enduring" positions, which enables us to assess whether tendencies toward de- valuation are less pervasive in recent cohorts of jobs. Finally, we hypothesize that the effects of changing demographic composition on pay rates depend on the extent of employ- ment growth in a job. One common expla- nation for lower pay in occupations having a disproportionate share of women or racial minorities is the "crowding'' that occurs when lower status workers enter a line of work and drive down wages (Bergmann, 1974; Hodge and Hodge, 19651. By creating excess supply, it is argued, women and minorities compete against white men and thereby lower wages. Although it is difficult to operationalize and measure the relation- ship between labor supply and demand in a given job (see Nakamura and Nakamura, in this volume), this line of argument sug- gests that the entry of lower status workers might devalue a job most when employment is declining. Moreover, it is likely that the entry of women or nonwhites into a job would be more salient under conditions of decline than under expansion, exacerbating the tendency for their entry to diminish the perceived worth of the position. In contrast, rapid growth in a job might ease competitive pressures and crowding, thereby reducing the tendency for the entry of women and minorities to depress wages. DATA AND METHODS The Sample We analyze data describing staffing pat- terns, published pay schedules, and other
110 characteristics of all jobs having incumbents in the California state civil service system in two time periods. As of March 31, 1979, there were 2,844 jobs in use and 113,569 full-time incumbents. As of March 31, 1985, there were 3,188 jobs and 123,212 full-time incumbents. The data files provide infor- mation on the staffing patterns (by ethnic group an(1 sex) and the wage range (i.e., the minimum and maximum pay prescribed by the civil service) for all jobs in every agency. We analyze data describing state employment as of both dates, and in our longitudinal analyses, we look at the effects of changes in demographic composition on prescribed pay rates over that 6-year period. Because California is currently involved in a suit brought by the California State Employees Association (Service Employees International Union), the largest compara- ble worth lawsuit in U. S. history, the results of our study have added relevance. We coincidentally obtained these data from the state for basic research purposes just before the lawsuit was filed. We have had regular conversations with both parties to the liti- gation, keeping them informed about our analyses and seeking answers to specific questions about the state personnel system, union activities, an<] the like. We have been steadfastly nonpartisan throughout the lit- igation, however. (The state die] not impose any quid pro quo for providing us with the data other than requesting that we apprise them of our results and honor certain routine confidentiality requirements.) Operationalization We operationalize job worth as the pre- scribed monthly starting pay in a job class. ~ Unlike a measure based on the average wage received within each job, our oper- iWe conducted supplementary analyses measuring job worth in terms of the prescribed maximum (rather than minimum) pay, as well as in a logarithmic metric. The results were unchanged. PAY EQUITY: EMPIRICAL INQUIRIES ationalization is not affected by sex or race differences in human capital, seniority, or productivity within a job classification. Note that if jobs dominated by white men provide greater chances of beginning employment above the prescribe(l starting pay or of ascending the pay lad(ler more quickly, those sources of advantage are not captured by our analyses. Thus, we are conservative in assessing the extent to which work done by women and minorities is "clevaluecl." We measure sex composition as the per- centage of full-time incumbents in each job who are female. Race composition is mea- surecI by three variables indicating the per- centage of full-time job incumbents who are black, Hispanic, or in "other" ethnic mi- nority groups (mainly Asians within most job categories). We also look at interactive effects of sex and race composition. One might expect, for instance, that although a predominance of women or minorities in a job lowers its perceived worth, there is no additional "penalty" (i. e., interaction effect) associated with having a predominance of female minorities. Accordingly, we also in- clucle variables denoting the percentage in each job who are black females, Hispanic females, and females belonging to other . . minority groups. We control for job content somewhat in- directly because California has no formal point system for judging job worth. To date, the state has gauged pay equity principally by comparing "the results . . . of compa- rable worth or point factor studies per- forme(1 in other jurisdictions to current sal- ary relationships of classes which have similar duties an(l responsibilities within State ser- vice" (California State Department of Per- sonnel Administration, 1982:141. Lacking formal ratings of job worth, we can only conduct an indirect and preliminary stu(ly of this facet of comparable worth. We do so in two ways. First, we rely on the occupational classification system used by the state civil service system itself. Our regression analyses inclu(le vectors of dum-
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE my variables (at varying levels of detail) that control for detailed job familiesjobs that the state civil service identifies as having comparable duties and requirements (see the appendix). This procedure may under- state pay inequities to the extent that people are segregates! across jobs for reasons un- related to skills, abilities, ant] voluntary choices. Therefore, for a subset of jobs, we report somewhat more refined analyses that control for the detailed educational and ex- perience requirements of jobs listed in of- ficial job descriptions for a sample of state agencies and civil service jobs. If differences in prescribed pay across job families solely reflect differences in job requirements, then controls for educational ant] training de- mancis should eliminate (or severely reduce) any net effects of job family on base pay rates. An effect of demographic composition on pay rates net of experience and educa- tional requirements would be prima facie evidence of unequal pay for comparable work. Job growth is measured by the percentage increase in employment between March 1979 and March 1985 within each job class (for classes having one or more incumbents at both time points). CROSS-SECTIONAL ANALYSES Effects of Demographic Composition Table 5-1 provides descriptive statistics for base pay rates anc] demographic com- position for three sets of jobs: positions having incumbents as of March 31, 1979, positions with incumbents as of March 31, 1985, and positions that were occupier] at both time points (i. e., encluring jobs). The statistics reported in parentheses weight each job according to its number of incum- bents; the other statistics are unweighted. These descriptive statistics are provided principally to assist readers in interpreting the magnitude of elects reported in the regression analyses that follow. ~1 In 1985 the average full-time California civil servant was in a job in which the entry pay rate had increased roughly 38 percent since 1979. As seen from the second] row of Table 5-1, however, adjusting 1979 pay rates to 1985 clolIars reveals a real decline in pay rates over this period, a result that is not altogether surprising given similar trends in the private economy and the im- pact of tax reform (Proposition 13) on the state budget.2 '~ ~ ~ r Doughy tour out of nine full-time civil servants in 1985 were female, ant! over 34 percent were minorities, up from 24 percent in 1979. Just over half the black employees were female, and women represented just under 50 percent of the remaining ethnic categories. Table 5-2 reports results from ordinary least-squares (OLS) regressions in which the dependent variable is the prescribed 1985 monthly minimum starting pay in each job classification (in clolIars). Coefficients in pa- rentheses are from regressions that weight each job according to its number of incum- bents; those results convey relationships that pertain to the typical civil service work- er, rather than the typical civil service job.3 Table 5-2 summarizes relationships among 3,188 jobs that employed 123,212 full-time workers as of March 31, 1985. Because California has no systematic pro- cedure for quantifying job requirements, we attempt to capture compensable differ- ences in job content by aching control vari- ables at increasing levels of occupational 2The Bureau of Labor Statistics calculated the Con- sumer Price Index in California to be 216.2 in 1979 and 320.0 in 1985 (1967 = 100), which implies a 48.01 percent cost of living increase over the 6-year period (California State Department of Finance, 1986:61). We have thus inflated 1979 pay rates by that amount in adjusting them to 1985 dollars. 3In all weighted analyses, we resealed the weights to sum to the actual number of jobs involved. This ensures that weighted effects do not appear statistically significant simply because the sample size is inflated when jobs are weighted by employment.
At Coo. ·_I At - . At pa At At By - o ~ - ~ - ~ ~ - At o ~ - * o no ~ - 5 Ct .~ be o a: . . in c) ._ V) ._ Cal Cal .> .= V: C) ,= a' 0 C) ._ ._ U: - ._ C), As, == ~ . ret no ~ ~ a . ~ ~ . ~ ~ .= _ Con . .Q . 0 _ C) ~ ~ a 0 - .= _ _ X X . CD CO (D Cal CO Cal X _^ Cal CO CO Cal _ U: _ O CD to X ~ ~ i . . . . . . . ~ ~ rat x co X C) ~ Lr) ~ CO X ~ _ _ ~ ~ ~ O . . . . . . . C~ U: CD ~ X ~ X C~ _ _ _ ~ ~ X _ c~ r~ CO ~ X C~ CD ~ ^ C~ ~ C~ ~ ~ C~ ~ O _ C~ O `-D ~ ~ _ ~ _ _ _ ~ ~ ~ _ _ ~ U: ~ C;) ~ CD ~ oO X ~ _ C~ ~ ~ O CS) . . . . . . . . . (D CD ~ ~ 00 ~ ~ C~ C~ oo ~ C~ ~1 ~ CD O u, O ~ C] ~ ~ C~ O ~ oo ~ C] O _ . . . . . . ~ O X X CD ~ ~ ~ ~ ~ ~r c~ C~ ~ ~ _ C] oo CO C~ ~ ~ ~ CM ~ C~ oO O C~ C';i '~ C~~ ~ c} ~ ~ C~ _ _ _ 1~ ~4 oo X C~ C~ ~ ~ ~ ~ O O ~ _ CO ~ C~ O X C~l ~ 1~ ~ ~ ~ ~ ~ o O C~ oo ~ ~ ~ C~ C~ ~ _ C~ oo ~ ~ ~ ~ ~ ~ ~ C~ (= ~ _ _ C'1 I.~ h O ~_ ~ ~ _ O . . . . . . . . . O ~ C~ U: ~ ~ _ o _ _ ~ _ X (D _ C~ - C~ COo oo _ C~ ~ C~ ~ ~ O ~ C] a) =4 ~) =) O C~ C~ LO . . . . . . . . C~ ~ ~ ~ ~ CD CC CC ~ CO ~ ~ ~ ~ ~ ~ X CM ~ X ~ C~ X X ~ C~ ~ (D ~ ~ CO CD CC X O ~ CD ~ C~ _ _ _ _ Q~ X C~ X _ ~ (= CO O o C<~ C~) C~ C~ _ C<l C~1 X ~ C~ C~ C~ O ~ _ C~ O CO ~ ~ _ CD ~ _ _ _ X - ~ / O X O LO CO ~ X C~ _ C~ ~ O . . . . . . . . O0 ~ C~ C~ C~ X C~ C~ ~ x r~ 0 c~ ~ O c~ 00 ~ X ~ C~ C~ . . . . . . . . . _ CO ~ 00 CO ~ ~ ~ ~ 00 _ C~ ~ ' ~ ~ `, c~ 0 C;) O r~ ~ ~ ~ oo ~ ~ ~ ~ ~ C~ ~ ~ . . . . . . . . . x c~ ~ co cO ~ ~r cs C~ C~ _ ~ _ ~ =) ^ X x c~ ~ ~ c~ ~ 0 ~r co _ O U) ~ _ . . . . . . . . . _ =, _ C~ X 00 ~ ~ ~ ~ ~ ~ ~ ' ~ C~ _ _ ~ _ ~ X ~ ~ O ~ ~ C~ ~ ~ U) C~ X CD C~ _ 0 3) CD ~ ~ ~ ~ ~ ~ C~ O _ O ~ _ X _ C~ ~ bC ~ bC _ t~ _ e3 ~ Ce ceS - :~ C) _ C~ ~ ~ (S L4 C) ~ CC ~ - v ~ v C~ ~ ~ C-s P9 ~ O O ~ ~ ~ O ~~ ~ ~ ° ~ == ° ~ ~ a; ~ ~o £ ~ ~ ~ ~ ~ V V) V~ V V V V V V V 4 - 40 a~ G) s~ s~ ~ ~ s~ s~ ~ 0 0 50 50 ~ a: ~ ~ ~ c; ~ ~ C4 ~ ~ ~ ~ ~ ~ ~ Z Z =5 .r, C~ ._ V) ._ C~ U: _ o C~ C) m _ 0 v C) V) C~ V) C~ _ O 00 _ 0 ~_ _ V: C) 3 - _ CD .. , ~ X _ C) , C) 5 C: ~4 ~0 ;^ . c~ a: ._ ~ U) 0~ .= ~o £ 5~ C_, ~ ._ ~ Ct £ C; ~ ._ ~ =, ~ C) bC _ 3 =- a; ~ C~ ~ ~ o - X C~ 13) C) s~ V, ._ s C~ ~ ~ Z U) .. v, u) ~ ~ ~o o ~ z ~
· - - c~ CJ) 1 := ~ an A ~ ~ - o o · ~ - v) o · ~ - o v o o c) At ~ ~ to ~ - c] ~ ~ ~ - ~ - ~ ~ - -¢ · at =; ~ - o - - at - ~ c; if · a- ~ o~ - o c; - - . - v: . - c; - ~ - - - be - o co o - co o Do en . . . . G) _ Cal O 1 1 1 ~ ~ ~ 1 . 1 en x - 1 - ~ en x ~ ~ . . . 'o ~ ~ 1 1 1 - ~ ~ LO - cO o en x A - en 1 1 1 c C~ G. CD CD ~ 1 1 1 CD ~ C~ _ . . . . C~ ~ _ C~ C~ _ C~ _ 1 1 1 1 o ~ c~ G) c~ (D 0 CO 1 1 1 1 CD _ ~ r~ . . . X 1 1 ~ 1 ~ _ X _ ~ C~ G: ~D 1 1 1 1 C~ C) C~ C,) t_ ; ~ X ~ C: ~ X C~ X X > C~ C5 ~ C) ~ C~ ~ U) C~ U: ~ m ~ ~ ~ 113 x o x * ~ _ CD ~ CO _ C~ r~ . . * X C~ C~ o 1 1 * x C~ o * * ~ o . . C~ o 1 1 ~ * CC X X X C~ CC ~ _ . . . . _ CC ~ C~ _ _' C~ ~ C~ _ oo ~ C~, C~ x - 1 oc 1 + C) . C~ · X bC ~ CO C~ ~ o _ C~ * * ~ CO _ _ . . _ o 1 * * o X CM . . C] C~ * * U: C~ X o _ o 1 1 ~ * C~ C~ X . . U: _ C] C~ o C~ C~ C~ ~ C) C~ ~ _ C) CM _ Cc . . . . X CO O CD C~ 1 1 1 ~ ~ _~ I ~ C o U) X o s~ x ~ r~ ~ c' ~ · 1 1 1 1 _ _ o (~] m. ~ x ~ ~ 1 - 1 ~' 1 ~ ~ _' C ., X ~ _ _ C':, ~ bt _ ,C, ~ . C~ ~ ~ bC X ~ X . ~ ~ V, x oc ~r co c C) ~ CO ~ ._ It · C) 1 1 1 1 bC '_ X CO ~ X X ~ CD _ ~ _ 1 _ 1 ~ 1 C~ CD ~ CC . . 1 1 ~r CO 1 ~ ,~, ,~, . ~ ~ C C) C) C~ C Cd C ~ ~ + G) ~ ~ O u: _ ~ C = - X ~ =' ~ C,) X C ~ CX) 0 ' - ~ C~ X _ ~ bC ~ ~ . . ~ t_ CO C~ C) CO CD 1 o C~ 1 - C~ CO CO 1 U) 1 t ~ I C) _ <,) O ~ C~ C: + C C ~ X C~ ~ 0 + C) _ ~ _ ~ ~C ~ X CC, ~V} C: C ~ X ~ ~ _ - c o - C~ ~ _ .C~ ~ C~ ~ ~ _ C, · ~ ~i - ~ , C C~ bC . _ bf ~ C.) ~ C) bt .o . . _ ~ C) . .) C C~ ._ C) ~ ~ C.) _ C V, C C) ~ C C) C C; C~ ~ _ _ ._ ._ ~ .o .o ~ ,~ N C~ bO ~ U, ._ .' ~ ~ .o ~ ~ X X X ._ ~ ._ C ~ C ~ .. c, ~ C) C) C) O ~ C) C~ Z U) C~ C~ t: ~ ~ - o C) o V, bC - (~0, C) - Ct - o ·_ - · _ - C V) - C) ·_ C) C __ - C - ;> C: o - C~ C~ - ._ V2 V: - ._ C: CC~ C~ - C - . ~ LO o A *
114 detail.4 Accordingly, Table S-2 shows the effects of sex and race composition on pay rates under various specifications. Models A through D regress pay rates on various combinations of demographic variables: sex only (model A), race only (mode! B), main effects of sex and race (model C), and main and interaction effects of sex and race (model D). Model E adds controls for the 20 major job categories used by the state for reporting to the Equal Employment Opportunity Commission (EEOC). (See the appendix for job categories.) These controls capture dif- ferences in pay rates associated with major families of work roles, such as "supervisory clerical," "clerical," and "semiskilled" work- ers. Mode! F controls for 99 major schematic classes characterizing specific categories of work activity, such as "machine operations" within "office and allied services" or "police and law enforcement" within "regulatory and public safety" (see appendix). Mode} G combines the occupational controls con- tained in models E and F. Finally, to pro- vide the most detailed controls for job con- tent, mode} H includes controls for 281 detailed schematic classes, which represent detailed subdivisions of the 99 categories controlled in models F and G. These cat- egories demarcate very specific job cate- gories, such as "duplicating" within "ma- chine operations" or "fish and game" within "police and law enforcement" (see the ap- pendix). Model I adds dummy variables for the major EEOC categories to these de- tailed occupational controls. Note that our controls for detailed oc- cupation may absorb some effects of de- 4Not surprisingly, there is considerable sex and race segregation across occupations, particularly at detailed levels of disaggregation. Consequently, there is some- times extreme collinearity in models that attempt to disentangle effects of demographic composition from those of occupational differentiation. This reduces the chances of discovering statistically significant demo- graphic effects and also suggests that the Precise point estimates we report may be unstable. . . vv 1 . 1 1 PAY EQUITY: EMPIRICAL INQUIRIES mographic composition that are of interest. In other words, our analyses only assess comparable worth within occupational fam- ilies, not across disparate types of jobs, such as nurses versus gardeners. All differences in pay rates across cletailecI occupational groupsincluding some that may have to JO with the sex and race of incumbents- were removed from moclels F through I before we assessed the effects of demo- graphic composition. Thus, our assessments of bias against female- and minority-domi- nated jobs5 are quite conservative, and we report results from mo(lels that control for detailed occupation only because those con- trols very much stack the (leek against fincI- ing net relationships between (lemographic composition and pay rates. Yet, even after inclucling the most (le- tailed occupational controls, we fine] strong effects of race and (especially) sex compo- sition on pay rates (Table 5-2~. Jobs dom- inate(1 by men pay consi(lerably more than otherwise comparable jobs clominatec] by women, and jobs dominated by whites pay substantially more than jobs dominated by blacks, Hispanics, or members of other eth- nic groups.6 Moreover, these effects tend to be stronger, rather than weaker, when jobs are weighted by the number of incum- bents, which suggests that it is actually in the most populous jobs within the civil service that these effects are manifested, rather than in a few anomalous single-person classes. Controls for occupational categories reduce but by no means eliminate these emographic effects. The metric effect of sex composition in the most elaborate spec- ification (model I) is nearly half the size of the bivariate effect (mocle! A). Controls for awe use the terms female-dominated and minority- dominated to refer to jobs with disproportionate num- bers of women or nonwhites, rather than meaning to imply that females or minorities are necessarily nu- merically dominant in any given classification. 6In supplementary analyses, we found no systematic evidence of nonlinearities in the effects of percent female or percent nonwhite on pay rates.
WAGE RATES lN THE CALIFORNIA CIVIL SERVICE occupation absorb somewhat more of the effects of race composition. Even with the most detailed controls for job content, how- ever, the effects of race and sex composition are substantively and statistically significant. The results in Table 5-2 also suggest that, for the most part, the effects of sex and race composition on pay rates are adclitive. One apparent exception concerns the weighted effects for "percent other female," which indicate that jobs staked disproportionately by women in this minority group (predom- inantly Asians) involve smaller penalties than are implied simply by the additive effects of sex and race composition. Given the high levels of eclucational and labor force attain- ment among California's Asian work force, this effect is not altogether surprising. Cer- tain models in Table 5-2 also indicate that in jobs dominated by black women, there is some "reclundancy" or overlap in the penalties associated with the presence of women and blacks. (Some numerical ex- amples of the pay penalties associated with (differences in (lemographic composition are provided below.) Educational and Experience Requirements The results in Table 5-2 might, however, merely reflect fine-grained segregation within job ladders. In other words, if pay rates reflect eclucational and experience require- ments for positions within a job ladder, and women (or minorities) are concentrated (lis- proportionately at the lower rungs of that lacicler, one would find a negative effect of percent female (or nonwhite) on wage rates simply by failing to control for the di~er- ences in requirements at various levels of the ladder. To examine this possibility, we cocled the detailed educational and experience require- ments from job descriptions for a random sample of state civil service jobs. We sup- plementecl this sample with a complete census of jobs within four representative agencies. ~5 We used this strategy to ensure a represen- tative sampling of civil service jobs, as well as a complete set of positions linked within detailed job ladders (from the four represen- tative agencies). Analyses of pay cletermina- tion within this subset of jobs enablecl us to determine whether the findings in Table 5- 2 were due to the concentration of women anc] minorities in jobs that are toward the bottom of detailed job hierarchies and that required less education an(l experience. California civil service job specification sheets typically list various combinations of education ant] experience that can be used to satisfy entry requirements. We charac- terized each job's requirements in terms of the particular combination involving the most experience and the least schooling.7 Because women are more disacivantagecl (relative to men) in labor force experience than in formal schooling, our conservative procedure increases the chances that train- ing requirements will absorb the gross effect of sex composition on pay rates. We coded educational ancI experience requirements for 406 positions that em- ployed 32, 719 full-time civil servants in 1985 (27 percent of the work force whose jobs were analyzed in Table 5-2~. For those positions, Table 5-3 reports regression es- timates similar to those in Table 5-2, but with these more precise controls for job requirements. Not surprisingly, educational and expe- rience requirements explain much of the variation in pay rates across jobs (84 percent 7For instance, a job specification might have listed the following combinations of education and experience as permissible: college graduate or 2 years of college and 2 years of experience in related classification or high school degree and 4 years of experience in related classification. We would have coded the last of these options and characterized the job as requiring 12 years of education (minimum) and 4 years of experience (maximum). In supplementary analyses, we also con- trolled for whether a specific curriculum, license, or other credential was required; those controls did not alter our results.
116 PAY EQUITY: EMPIRICAL INQUIRIES TABLE 5-3 Effect (in dollars) of Demographic Composition and Education/Experience Requirements on Prescribed Minimum Monthly Starting Pay in 406 California Civil Service Jobs, 1985 Elect of Sex Race Model Including Sex x Race Percent Percent Female Black Percent Percent Percent Black Hispanic Other Female A. Sex B. Race C. Sex, race D. Sex x race E. Experience and educational - 12.49 (- 14.11) ( - 39.85) - 11.84 (- 11.71) -9.45 (- 14.76) - 10.01 (- 11.35) - 10.34 (- 13.38) requirements F. Sex x race + requirements - 6.97 ( - 7. 18) - 4. 17 ( - 4. 24) G. Sex x race + requirements - 5. 72 ( - 6. 76) - 4.52 ( - 3. 72*) + controls for 20 job categories H. Sex x race + requirements -5.99 (-6.44) + controls for 20 job categories + controls for major schematic (in- dustry) group a - 9.30 ( - 20.23) - 6.49 ( - 20.06) -8.07 (- 18.27) -6.50 (- 12.59) -4.49* (- 16.46) -5.68 (- 13.15) 6.22 (0.01*) 2.73* (- 1.71*) -5.39 (-8.28) 3.59* (-0.20*) 2.34 (-4.05*)-3.12 (-5.30) 4.41* (3.32*) -4.53 (-3.67*) 2.38 (- 1.60*) -2.21* (-3.83) 4.22 (-4.24*) NOTE: Coefficients in parentheses are from regressions weighting each job by the number of incumbents. * p > .05; all other coefficients significant at .05 level (two-tailed). Coefficients of determination are adjusted for degrees of freedom. aSee appendix. This categorization includes the following level of detail: office and allied services, regulatory and public safety, and so on. according to mode} E, when jobs are weight- ect according to the number of incumbents). Demographic composition continues to have a sizable effect on job worth, however, even after educational and experience require- ments are controlled. As in Table 5-2, the elects of race composition are reduced more by controls for job content than is the case for sex composition. The pattern of results, however, is generally the same as in Table 5-2. Moreover, controls for major EEOC job class and major schematic (in(lustry) group in Table 5-3 generally do not add much explanatory power once educational anal experience requirements are already controlled, nor does their inclusion seem to absorb much of the effects of demographic composition on pay rates (compare models F and H in Table 5-3~. This gives us some confidence in our use of dummy variables for job category and schematic class (Table 5-2) as proxies for differences in job content, since their effect on pay rates appears to be due overwhelmingly to differences in educational and experience requirements across job categories and schematic classes. The fact that the demographic effects persist in Table 5-3 suggests that they are not simply capturing the segregation of women and minorities in positions within (letaile(l job la(lders that require less human capital. 8 Moreover, as in Table 5-2, the 8We conducted supplementary analyses limited to the subset of jobs in the four state agencies for which we had information on all jobs, which enabled us to look at each job's position within specific detailed ladders. These analyses yielded results comparable to those reported in Table 5-3, and there were no major differences in the extent to which the effects of de- mographic composition on pay rates were mediated by educational and experience requirements.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE TABLE 5-3 Continued ~7 Effect of Sex x Race Requirements Percent Percent Other Years Experi- Years Education Model Including Hispanic Female Female ence (Maximum) (Minimum) R2 A. Sex .28 (. 58) B. Race .09 (.34) C. Sex, race .34 (.67) D. Sex x race -28.45 (-5.40*) -3.62* (2.05*) .35 (.67) E. Experience and educational 201.63 (208.81) 193.58 (157.17) .60 (.84) requirements F. Sex x race + requirements Sex x race + requirements + controls for 20 job -9.33* (4.75*) 1.99 - 8.94* (7.40*) - 1. 17* categories H. Sex x race + requirements -7.95 (638*) -1.45* + controls for 20 job categories + controls for major schematic (in- dustry) groupa (7.71) 156.36 (153.20) 179.63 (135.24) .69 (.91) (4.43*) 119.30 (165.08) 155.94 (152.64) .76 (.92) (1.60*) 120.00 (160.38) 126.53 (131.67) .79 (.93) NOTE: Coefficients in parentheses are from regressions weighting each job by the number of incumbents. * p ~ .05; all other coefficients significant at .05 level (two-tailed). Coefficients of determination are adjusted for degrees of freedom. aSee appendix. This categorization includes the following level of detail: office and allied services, regulatory and public safety, and so on. results in Table 5-3 are even stronger when jobs are weighted by their number of in- cumbents, which suggests that our findings are not due simply to several anomalous jobs with few incumbents. Penalties Associates} with the Presence of Women and Nonwhites in a job The penalties against female- and minority-dominated jobs that are implied by our regression models are illustrated in Table 5-4, which reports the predicted minimum monthly starting pay for jobs under various hypothetical demographic compositions. The estimates in Table 5-4 are based on (unweighted) regression re- sults from three of our models. The first column uses the coefficients from mocle} D of Table 5-2, which controls only for (demographic composition; thus, these pre- dictions pertain to the typical civil service job. The seconc! column uses coefficients from model G of Table 5-2, which controls for demographic composition, the 20 EEOC job categories, and 99 cletaile(l occupa- tional schematic categories. (The predict- e(l pay rate pertains to a nonsupervisory clerical position within the occupational category "office and allie(l services gen- eral"~. The thir(l column uses coefficients from mociel H of Table 5-3, which controls for demographic composition, educational and experience requirements, E E OC job categories, and major schematic catego- ries. (The job in question is assumed to be a nonsupervisory clerical position with- in "office and allied services" that requires 13 years of education tminimum1 and 4
118 PAY EQUITY: EMPIRICAL INQUIRIES TABLE 5-4 Predicted 1985 Monthly Starting Pay Rate for California Civil Service Jobs as a Function of Demographic Composition, Based on Regression Results in Tables 5-2 ant] 5-3 Predicted Monthly Starting Pay According to (Unweighted) Results of Hypothetical Demographic Composition of Job 1. Average full-time state employee's job (38.56, 27.13, 5.18, 6.10, 6.61, 5.63, Model D, Table 5 2a $2,358 5.46, 5 33)3 2. Average full-time white male's job 2,637 (60.57, 12.68, 5.47, 2.69, 7.85, 1.99, 6.66, 2.09) 3. Average full-time white female's job (18.02, 46.31, 2.87, 8.71, 3.63, 8.81, 3.19, 8.47) 4. Average full-time black male's job (40.68, 15.03, 15.04, 6.10, 11.49, 3.94, 5.23, 2.50) 5. Average full-time black female's job (17.01, 38.72, 5.18, 12.59, 5.21, 10.18, 3.06, 8.04) 6. Average full-time Hispanic male s job 2,364 (45.82, 14.91, 9.02, 4.81, 13.44, 9.91, 5.63, 2.47) 7. Average full-time Hispanic females 2,000 job (13.62, 42.47, 3.63, 11.04, 4.59, 12.84, 2.93, 8.88) 8. Average full-time "other" male's job 2,519 (47.02, 15.83, 4.96, 3.41, 6.81, 3.02, 14.61, 4.35) 9. Average full-time "other" female's job (15.09, 43.10, 2.43, 9.20, 3.06, 9.37, 4.46, 13.30) 10. 100 percent white male . 90 percent white male, 10 percent white female 12. 90 percent white male, 10 percent black male 13. 90 percent white male, 10 percent black female 90 percent white male, 10 percent Hispanic male 15. 90 percent white male, 10 percent Hispanic female 16. 90 percent white male, 10 percent "other" male 17. 90 percent white male, 10 percent "other" female 2,089 Model G. Table 5 2b $1,505 1,633 1,378 1,547 1,374 1,532 1,348 1,576 1,353 2,984 2,890 2,880 2,875 2,935 2,778 2,936 2,846 1,787 1,741 1.750 1.730 1,769 1,719 1,763 1,719 Model H. Table 5 3C $2,050 2,230 1,860 2,136 1,863 2,107 1,817 2,155 1,821 2,387 2,327 2,342 2,324 2,411 2,271 2,365 2,290 aResults pertain to the typical civil service job. bResults pertain to a nonsupervisory clerical job within the occupational category "office and allied services: . ,, general. CResults pertain to nonsupervisory clerical job within ''office and allied services," which is assumed to require 13 years of education (minimum) and 4 years of experience (maximum). dNumbers in parentheses give percent white male, white female, black male, black female, Hispanic male, Hispanic female, "other" male, and "other" female, respectively.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE years of experience fmaximum], roughly the statewide average.19 Row 1 of Table 5-4 reports predictions that assume that the position had the sex and race composition that characterized the job held by the average full-time state em- ployee in March 1985. Rows 2 through 9 report predictions that assume that the po- sition had the sex and race composition that characterized the job helc] by the average member of specific sex and race subgroups- for instance, the average white male or female Hispanic. By comparing these pre- dictions, the reacler can determine how the pay rate in a job is predicted to vary if that job had the demographic composition that characterizes female (versus male) or non- white (versus white) civil servants. Table 5-4 documents that a job is expected to pay substantially more if it has the de- mographic composition characteristic of white men in the state civil service than if that same job had the demographic composition typical of any other group. For instance, according to model G of Table 5-2, which controls for cletailed occupational clistinc- tions, the job would pay roughly $255 less per month if it hac! the demographic com- position characteristic of white females than if it had the mix typical of white males, and slightly greater penalties are predicted if the job had the mix characteristic of non- white females (blacks, Hispanics, ant] "oth- er" females). Model H of Table 5-3, which controls for educational and experience requirements but not for occupational distinctions at such a detailed level, estimates the penalties to be even larger. Compared with the average white male's job, the position would have an entry pay rate of roughly $370 less per 9The use of different implicit reference points in computing predicted pay rates in each column of Table 5-4 affects the absolute size of the predicted values in the table, but not the size of the penalties associated with differences in demographic composition, which is the major focus of the table. 119 month if it hac! the composition typical of white females or black females, and less still if it had the demographic composition characteristic of Hispanic and "other" fe- males. Considering that these models con- trol for detailed occupational differences and/or formal educational and experience requirements, these penalties appear quite large ant] suggest that the typical female or nonwhite civil servant pays a substantial price by virtue of being in a job classification not dominated by white males. Rows 10 through 17 of Table 5-4 provide a cli~erent means of assessing the deval- uation of female- ant] minority-dominated jobs. Row 10 reports the predicted pay rate in the hypothetical job if it were hel(l ex- clusively by white males (i.e., if all seven variables capturing sex-race combinations in Tables 5-2 and 5-3 equal zero, implying that all incumbents are in the omitted cat- egory of white males). Under model G of Table 5-2, the starting pay rate would be about $282 higher per month if the job were hell] only by white men than if it had the demographic mix characteristic of the state civil service workers overall; un(ler model H of Table 5-3, the estimated premium is of roughly the same magnitude. Contrasting rows 11 through 17 with row 10 shows the estimate(1 effect on the pay rate if the demographic composition is al- terecI so that the white male share of em- ployment in the job declined by 10 per- centage points and those white males were replace(1 by females or minority males. For instance, row 11 reports the preclicted pay rate if the variable measuring the percentage of white females in the job equals 10 an the six remaining sex-race variables in Ta- bles 5-2 and 5-3 equal 0, implying that 90 percent of the job incumbents are in the omitted category of white males. If the job composition changed from 100 percent white males to 90 percent white males/10 percent white females, the pay rate is pre(lictec3 to (recline by nearly $100 per month (base(l on moclel D, Table 5-2), and by roughly
120 $50 to $60 per month even after controlling for detailed occupation and formal job re- quirements (mode} G of Table 5-2 and mode} H of Table 5-31. The penalties are estimated to be even greater if the replacements were nonwhite women. The net penalties asso- ciated with nonwhite men in a job are less severe, but still nontrivial. For instance, replacing 10 percent of the white men with black men is predicted to lower the pay rate by $37 to $45 per month after con- trolling for occupation and job require- ments. In sum, Table 5-4 dramatizes the large gross and net penalties against female- and minority-clominated jobs implied by our regression results in Tables 5-2 and 5-3. Table 5-4 shows how much more female- anc! minority-dominated jobs would pay if they were dominated by white males. Note that most of the estimated penalties in Table 5-4 would have been even larger if the weighted (rather than unweighted) regres- sion results in Tables 5-2 and 5-3 were used to calculate predicted pay rates. TREND ANALYSES We examiner] temporal changes in the elects of demographic composition on pay rates in several ways. (For the sake of brev- ity, we merely summarize these results.~°) First, we estimated cross-sectional regres- sions similar to those in Table 5-1 for two sets of jobs: enduring jobs (positions that were occupied in 1979 ant] 1985) and job births (positions that entered the civil ser- vice between 1979 and 1985~. Contrasts between these sets of jobs provide inter- esting information on the nature and mag- nitude of changes in pay inequities (for additional details, see Baron and Newman, i°Details are available from the authors on request (also see Baron and Newman, 1988~. All results re- ported in this section are based on analyses that weighted jobs in proportion to their number of in- cumbents in a given year and are significant beyond the .05 level (two-tailed tests). PAY EQUITY: EMPIRICAL INQUIRIES 1988~. Second, we undertook longitudinal analyses of changes in demographic com- position and pay rates for jobs having in- cumbents in 1979 and 1985. Enduring jobs: Cross-Sectional Regressions, 1979 and 1985 For enduring jobs, we examined whether the elects of demographic composition on pay rates differed for 1979 and 1985. In these analyses, the depenclent variable was the logarithm of the prescribed minimum monthly starting pay rate. (Using logarithms enabled us to examine how demographic composition affects pay rates in proportion- ate, rather than absolute, terms, thereby circumventing any confusion caused by the inflation of pay rates between 1979 and 1985. These analyses reveal that the wage penalties associated with the presence of blacks and male Hispanics in jobs had ac- tually increased significantly by 1985, even after controlling for the 20 major EEOC job classes and 99 occupational schematic categories. In 1979, replacing 10 percent of white males with Hispanic males implied a decline of roughly 1.1 percent in the pre- scribecl starting pay rate, net of occupational controls; by 1985 that penalty was about 3.7 percent. Similarly, a hypothetical 10 percent replacement of white males by black mates was associate(l with a 2.6 percent decline in the entry pay rate in 1979, com- parecl with a 3.9 percent decline in 1985. It appears, however, that jobs (lominatec3 by black females fared especially badly: a 10 percent increase in black females was associated with a 4.6 percent (recline in the 1979 pay rate, compared with a 9.2 percent decline in 1985. 1lIn these analyses, the effects of the job and oc- cupational controls were constrained to be equal in 1979 and 1985. Supplementary analyses based on con- stant dollars revealed the same general pattern of results. Except as noted in the text, models that allowed the effects of the job and occupational controls to vary across years also showed the same general pattern of results.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE In contrast, the wage penalties associated with the presence of white and Hispanic women in jobs decreased significantly, even after controlling for the 20 major EEOC job classes and 99 occupational schematic categories. In 1979, a 10 point increase in the percent white female lowered the start- ing pay rate by about 3.3 percent, net of occupational controls; by 1985, the corre- sponding penalties had declined to 2.7 per- cent. Improvement seems to have been particularly prevalent in jobs having a dis- proportionate share of female Hispanic in- cumbents: In 1979, a 10 point increase in the proportion of female Hispanics was as- sociated with a 12.7 percent decline in the starting pay rate; by 1985, that same change in demographic composition was associated with a net penalty of 5 percent. These results suggest that the pay equity movement, which has been defined as a "women's" issue, may have lowered the pen- alties associated with jobs dominated by white and Hispanic women between 1979 and 1985. Indeed, at the beginning of 1985, the state upgraded pay rates in selected female- dominated (especially clerical) occupations as part of pay equity agreements with several collective bargaining units. In supplemental analyses, we allowed the effects of major job category on pay rates to diner between 1979 and 1985 in orcler to capture this realignment. The results document a considerable up- gracling of clerical positions relative to other jobs; that upgrading, in turn, is responsible for the reduced penalties against enduring female-dominated positions observed in 1985. During this same period, however, the pen- alties associated with nonwhite men and black women in jobs actually increased. Perhaps in comparison to nonblack women, black women were underrepresentec] in the female-c30mi- 1 . . . ] . ] natect tines tnat were upgracled within the civil service pay hierarchy in 1985.~2 i2The special pay equity adjustments were concen- trated primarily in female-dominated clerical and tech- nical titles. Civil service positions having large numbers 121 Comparing 1979 and 1985 cross sections also suggests that less of the gross penalty associated with female- and minority-dom- inated jobs is mediated by major job cat- egory in 1985 than in 1979. In other words, the pay disparities associated with demo- graphic composition in 1985 do not reflect the mere exclusion of women and minorities from higher paying lines of work as much as in 1979. Comparing 1985 Penalties in Enduring Jobs and job Births This comparison provides some evidence that penalties against female- and minority- dominated jobs are less severe in more recently created positions, consistent with our hypotheses about the effects offounding conditions on job definitions. In weighted analyses controlling only for demographic composition, the 1985 monthly penalty for a 1 percent increase in white females in a job was $4.27 less among jobs founded between 1979 an(1 1985 than in enduring jobs, and the monthly penalties associated with the presence of black males and black females were also reduced: $19.74 and $16.88 less per percentage point, respectively. Controls for major job category and (letafled occupation me(liate these effects almost en- tirely; in other words, many of these newly created positions were in high-wage lines of work, some of which employee! female an(1 minority workers. Thus, the lower pen- alties associated with recently create(l jobs appear to reflect an upgrading of the state's occupational structure and greater access of women an(1 minorities to these higher wage positions than in the past, rather than a and percentages of black women, in contrast, include a number of low-level, service-oriented jobs (e.g., in nursing and food preparation), as well as positions in several state agencies that have been sanctioned in the past by the SPB for poor affirmative action per- formance.
122 pattern of increased equity in pay setting across-the-boarc] in newer jobs. Longitudinal Analyses For enduring jobs, we examined how a position's initial (lemographic composition and changes in its demographic composition between 1979 and 1985 affected changes in the prescribed pay rate. The causality could run in the opposite direction, however: If men have superior opportunities and wom- en are therefore willing to accept Tower wages, then a clecline in the relative stand- ing of a job in the wage hierarchy might make it more likely that women will clom- inate that position. Pfeffer and Davis-Blake (1987), for instance, concluded that when eclucational organizations increased their wages (relative to wages paid by competi- tors), they were better able to compete for desired male administrators (also see Blau's 1977 study of clerical workers). This argu- ment may make sense when applied to the allocation of individuals within a particular occupation. We think it unlikely, however, that changes in the prescribe(1 pay of dif- ferent civil service jobs would affect de- mographic composition by changing the la- bor supply available in each position for instance, that an increase in the relative wages of some occupation wouIc] induce male (but not female) clericals or school- teachers to change to that other line of work. Such major career adjustments seem particularly unlikely within the rather short interval (6 years) spanned by our data. Ac- cordingly, we do not pursue that causal argument in this paper. Table 5-5 reports the results of analyses that regress 1985 pay rates on 1979 de- mographic composition and changes in de- mographic composition between 1979 and 1985, controlling for 1979 pay rate. (These analyses weight jobs according to the num- ber of incumbents. ) Not surprisingly, there is tremendous inertia in the wage hierarchy within CaTifornia's civil service system; in- PAY EQUITY: EMPIRICAL INQUIRIES dee(l, one can explain 96 percent of the variation in 1985 base pay rates simply by knowing the 1979 pay rates. Given this stability, the deck is somewhat stacked against finding net effects of demographic compo- sition (and changes in composition) on pay rates. Nonetheless, Table 5-S reveals that the demographic composition of jobs in 1979, as well as changes in composition between 1979 and 1985, had significant effects on changes in prescribed wage rates. Mode! 1 assesses the effects of 1979 demographic composition and changes in demographic composition on 1985 pay rate, controlling for 1979 pay rate. lobs that became in- creasingly dominated by women, blacks, en cl "other" females experienced less growth in their prescribe(l entry pay rates than (licl other jobs. In ad(lition, jobs with a (lispro- portionate share of blacks (especially black women), Hispanics, or female members of other minorities in 1979 hac] their pay rates rise less quickly than did other jobs. Con- versely, model 1 indicates that jobs clomi- nated in 1979 by females, particularly female Hispanics, actually ha(1 larger net wage gains, as dill jobs that increaser] their percentage of female Hispanics by 1985. i3 Labor market conditions might explain these trencls: Women an(1 minorities may simply have entered occupations in which there was an oversupply of workers, relative to other kinds of work, an(1 consequently pay rates grew less quickly in those lines of work. To examine that possibility, mocle! 2 of Table 5-5 controls for major job category an(l cletailed occupational schematic code, as well as 1979 base pay rate. Thus, this mo(le! assesses the effects of demographic composition and change on pay rates by comparing jobs in the same specific line of Supplementary analyses that measured 1985 and 1979 pay rates in a logarithmic metric revealed the same general pattern of results, as did analyses allowing for nonlinear effects of change in demographic com- position between 1979 and 1985.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE TABLE 5-5 Effects (in dollars) of Demographic Composition (1979) and Changes in Demographic Composition (1979-1985) on 1985 Prescribed Minimum Monthly Starting Pay in 2,285 California Civil Service Jobs 123 Model 1a Model 2a Variable b p b p Percent female, 1979 .495 .010b - .041 .848 Percent black, 1979 - 1.192 .012 - 1.981 <.001 Percent Hispanic, 1979 - 6.251 <.001 - 1.599 .002 Percent other minorities, 1979 .893 .070 - .769 .091 Percent black female, 1979 - 1.810 .085 .799 .366 Percent Hispanic female, 1979 - 11.267 <.001 4.423 <.001 Percent other female, 1979 - 1.902 .063 .532 .511 Change in percent female - 2.446 <.001 - 2.661 <.001 Change in percent black - 4.071 <.001 - 1.467 .006 Change in percent Hispanic - .319 .592 .038 .931 Change in percent other .882 .098 - .041 .918 . . . minorities Change in percent black female 1.394 .230 .731 .371 Change in percent Hispanic 2.859 .006 .088 .910 female Change in percent other female - 1.874 .051 - 1.547 .021 aAnalyses weight each job according to its number of incumbents in 1979. Model 1 controls for 1979 prescribed pay rate; model 2 also controls for 20 major job categories and 99 detailed occupational schematic codes (see appendix). bTwo-tailed significance level of regression coefficient. work. Consequently, any effects of demo- graphic composition net of these cletafled occupational controls seem unlikely to re- flect differences in labor supply and demand throughout the occupational structure of the state civil service system. Mode} 2 indicates that the negative effects on pay rates associated with the movement of women, blacks, and other females into jobs persist, even after controlling for the kinds of jobs involved. (These weighted effects are stronger than in unweighted anal- yses, which suggests that the entry of wom- en and minorities occasioned the largest penalties in heavily populated job cIassifi- cations.) Moreover, the few positive effects in model 1, such as the effect of percent female in 1979 on 1985 wage rates, were due almost entirely to the kinds of positions involved and they vanished once we con- trolled (moclel 2) for detailed occupational type. In other words, jobs that were dom- inated by women or minorities in 1979 and that had unusually rapi(1 wage growth by 1985 were concentrated in parts of the civil service structure for which wage gains gen- erally were bigger over this period. As noted above, for instance, pay equity adjustments in female-dominated (especially clerical) po- sitions occurre(1 at the beginning of 1985 and may explain the gross positive effect of percent female in 1979 on 1985 pay rates. Given that such adjustments occurred, it is all the more striking to fine] that increases in female representation between 1979 and 1985 had strong negative effects on the relative pay of civil service jobs. A (different interpretation of these de- mographic effects is that they simply reflect some (recline between 1979 en cl 1985 in the requirements and duties of jobs in which the percentage of women or minorities in- crease(l. Accorclingly, we reestimated the mo(lels in Table 5-5, including a dummy variable indicating whether each position ha(1 its job specification revise(l (luring the
124 1979-1985 period. As we expected, there was a slight negative effect of job revision on growth in pay rates that is, jobs that were revised during this period apparently did have their educational or experience requirements lowered, which resulted in diminished wage growth. Including this control variable, however, cTid not change the estimates reported in Table 5-5, which suggests that the results are not spurious. Examining the Crowding Hypothesis: The Interaction Between Employment Growth and Changes in Demographic Composition We predicted less devaluation in jobs undergoing employment growth than in jobs that were declining because growth should diminish the perceived threat of competi- tion and crowding. To test this hypothesis, we estimated one final version of the lon- gitudinal models reported in Table 5-5, this time adding interaction terms between em- ployment change and changes in demo- graphic composition (percent female and percent nonwhite) in each job between 1979 and 1985. The results generally supported our hy- pothesis. When employment change be- tween 1979 and 1985 was measured in per- centage terms, there was a significant positive interaction between growth and change in percent female. As hypothesized, the de- valuation associated with women entering a position is most severe when employment is declining. This effect is present even in models that control for demographic com- position in 1979, changes in demographic composition by 1985, 1979 wage rate, the 20 major job categories, the 99 detailed schematic codes, and whether each job was revised between 1979 and 1985 (b = 1.004; p < .001, two-tailed). Measuring employ- ment change in absolute rather than per- centage terms also produces a significant positive interaction effect with change in percent female, but only in the full mode} PAY EQUII'Y: EMPIRICAL INQUIRIES (i.e., mode} 2, Table 5-5~; the effect is in the opposite direction in models that do not control for detailed] occupation. In the case of race composition, the in- teraction between proportionate employ- ment change and changes in percent non- white is also substantial, significant, and in the predicted direction, even in a model that controls for 1979 demographic com- position and changes in demographic com- position, 1979 wage rates, whether a job was revised between 1979 and 1985, and the 20 major job categories (b = .365; p = .034, two-tailed). The effect becomes insignificant, however, once controls for the 99 detailed occupational codes are included (b = .168; p = .23~. Moclels measuring employment change in absolute terms, how- ever, consistently reveal significant positive interactions with changes in percent non- white, as hypothesized.~4 In sum, our longitudinal results are con- sistent with numerous studies suggesting that the entry of women and minorities into positions clevalues them. This effect appears to be moderated, however, in recently cre- atec] jobs and in lines of work that are growing. By estimating the moclels on a stable set of positions within a single a(l- ministrative system, and by controlling in detail for the kinds of jobs involved (an(l the prevalence of job revisions), these anal- yses strongly point to a real organizational devaluation in the perceived worth of po- sitions as women an(l minorities enter them. By focusing on changes in pay rates rather than average wages actually receive(1 in each job, and by including 1979 pay rate and demographic composition as control vari- ables, our longitudinal analyses preempt a criticism sometimes levele(1 at this litera- ture namely, that the apparent wage "pen- i4These results weight jobs based on the number of incumbents and are therefore unlikely to be affected by extreme values for percentage growth associated with very small job categories (e. g., a job that increased from one to five incumbents between 1979 and 1985~.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE allies" against female- and mate-dominated jobs simply reflect lower requirements and labor quality in those jobs. SUMMARY AND IMPLICATIONS Our results document substantial wage penalties associated with the presence or entry of women and nonwhites in job cias- sifications within the California civil service. We have construed the notion of compar- ability across jobs very narrowly. That is, for the most part, our analyses have com- pared wage rates among civil service po- sitions that involve essentially similar duties and requirements and that diner only in their demographic composition, rather than comparing disparate jobs in different parts of the state's occupational structure. We have reported results from models that con- trol for educational and experience require- ments of jobs and for occupational distinc- tions, at varying levels of detail, in assessing the elects of demographic composition on wage rates. Readers can make their own judgments about the extent of devaluation, (lepending on how broadly or narrowly they think "comparability" among work roles should be construed. Perhaps our most striking result, however, is that the net penalties against female- and minority-clom- inated jobs appear severe regardless of how strict a standard of comparability one em- ploys. Our analyses have also taken the occu- pational classification system and job spec- ifications of the California civil service sys- tem at face value. Yet, there may be gender or race biases in how positions are de- scribed, classified, and have their require- ments assessed. A recent court decision concerning the comparable worth lawsuit pending against California relied heavily on a 1934 report drafted by a state government official that documented historically how gender distinctions were incorporated into job specifications and wage rates within the state civil service system (Becker, 1934~. 125 The pervasive effects of sex and race com- position that we have documented appear to represent the continuing legacy of those historical practices. Our analyses, on the other hand, also indicate that the tendency for female- and minority-dominated jobs to be devalued is not inevitable. We found, for instance, that underpayment for female- and minority- dominated positions appeared to be greatest in large "generic" jobs, compared with po- sitions with few incumbents- that is, the weighted results were invariably stronger than the unweighted ones (also see Baron and Newman, 1988~. Large job cIassifica- tions (e.g., secretarial positions) are most closely tied to the external labor market and may therefore reflect widely held ste- reotypes about their appropriateness for women or minorities. Moreover, they are also likely to be key job classifications within the civil service that serve as anchors for setting wage rates in other classifications (Dunlop, 1957~. Thus, politically and eco- nomically, it is most costly for reformers and personnel specialists to attempt to re- calibrate the worth of those positions. In other research (Baron and Newman, 1988), we have shown that the degree of devaluation also depends on the age, union coverage and task ambiguity of the job. Thus, devaluation is not universal, which underscores the importance of examining the organizational and institutional factors governing pay setting and related personnel activities in studying pay equity. Those fac- tors have too often been overlooked in in- dividual- and occupation-level research on sex and race inequality. Research that ex- amines the characteristics of jobs, organi- zations, and institutional environments that favor or discourage devaluation would be of inestimable value to scholars, policy- makers, and personnel practitioners. Our analyses of trends provided some evidence of improvement and some evi- dence to the contrary. On the one hand, we found that the entry of women or mi-
126 norities into state civil service jobs had significant negative effects on the assessed worth of those jobs, especially when em- ployment was not growing rapidly enough to embrace those entrants without threat- ening white males who previously monop- olized those jobs. To cast some light on the processes underlying our longitudinal re- sults, we took a detailed look at particular jobs in which large demographic changes between 1979 and 1985 were accompanied by large changes in prescribed pay. A num- ber of these were occupations in data pro- cessing, where the rapid entry of women into specific lines of work appears to have devalued those positions relative to other jobs within computing. Using census data, Strober and Arnold (1987b) have c30cu- mented a similar channeling of women into the least lucrative positions within the com- puting sector. Our analyses suggest those positions may have become relatively less lucrative precisely because women were entering them, perhaps causing the work to be perceived as more clerical than tech- nical in nature and thus warranting lower relative pay. Demographic changes were also accom- panied by changes in pay for several "uni- formed law enforcement and correctional titles, located in agencies that had been sanctioned in the past for failing to hire women and nonwhites. It appears that as gains in female and minority representation occurred within uniformed classes, those classes were left behind in the pay hier- archy, relative to other positions within the same occupational families. Finally, we found a few instances in which increaser] male representation in certain titles previously dominated by women ap- parently led to a revaluation upward of the pay of those positions, relative to lower level, female-clominated titles within the same occupational family. These results sug- gest that as an occupation becomes more integrated by sex or race, there is increasing stratification and devaluation of work per- PAY EQUITY: EMPIRICAL INQUIRIES formed by women or nonwhites within the occupation, a conclusion buttressed by other recent studies relying on diverse data sources (Bielby ant] Baron, 1986; Reskin and Boos, 1987; Strober and Arnold, 1987a; Tienda anal Ortiz, 19871. The felicitous effects of growth in tem- pering the (levaluation of jobs entered by women and nonwhites is consistent with other research, which indicates that women and minorities benefit disproportionately from economic expansion (see Baron, 19841. If declining employment in a job exacerbates tendencies toward devaluation, however, it is interesting to speculate about how current trends toward reducing permanent em- ployment and privatizing public sector func- tions will affect progress toward pay equity in the civil service. By exacerbating crowc3- ing, the trend toward "contingent" em- ployment in the public sector could con- ceivably have the effect if not the intent- of creating further obstacles to pay equity in state government. Although the movement of women anal minorities into jobs between 1979 and 1985 appears to have devalued those positions, we (li(l fin(l that recently created occupa- tions in state government impose(1 less cle- valuation than did older civil service posi- tions, which suggests some progress toward equity. It may simply be that these newer jobs have not existed long enough for ste- reotypes concerning race an(1 sex appro- priateness to (levelop. Strober and Arnold (1987a) have suggested, for instance, that such stereotypes evolved in computing oc- cupations after an initial period of ambiguity about the appropriate personnel in various computer-relate(l jobs. Taken together, our fin(lings concerning trends seem to indicate that in a large bureaucratic labor market, such as the California civil service, progress toward pay equity is harder to achieve among enduring, inert jobs, especially under cir- cumstances of declining employment, than among newly created positions in the sys- tem.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE Our longitudinal results also cast doubt on the claim that devaluation in state gov- ernment merely reflects external market forces. Even controlling for 1979 pay levels and detailed occupational distinctions, which presumably capture market-basec3 differ- ences in compensation, we found that in- creasing representation of women and non- whites in a job lowered its position within the state government pay hierarchy. More- over, many civil service positions have few, if any, counterparts in the private sector, and so the very notion of a market wage is itself oRen problematic in this context. The state government in some instances has even acknowledged biases in assessing prevailing wage rates for femaTe-dominatecT jobs, yet failed to mitigate those biases (see Baron and Newman, 1988; also cf., Bridges and Nelson, 1988~. Thus, our results point to organizational practices and procedures- and not simply to inexorable or objective market forces as sources of devaluation in jobs held disproportionately by women and nonwhites. In one sense, the results we have reported are not noteworthy because they are con- sistent with prior studies, even research using other kinds of data (particularly U.S. census occupational trends). Such consis- tency is not always the norm in social science research. What is striking is how closely our results parallel those obtained in the few other longituclinal studies of organiza- tional wage determination, such as Prefer and Davis-Blake's (1987) study of wage de- termination among college and university administrators between 1978 anal 1983 and Rosenbaum's (1985) analysis of gentler com- position and wages in a large corporation between 1962 and 1975. Taken together, these studies document very similar pro- cesses by which work done by women and nonwhites is devalued, but across quite diverse organizational settings: state gov- ernment agencies, eclucational organiza- tions, and a large private firm. Clearly, more research is required along these lines, in- 127 clucling historical and ethnographic studies of how job classification and compensation systems are designed in organizations, the circumstances under which they change ver- sus remain inert, and their effects on in- clividuals' careers. ACKNOWLEDGMENTS This research was supported in part by a grant from the Pane! on Pay Equity Re- search, National Research Council (NRC), and by research funds from the Stanford Graduate School of Business. This paper was completed while the first author was a Fellow at the Center for Advancer! Study in the Behavioral Sciences; support from the Alfred P. Sloan Foundation and from the Center staffis gratefully acknowleciged. Officials at the California State Personnel Board, Department of Personnel Admin- istration, Controller's Office, and the Cal- ifornia State Employees Association were extremely generous in providing access to data and answering innumerable questions, but they bear no responsibility for the con- clusions of this paper. Ann Bucher and [ill Fukuhara ably assisted in preparing the manuscript. The authors appreciate helpful assistance and suggestions from members of the NRC pane! and staff, Bill Bielby, Albie Lau, Don Palmer, Marilyn Pearman, Jeff Pfeffer, Peter Reiss, lean Ross, and especially, Brian Mittman. REFERENCES Baron, James N. 1984 Organizational perspectives on stratification. Annual Review of Sociology 10:37-69. Baron, James N., and Andrew E. Newman 1988 For What It's Worth: Organizational and Oc- cupational Factors Affecting the Value of Work Done by Women and Nonwhites. Unpub- lished manuscript, Graduate School of Busi- ness, Stanford University, Stanford, Calif. Becker, F. M. 1934 Administration of the Personnel Program in the State of California. Sacramento: State of
128 California Archives (Administration, Person- 1985 net Board Historical). Bergmann, Barbara 1974 Occupational segregation, wages and profits when employers discriminate by race or sex. Eastern Economic Journal 1:103-110. Bielby, William T., and James N. Baron 1986 Men and women at work: Sex segregation and statistical discrimination. AmericanJour- nal of Sociology 91:759-799. Blau, Francine D. 1977 Equal Pay in the Office. Lexington, Mass.: D. C. Heath. Bridges, William P., and Robert L. Nelson 1988 Markets in Hierarchies: Organizational and Market Influences on Gender Inequality in a State Pay System. Paper presented at a meeting of the American Sociological Asso- ciation, Atlanta, Ga. California State Department of Finance 1986 California State Abstract 1986. Sacramento: State of California. California State Department of Personnel Administra- tion 1982 Report to the California Legislature and Ex- clusive Representatives of State Employees on Information Relevant to the Salaries for Female-Dominated Jobs. Sacramento: State of California. Deaux, Kay 1985 Sex and gender. Annual Review of Psychology 36:49-81. Dunlop, John T. 1957 The task of contemporary wage theory. Pp. 3-27 in John T. Dunlop, ea., The Theory of Wage Determination. London: Macmillan. Hannan, Michael T., and John H. Freeman 1984 Structural inertia and organizational change. American Sociological Review 49:149-164. Hartmann, Heidi I., Patricia A. Boos, and Donald J. ~ . 1 relman 1985 An agenda for basic research on comparable worth. Pp. 3-33 in Heidi I. Hartmann, ea., 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. Hodge, Robert W., and Patricia Hodge 1965 Occupational assimilation as a competitive process. AmericanJournal of Sociology 71:249- 264. Kanter, Rosabeth M. 1977 Men and Women of the Corporation. New York: Basic Books. Killingsworth, Mark R. PAY EQ UITY: EMPIRICAL INQ UIRIES The economics of comparable worth: Ana- lytical, empirical, and policy questions. Pp. 86-115 in-Heidi I. Hartmann, ea., Com- parable Worth: New Directionsfor Research. National Research Council, Committee on Women's Employment and Related Social Issues. Washington, D.C.: National Academy Press. McArthur, Leslie 1985 Social judgment biases in comparable worth analysis. Pp. 53-70 in Heidi I. Hartmann, ea., Comparable Worth: New Directions for Research. National Research Council, Com- mittee on Women's Employment and Related Social Issues. Washington, D. C.: National Academy Press. Major, Brenda, Dean B. McFarlin, and Diana Gagnon 1984 Overworked and underpaid: On the nature of gender differences in personal entitlement. Journal of Personality and Social Psychology 47:1399-1412. Parcel, Toby L., Steven Cuvelier, Jennifer Zorn, and Charles M. 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. Pfeffer, Jeffrey, and Alison Davis-Blake 1987 The effect of the proportion of women on salaries: The case of college administrators. Administrative Science Quarterly 32:1-24. Reskin, Barbara F., and Patricia A. Boos 1987 Status hierarchies and sex segregation. Pp. 1-21 in Christine Bose and Glenna Spitze, eds., Ingredients for Women's Employment Policy. Albany: State University of New York Press. Rosenbaum, James E. 1985 Jobs, job status, and women's gains from affirmative action: Implications for compa- rable worth. Pp. 116-136 in Heidi I. Hart- mann, ea., Comparable Worth: New Direc- tionsfor Research. National Research Council, Committee on Women's Employment and Related Social Issues. Washington, D.C.: Na- tional Academy Press. Stinchcombe, Arthur L. 1965 Social structure and organizations. Pp. 142- 193 in James G. March, ea., Handbook of Organizations. Chicago: Rand McNally. Strober, Myra H., and Carolyn L. Arnold 1987a The dynamics of occupational segregation among bank tellers. Pp. 107-148 in Clair Brown and Joseph A. Pechman, eds., Gen- der in the Workplace. Washington, D.C.: The Brookings Institution.
WAGE RATES IN THE CALIFORNIA CIVIL SERVICE 1987b Integrated circuits/segregated labor: Wom- en in computer-related occupations and high- tech industries. Pp. 136-182 in Heidi I. Hartmann, ea., Computer Chips and Paper Clips: Technology and Women's Employ- ment. Vol. II. National Research Council, Committee on Women's Employment and Related Social Issues. Washington, D.C.: National Academy Press. Tienda, Marta, and Vilma Ortiz 1987 Intraindustry occupational recomposition and APPENDIX I. Job Category Codes 01 Clerical 02 Supervisory clerical 03 Semiskilled workers 04 Crafts/tracles 05 Supervisory crafts/trades 06 Professional 07 Supervisory professional 08 Subprofessional/technical 09 Supervisory subprofessional/technical 10 Law enforcement II. Sample Schematic Arrangement of Classes (Al Office and Allied Services A. General B. Typing C. Stenography and Secretarial D. Legislative E. Payroll F. Personnel-clerical G. Machine operations 1. Key data 2. Mailing 3. Microfilm (BJ Regulatory and Public Safety A. Police and law enforcement 1. Highway patrol ]29 gender inequality in earnings. Pp. 23-51 in Christine Bose and Glenna Spitze, eds., In- gredients for Women's Employment Policy. Albany: State University of New York Press. Treiman, Donald J., and Heidi I. Hartmann 1981 Women, Work, and Wages: Equal Pay for Jobs of Equal Value. National Research Coun- cil, Committee on Occupational Classification and Analysis. Washington, D. C.: National Academy Press. 11 Supervisory law enforcement 12 Field representative 13 Supervisory field representative 14 Administrative staff 15 Administrative staff supervisory 16 Administrative line 17 Janitor/custodial 18 Supervisory janitor/custodial 19 Laborers 20 "Career opportunity development" classes I 4. Duplicating 5. General office H. Storekeeping 1. General 2. Equipment Communications Fiscal 1. Cashiering 2. Account recor(lkeeping and review K. Miscellaneous office services and allied 2. Fish and game 3. State police
130 B. Criminal identification and investigation 1. Administration 2. Fingerprints 3. Criminalists 4. Polygraph 5. Law enforcement consult. Special investigator PAY EQUITY: EMPIRICAL INQUIRIES C. Field representation D. 1. Collection/tax acimin. 2. Real estate Inspection E. 1. Regulation of business/professional activities 2. Public health and safety
Commentary JEAN ROSS The study by Baron and Newman analyzes the relationship between demographic fac- tors and wage setting practices within the State of California's civil service system. The system is that of a relatively centralized, large public sector employer. The wage (lata user] consist of the starting pay assignee] to each job classification at two points in time, March 31, 1979, and March 31, 1985. The categories for analysis inclucle sex and race, broken down into whites, blacks, Hispanics, and other minorities (primarily Asians). Within the literature on wage and job segregation, there is a paucity of case study research. As a result, there is considerable clebate over the extent to which findings of research using national survey data reflect firm, geographic, and other factors rather than within-firm job segregation or wage discrimination. To the extent that within- employer research exists, it is often pre- pared by advocates, such as labor unions, ant! frequently lacks depth and a theoretical framework. This paper is an important con- tribution to the field because it provides an in :lependent and analytic assessment of pay practices within an organization. It is also important because of the nature 131 of the employer studied. The California civil service system is large and diverse, em- ploying approximately 150,000 workers in nearly 5,000 job titles. The state draws from a variety of labor markets, urban and non- urban, for its work force. Baron an(l Newman's findings gain addecl significance as a result of the history of pay equity as an issue for employees in the system. The state is currently involved in litigation with the largest of its employee unions, the California State Employees As- sociation (CSEA)/Service Employees Inter- national Union (SEIU) Local 1000. The law- suit grew out of activism on this issue which began in the late 1970s. The lawsuit was filed in 1984, and the union negotiated a collective bargaining agreement in 1985 containing wage adjustments for a number of predominantly female jobs. Thus, the wage disparities found in this stucly persist despite attempts to remedy disparities. As a result of the activism around pay equity, one would expect that disparities, if any- thing, would be less than would be found in the labor market at large. Baron and Newman fin(l large ant! sig- nificant undervaluation on the basis of race
Z32 and sex even when controlling for occu- pational grouping and education and ex- perience. Effects are greater when jobs are weighted by number of incumbents. As Baron and Newman note, wage disparities are more costly to eradicate in large jobs. Large jobs are also more likely to be am- biguously described and to be sex stereo- typed because they have a large number of female or minority incumbents, whereas smaller jobs (and their incumbents) are more likely to be seen as worthy of individualizes] treatment. The level of job segregation in the Cal- ifornia system is severe the average white male's job is 12.7 percent white female and the average white female's job is 18 percent white male. The pay gap between the av- erage female and average mate's job is $548 per month, without occupational or other controls. Gaps for blacks ($216 for males) and Hispanics ($273 for males) are smaller, but still statistically significant. Occupa- tional and educational controls reduce, but do not eliminate, wage clevaluation. Edu- cation ant] experience explain 84 percent of the variation in wage levels, but sex and black ant] other minority are still statistically significant variables in explaining the resid- ual. Baron and Newman control for job con- tent using the state's finely clifferentiate schematic system, which categorizes posi- tions according to 1 of 281 classifications. They acknowledge that analyzing wage dis- parities controlling for such a finely clisag- gregate(l occupational breakdown may in- corporate any built-in biases in the cias- sification system (as opposed to the wage setting process). This is supported by the findings that sex and race have a much greater impact when only e~lucational an experience controls are used. The results of this analysis strengthen the overall fin(l- ings by showing that race and sex have significant and negative effects even within narrowly defined occupational groups. The depiction of within-occupation disparities is PAY EQUITY: EMPIRICAL INQUIRIES one of the strongest and most unique con- tributions of the paper. Analysis ofthe impact of individual human capital as a factor in wage setting was not the focus of this study. In the civil service systems with which I am familiar, however, blacks are frequently segregated into po- sitions with low eclucational requirements, regardless of the education level of the incumbents. Women, on the other hand, frequently are overeducated relative to job requirements, an(l given the supply of wom- en with higher educational attainment, em- ployers can finely differentiate job require- ments without increasing pay for jobs. They can then hire women whose education ex- ceecis states] requirements. It is also not surprising that between 1979 ant! 1985 the education and experience re- quired for jobs was reducecI. My guess is that if you looked at human capital indicators for incumbents, they would increase despite the decline in formal requirements. At SEIU, we have seen requirements downgraded when pay equity is raised as an issue in a number of workplaces. Because of labor market factors, the employers can continue to hire at a higher level, but subvert the argument that the position is truly more skilled. In Los Angeles County, for example, the educational requirement for the job of social worker was (lowngradecl to require only a bachelor's degree, even though the county hires persons with master's (legrees in social work almost exclusively. The study differentiates between new and enduring (or stable) jobs (luring the period analyze(l a useful breakdown. Quite in- terestingly for enduring jobs, (levaluation of black female and minority mate jobs, even controlling for job content, increased between 1979 an(1 1985, whereas, perhaps in response to pay equity pressure, the penalties associated with white and Hispanic female jobs clecTined. For new jobs, the results support the researchers' hypothesis that the penalties are less severe. The fin(l- ing that new jobs incur less severe penalties
COMMENTARY could reflect the fact that new jobs have not had time to become stereotyped and thus devalued. Occupational controls me- cliate these results, which reflects the fact that job births represent creation of higher level jobs filled by women or minorities, rather than upgrading of entire hierarchies. Also, as would be expected, 1979 salary explains most of the variation in 1985 salary (96 percent). This does not, however, re- duce the significance of demographic fac- tors. lobs that were increasingly held by women or minorities experienced less salary growth. This remained true even when con- trolling for occupation. An earlier version of Baron and Newman's paper tested some interesting hypotheses that jobs that are ambiguously defined (often "large" jobs with many incumbents), profes- sional and administrative jobs, and nonu- nionized jobs are all more likely to be cle- valued more on the basis of race and sex than other jobs. For example, they fincI that unionization counteracts the devaluation for sex, but not for race. In this context, it is interesting to note that within CSEA/SEIU, pay equity has been defined as an issue of sex (as opposed to sex and race) and the lawsuit filed against the state was filed on the basis of sex discrimination. Baron and Newman note that their results show that no matter how broadly or narrowly the question is framed, sex and race are associated with wage devaluation. The ex- istence of disparities, even using detailed occupational and educational/experiential controls, points to the importance of having correctly classified jobs and of assigning them correctly to a job family and pay rate at the point of creation or revision. Often, female job characteristics are not adequately valued or are explicitly devalued. To remedy wage disparities, we need to develop a better structure for assessing the skill and content of ambiguously defined jobs. Case ]33 study research on the work process can help contribute to this task. Some of the needed research exists for clerical work, but very little exists for other types of work predom- inantly held by women, such as social ser- vices, education, and health services. Baron an(l Newman's analysis also illus- trates how difficult it will be to implement an equitable values system by which wages can be determined. Penalties against wom- en's (ancl minorities') work are deeply im- beddec] and widespread. Efforts at affir- mative action are not really eliminating much of the problem. Occupational segregation remains severe and the continued wage gap within occupations remains large. This is supported by the recent Bureau of the Cen- sus (1987) report showing wide disparities within a single occupation for women and men, even when women had no gap in employment history. From the standpoint of an advocate, such as myself, case study research such as this is valuable for several reasons. First, it points to specific instances of sex- and race- basec3 wage discrimination and estimates the magnitude of wage disparities. Second, it provides a model research strategy that could be replicated for other civil service systems that are similarly organized, and many are. Third, its strong base in the scholarly literature helps to point to hy- potheses concerning phenomena (e. g., change over time, job size, job age, an(l unionization) that can provide useful in- sights in a variety of workplace settings. REFERENCE Bureau of the Census 1987 Male-female differences in work experience, occupations, and earnings: 1984. Current Population Reports, Household Economic Studies. P-70410~. Washington, D. C.: U. S. Government Printing Office.
