Click for next page ( 180


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 179
8 Comparable Worth and the Structure of Earnings: The {owe Case PETER F. ORAZEM and J. PETER MATTILA Comparable worth pay plans have been implemented in several states since the early 1980s. To our knowledge, however, no study exists of the actual impact of such plans on the pay structure of state govern- ment. We examine the case of comparable worth in Iowa, both as proposed in 1984 and as actually implemented (in compromise form) in 1985. In particular, we identify the relative winners and losers from comparable worth by analyzing the impact on earnings for men, women, minorities, unionized em- ployees, an(1 particular occupational groups, such as supervisors and professionals. In addition, we are able to determine whether the relationships between the state pay structure and such variables as market wages, educational attainment, and work experi- ence are alterec] by the plans. After discussing the development of com- parable worth in Iowa in the first section, we cletafl our hypotheses and the issues that motivated this paper. We then outline our methodology and data and present our re- sults. In the final section we summarize our findings an(1 (liscuss their applicability to other settings. 179 HISTORICAL BACKGROUND In 1983, the Iowa state legislature passed a bill stipulating that the state shall not discriminate in compensation between pre- dominantly male and female jobs deemed! to be of comparable worth. Toward that end, the state engaged Arthur Young and Company to evaluate the inherent "value" to the state of the more than 800 job cIas- sifications in the Iowa State Merit Em- ployment System. Arthur Young was di- rected to ignore market wages in conducting its analysis because the market was pre- sumed to reflect discriminatory practices of employers in the private sector. The Arthur Young consultants elected to use a point system to establish the relative value of the jobs in the state system. Each job was assigned a point total based on information on the job's level of 13 job characteristics weighted by the importance of the characteristic to the state. Three factors measured skill, five measure(l effort, three measurer] responsibility, and the re- maining two measured working conditions. (The specific factors are listed at the bottom

OCR for page 179
180 of Table 8-54. The unit of measurement for each characteristic was a subjective scale. The evaluation was performed by four- person teams of employees, who had at- tended a 3-day training program in job evaluation. Each team had two men and two women, and an effort was made to balance the groups by age and location. The teams' evaluations were based on question- naires that had been filled out by a sample of incumbents for each classification in the state merit system. Because earlier job eval- uations may have undervalued the char- acteristics of female-dominated jobs, the questionnaire was specifically designed to identify aspects of jobs traditionally held by women. The consultants assisted the eval- uation teams and tried to ensure that each team used the evaluation system in a con- sistent manner. The result was a measure for each of the 13 job characteristics for each job. Arthur Young examined several methods for establishing weights for the different job factors. The procedure that ultimately proved to be the basis for the Arthur Young rec- ommenciations was to have a committee (composed of representatives from the Iowa legislature and the governor's staff) establish factor weights independent of the job eval- uation. In fact, however, the committee modified the factor weights twice after ex- amining their impact on the final results. According to Arthur Young's (1984:30) final report, these alterations were based, at least in part, on "the different impacts on male and female jobs" and "the ways the factors actually acted in determining the final point totals." Such changes may raise (loubts con- cerning the objectivity of the pay analysis. The factor points were multiplied by the factor weights and summed to obtain total points for each job. These total points were then translated into pay grades and pay for each job. The final recommendation (April 1984) proposed that 10,751 employees should be given pay increases and 7,300 should have their pay decreased. Of those covered PAY EQUITY: EMPIRICAL INQUIRIES by a union contract, 8,800 were to receive pay increases an(l 5,000 were to receive pay cuts. We estimate that the cost of imple- menting this plan would have been $16.6 million, with the pay reductions. Overall, 79 percent of female-dominated jobs were scheduled for increases, and 40 percent were to receive increases of more than two pay grades. In contrast, 53 percent of the male-dominatec] and 48 percent ofthe mixed classifications were to receive increases, and 17 percent and 19 percent, respectively, were recommencled for increases of more than two pay Oracles. The committee also recommended, however, that employees' pay not actually be cut, but be "red circled" (i.e., they would receive no raises until their pay came into line with the recom- mendations). In this case, we estimate that the initial cost would have been $24 million per year. The final recommendations were made shortly before contract negotiations began in July 1984 between the state and the American Federation of State, County and Municipal Employees (AFSCME). Due to the high cost of this proposal and union opposition to pay cuts, a political compro- mise was negotiated between the governor's office and AFSCME in September 1984. The compromise provided that no one would suffer a reduction in pay grade and that the size of the pay increases would be reduced by one pay grade and one step. The com- promise was implemented in March 1985 at what we estimate to be a total cost of approximately $18.8 million per year. The compromise plan was implemented in March 1985, although pay recommen- dations for higher gracle jobs (especially upper-level supervisors) were not imple- mentecl until 1987. This create(1 the pos- sibility that some supervisors would be paid less than some of the employees they su- pervised. In adclition, a relatively small number of employees who were not part of the merit pay system were not stuclied by Arthur Young an(1 (lid' not incur pay a(l-

OCR for page 179
THE IOWA CASE justments in March 1985. Finally, the state an(1 AFSCME agreed to an appeals process, which ultimately resulted in further mod- ifications of pay grades in 1987. HYPOTHESES Although some view comparable worth as a straightforward application of common pay analysis procedures, the outcome in Iowa depended heavily on political expe- diency and union bargaining. Unlike private sector applications of pay analysis proce- dures, which routinely conduct market wage surveys in order to set wages in key jobs, the Iowa comparable worth plan ignored market wages totally. Ironically, by the time the dust had finally settled on the Iowa merit pay stu(ly, it was not at all clear what the ultimate impact of the effort had been. In particular, one presumes that women gained relative to men, but the extent of the gain is not clear. An important question concerns what would have happened to male and female earnings had the original com- parable worth proposal been enacted and whether the implemented compromise plan benefited women to the same extent as the original plan. A second] question is how comparable worth alters the union/nonunion pay dif- ferential. Some unions, particularly AFSCME, have been strong proponents of comparable worth, even though the concept would seem risky on the surface. As the Iowa experience indicates, a large propor- tion of workers covered by union contracts could lose as a result of comparable worth. Do unions support comparable worth be- cause the studies tend to increase returns to union status or do unions gain primarily by providing their members with protection against pay cuts, such as might have come to fruition under the original Arthur Young plan? A third question concerns how compa- rable worth affects the pay of blacks, His- panics, and other minorities. By increasing ]81 relative earnings for female-dominated jobs (e.g., clerical occupations), comparable worth plans may tend to decrease returns to char- acteristics of blue-collar jobs and other oc- cupations with relatively high numbers of minority incumbents. A fourth uncertainty in the comparable worth debate is how a comparable worth pay plan relates to market wages. Since the Iowa comparable worth study ignored mar- ket wages by construction, it should lower the impact of market forces on public sector wages. On the other hand, the compromise following the completion of the study could reverse those effects if the compromise was based in part on market opportunities. A fifth important question concerns the impact of comparable worth on the returns to education ancI experience. There are sev- eral reasons why the returns to a year of education, tenure, or general labor market experience might fall. The first is that the comparable worth pay structure rewards the minimum levels of education or experience required for a job, not the actual level of education or experience of the incumbent. Although this is true in general of rigid pay structures based on job attributes rather than indivi(lual attributes, it is possible that the previous pay structure may have re- flected the characteristics of individuals in the job as well as the characteristics of the job itself. By rewarding only the minimum level of experience or e(lucation requirec] to perform a given job, the comparable worth pay struc- ture resembles the pattern of returns to eclucation assume(l in the job signaling lit- erature (Spence, 1973~. An implication of the signaling models is that specific mini- mum threshold levels of education (e.g., high school diploma, bachelors, masters, or Ph. D. degree) will generate higher wages, but 1 year of education above the threshold level will have no value. This type of pay structure would therefore lower the returns to years of education or experience in gen- eral, but would tend to increase returns to

OCR for page 179
182 the attainment of the threshold levels. The returns to threshold levels of education would also increase if evaluators tended to put more emphasis on such credentials as a diploma, degree, license, or certificate than on the number of years of education or experience. The attainment of a degree, for instance, may be viewed as a more defen- sible requirement than requiring 11 or 15 years of education. In addition, comparable worth may lower returns to education and experience simply because the process tends to put more weight on other factors. In Iowa, none of the 13 job factors was assigned a weight of less than ~ percent regardless of how unimpor- tant it might be in explaining the existing pay structure. A goal of the Iowa study was to incorporate characteristics of female- dominated occupations that may not have been rewarded in the past. By weighting some factors more heavily, some other fac- tors must, by definition, diminish in relative importance. A sixth question concerns how compa- rable worth affects specific occupations. Be- cause of the above impact on education and experience, people often assume that lower skill, femaTe-dominated jobs gain, but no consensus exists on other occupations. We will examine how a wide variety of broadly defined occupations were affected in Iowa. We will also examine how professional and supervisory jobs were affected. METHODOLOGY AND DATA The goal of this study is to explore the impact of proposed and implemented ver- sions of comparable worth on the Iowa pay structure. It is important that this analysis be performed in a way that isolates the impact of the change in the pay system without simultaneously allowing other fac- tors to change. If, for example, implemen- tation of the new pay structure causes some employees to quit and others to enter the state system, then a comparison of pay structures before and after implementation PAY EQUITY: EMPIRICAL INQUIRIES will be biased by differences in the sample of employees. Further, any methodology that takes a snapshot of the pay structure before and after comparable worth will run into problems of biases due to changes in other exogenous influences, such as busi- ness cycles and political elections, that may also alter the state pay structure. The snap- shots would also fad! to capture the full impact of comparable worth since the plans may be introduced gradually or may not be implemented in full, such as in Iowa. We solved these problems by using the December 1983 pay schedule, which existed before the comparable worth study was ini- tiated. By determining the individual's ac- tual 1983 pay grade and the number of pay grades that the individual would have in- creased or decreased due to comparable worth, we could compute what the indi- vidual's corresponding biweekly earnings would have been in 1983 under each of the alternative plans. Given the recommended pay grades, we could compute three earn- ings rates for each employee: (1) the actual 1983 earnings, (2) the earnings rate asso- ciated with the final Arthur Young rec- ommendations, and (3) the earnings rate associated with the implemented state/ AFSCME compromise plan. In this way, our analysis avoids the biases associated with job entry and exit, business cycles, inflation, and other changes that occur over time. To explore how comparable worth altered the earnings structure in the Iowa merit system, we used the standard earnings func- tion approach pioneered by Mincer (19741. Earnings were presumed to be related to individual and job characteristics according to I 1nWk;^ = '7 km + hk,iX. + ek^Z. + ek — U ~ V —i — v—J - — y (1) where inWkij is the natural logarithm of the biweekly wage of individual i in job cIas- sification j in pay plan k, Xi represents a set of socioeconomic characteristics for in- dividual i, and Zj represents a set of job characteristics. For our purposes, elements

OCR for page 179
THE IOWA CASE of the set of variables X' are common ele- ments of the standard human capital mode} of earnings, such as inclividual education, experience, and job tenure (see Willis, 1986, for a survey of this literature). The param- eters akO, boo, and Cko are specific to the pay plan k, and ekij is the error term. The regression coefficients can be interpreted as the percentage change in earnings as- sociated with a one-unit change in a given characteristic, holding the other character- istics constant. One exception is the market wage variable, which was entered as the log of market wage. In this case, the coef- ficient can be interpreter] as an elasticity, indicating the percentage change in the individual s earnings associated with a 1 percent change in the relevant market wage. The coefficients can be compared across pay schedules to determine how the percentage returns are affected by comparable worth. Another way of making these comparisons is to explain the difference in pay between a comparable worth plan (Wk) and the actual 1983 pay plan (W°), using the form nwkij — 1nW0ij = alp + bkiXi + cozy + ukij. (2) In this case, bki and Cki measure the per- centage increase in wages in the kth pay plan relative to the 1983 wage, W°ij, asso- ciated with a one-unit change in the char- acteristic. A positive coefficient implies that the characteristic wins as a result of com- parable worth, and a negative coefficient means that the returns to the characteristic fall. Our source of information on individual pay grades, job classifications, job step, pay plan, and biweekly pay in 1983 were ob- tained from the Iowa merit system payroll tapes. The tapes also yiel(led information on an employee s race, sex, marital status, age, ant] employment clate with the state. Information on whether an employee held professional or supervisory status, whether the individual was covered by a union con- tract, an(1 whether the individual paid dues to a union or professional society through ]83 payroll deductions was also available from this source. The merit employment de- partment s personnel record] files were the source of data on educational background and work history. We ranclomly sampled from the personnel files and then merged those observations with the data from the payroll tape. Our final sample was 3,734 persons, roughly one- fifth of the state merit employees in De- cember 1983. Regents system employees of the state universities were exclucle(l from our sample. The final report of the Arthur Young study (1984) yielded information on the factor points an(l on the pay grade recommen- ciations for each job in our sample. The state provided tables of the 1983 pay schecI- uTe, the pay grades as implemented in 1985, and the number of male and female incum- bents in each job as of early 1984. We used independent wage survey data (generally private sector) from the lob Service of Iowa (1984) to measure median market wages. For a few occupations for which national markets exist and for which Iowa data were limited, we used the median wage from the 1983 Current Population Survey (unpub- lishec] data). By matching job descriptions in the state merit system with occupational descriptions in the wage surveys, we were able to measure the opportunity cost of state employment as the wage of the market job most closely associates] with the job in the state system. We were not able to account for differences between state and market fringe benefits, job security, or other vari- ables, which as compensating differences, could explain, in part, the direct pay dif- ferentials. Nor can we account for cost-of- living clifferentials. This should not be a major problem because the great majority of state employees live and work in the Des Moines metropolitan area. Nor clid the state or the Arthur Young study take living costs into account. Table 8-1 provides descriptive statistics for the variables used in our analysis. Fifty percent of our sample are females, which

OCR for page 179
184 Cal o - - :^ .~ V) .= 5 .= ,= _ V: C) ._ . _ C;: ~ a; i_ ~ o JO ~ _ 50 Cal Cal Can V) oo ~ ~ 2 G) oc CD ~ rid ~ ~ ~ ~ CC Cal ~ ~ ~ O car ~ CD ~ ~ O oO ~ ~ X A) Cal ~ ~ - 1 0 0 CD ~ X ~ ~ I) ~ ~ X 5) ~ Cal X ~ c: CD ~ Cal X X ~ CO O CD cot ~ ~ ~ ~ ~ co ~ car fir 0 ~ ~ ~ ~ ~ ~ 0 cut . . . . . . . . . . . . . . . . . . . . . . . . . O ~ ~ ~ ~ C>1 X O Cat Cal o ~ ~ CS — ~ Cat ~ fig 6g erg fig u: can ~ ~ ~ C5, ~ ~ ~ ~ fir ~ ~ co O c~ _ ~r ~ ~ ~ ~ oo cc oC CS ~ O c~ 0 O ~ ~ ~ ~ ~ c~ ~ x c~ ~ x ~ oo ~ c~ co _ c~ x O ~ CD O O _ O C~ ~ — ~ X ~ C~ C~ _ C~ ~ ~ _ . . . . . . . . . . . . . . . . . . . . . . ~ _ CO ~ C~ ~ (D CO ~ C~ u: c~ _ _ e3 6~3 00 _ C~ CD 0 CD ~ ~ ~ ~ 0 C~ — — — ~ ~ ~ ~ ~ ~ ~ ~ c~ x c~ ~ O oo ~ C;, c~ ~ O x _ ~ U: ~ C~ ~ C~ O d~ ~ _ O _ ~ ~ O _ C~ _ ~ ~ ~ ~ O X ~ ~ ~ ~ ~ c~ _ c~ ~ ~ x O c~ ~ ~ _ X ~ c~ _ ~ _ u: c~ . . . . . . . . . . . . . . . . . . . . . . . . . ~ ~ ~ ~ co c~ ~ 0 c~ c~ 0 ~ ~ r~ c~ ~ _ 69 ~3 co cs oo c~ 0 ~ ~ ~ O ~ _ ~ 00 _ ~ C~ — ~ X C~ _ C~ O ~ O CO X CD ~ ~ O oo cS' oo ~ co ~ c~ ~ c~ _ ~ ~ ~ ~ ~ x ~ ~ ~ c~ ~ — 0 ~ c~ c~ u: 0 0 0 ~ ~ ~ ~ c~ _ c~ ~ c~ ~ _ 0 c~ 0 . . . . . . . . . . . . . . . . . . . . . . . ~ CD ~ C~ O C~ ~ O ~ _ c~ ~ e3 co ~ _ eg eg ~ ~ 0 ~ ~ O oo co u: c~ d4 ~ 1 - 0 _ C] X c~ _ ~ U: CS 00 — C~J _ ~ CM CC:) 0 C~ C~ C~ O _ X C~ 0 U: _ LO ~ 0 00 ~ — C~ ~ 0 0 CO CS ~ ~ — CO _ C~ O X O ~ C~ ~4 C~ O O O O C~ O 0 =4 0 0 — 0 C~ ~ — O ~ O O ~— . . . . . . . . . . . . . . . . . . . . . . . . . ~ — C~ C~ CS _ C~ _ _ C:) ~ _ CD ~ 69 c~ _ ~ ~ o 0 _ 0 o ~ u: ~ CD C~ C~ CO CK1 U') ~ X ~ C~ ~ C~ 1~ C~) ~ ~ C~ ) 00 0 0 0 ~4 — X ~ ~ ~ ~ — CD 0 _ C~ 0 ~ CO O ~ 00 ~ ~ eo C~ C~ 0 0 — ~ X ~ ~ C~ O O O _ CM C~ ~ _ _ C~ 0 X ~ — O C~ O ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . Co _ ~ ~ 0 C~ X 00 — r~ ~ c~ co _ C~ CD CC 6g e3 C~ C~ 00 0 — ~, ~ ~ ~ ~ ~ _ ~ ~ o ~ — ~ 5 ~ 3 — ~ C ~ ~ ~ ° ~

OCR for page 179
THE IOWA CASE differs only slightly from the 47.3 percent prior to sampling. In contrast, private sector and local government units had work forces that were 42 percent female (in 1980~. Two percent of our sample are minorities (black, Hispanic, or American Indian). The average employee in our sample was 40.4 years old, had 13.3 years of education, 8.5 years of nonstate work experience, and 8.4 years of tenure with the state. Overall, 26 percent were college graduates, 5.3 per- cent had M.S. degrees, and 1.2 percent hacl Ph.D.'s. In our sample, 13.2 percent were supervisors and 20.5 percent were professionals. Because of problems encoun- tered in generating biweekly equivalent pay for part-time workers, our sample is pre- clominantly (97.7 percent) full-time workers. Overall, 6.5 percent had occupational li- censes, 19.6 percent had some vocational training, and 21.8 percent had military ex- perience. In total, 33.9 percent were single, and employees averaged 1.1 years out of the labor force after starting postschool em- ployment. We have three measures of union status. In the sample 77.4 percent were in non- supervisory jobs that were covered by union contracts. Approximately 15.2 percent paid dues to (nonprofessional) unions, and an additional 3.4 percent paid dues to profes- sional associations through a payroll check- off. On average, state employees in our sample were paid $636.14 biweekly in De- cember 1983. This corresponds to an hourly rate of $7.95 and an annual rate of $16,540. Table 8-1 also provides means and stan- dard deviations separately for each sex. In 1983, women averaged 78.1 percent of the average biweekly mate wage. This would have increased to 86.2 percent uncler the Arthur Young plan but increased to 81.8 percent upon actual implementation. Wom- en averaged 13.1 years of school compared with 13.6 years for men. Men were much more likely to have an M.S. or Ph.D. cle- gree, and women were more likely to have a license or vocational training. Men av- 185 eraged 9.8 years of tenure with the state and an additional 10.5 years of nonstate work experience. Women workecl less; they hacT 6.9 years of tenure and 6.4 years of other experience, on average. Women had 2.0 years out of the labor force versus only .3 years for men. Men were more likely to be supervisors and professionals. Although men were no more likely than women to be covered by a collective bargaining con- tract, men were more likely to be dues- paying members. Women were much more likely to be single and were slightly younger (39.5 years old) than men (41.4 years oIcl). Using our measures of opportunity wages, women averaged a $6.65 market wage and men averaged $9.26. If these wages accu- rately reflect actual relative market oppor- tunities, female employees would have been paid 28.2 percent less than men in alter- native market jobs but were only paid 21.5 percent less in their state jobs in 1983. This may suggest that the state was less discrim- inatory toward women even before com- parable worth, or that nonstate employers hire workers with a ctifferent set of personal characteristics. Female state employees were more likely to be concentrated in female jobs, as expected. The average value for "percent female job" was .78 for women and .22 for men. RESULTS An examination of sample means indicates that public employees gain on average from comparable worth. Although some would have surreal pay cuts under the Arthur Young/committee proposal, the average em- ployee in our sample would have gained $35.57 biweekly ($925 per year). In addi- tion, the variance of pay across workers was reduced. Had the original plan been im- plemented, 10 percent of the employees would have gained $3,515 per year, and another 10 percent would have lost $1,380 or more per year. Women would have gained

OCR for page 179
186 $63.45 biweekly and men wouIc] have gained only $6.38 on average. As discussed above, the state/AFSCME compromise of 1985 eliminates] all pay cuts in exchange for moderating the pay increas- es. On average, biweekly pay increased $39.74 ($1,033 annually); 10 percent re- ceived more that $2,850 annually. Recall that all of our analysis is in terms of the 1983 pay schedules so that these figures are estimates of real pay changes, not just nom- inal pay changes. Implementation tended to raise male gains to $30.27 biweekly and slightly reduce female gains to $51.31, com- pared with the more generous Arthur Young proposal. The compromise plan also tended to increase the variance in pay across work- ers relative to the proposed plan. Compared with the 1983 pay plan, the compromise appeared to increase the dispersion of pay within sex groups, but reduced the disper- sion of pay overall. Tabulations Table 8-2 provides more detailed data on average pay changes tabulated by several demographic and human capital variables. Un(ler both plans, women gaine(1 consicI- erably more than dill men. lobs that were 41 percent female or higher gained sub- stantially more than jobs that were only 40 percent female or lower. In fact, employees in the latter jobs would have lost pay, on average, had the Arthur Young recommen- (lation been implemented. (The proportion female in the "100 percent female job" category differs slightly from 1.0 due to a 2-month mismatch between data sources.) As for the racial impact, the earnings gap between minorities and whites increased as a result of the implemente(1 pay plan, but it wouIcl have been only marginally affected by the proposed plan. Clearly, any general conclusions regarding minorities must be drawn with caution because of the small number of minorities employed in the state system. PAY EQUITY: EMPIRICAL INQUIRIES As expected, the least educatecl and those with the least tenure gained the most. Un- der the Arthur Young plan, college gracI- uates would have lost income. A similar but less pronounced effect occurs for (nonstate government) work experience. There is also a clear tendency for those with the lowest paying market alternatives to gain the most. The compromise plan decreases the gains to the lowest wage occupations while ben- etiting higher wage occupations, but the pattern of larger relative increases for the lower paid occupations remains. Women are relatively more concentrated in the cat- egories with the least tenure, experience, and education (except high school ciropouts) and in the lowest paying jobs. Workers in health, social work, clerical, and education consistently gained the most under both plans. Those occupations also hacI relatively high proportions of female incumbents. At the other en(l of the spec- trum, workers in computation, finance, transportation, regulation, employment ser- vices, ant] law enforcement jobs would have lost income under the proposed plan and gained the least under the compromise plan. Those jobs tended to have higher concen- trations of male incumbents. Supervisors and professionals would have pained the least under the Arthur Young plan but ac- tually did better than their colleagues once the compromise plan went into operation. Unions gained more under the proposed plan but (li ;l less well than nonunion workers under the compromise plan. Somehow, dues- paying union members clid better than the average worker covered by the union con- tract or the average noncontract worker under both plans. It is also interesting that the pattern of shrinking wage differentials between high- an(l low-skill workers in the plan is similar to the typical effect of unions on wage differentials by skill level. Although strongly suggestive, the simple averages reported in this section shouIc] not be overemphasized because no control vari- ables were used. We now report the results

OCR for page 179
THE IOWA CASE of our multivariate regression analysis, in which indiviclual effects were estimated while holding other variables constant. Human Capital Mode} The basic results for Equation (1) are reported in column 1 of Table 8-3. The estimates are very consistent (at least in sign) with wage equations estimated by oth- er economists. This implies that wage pat- terns for state government employees in Iowa do not differ fundamentally from wage patterns typically studied in the private sector for the United States as a whole. In particular, the 1983 regression indicates that pay increases significantly with most of the educational and work experience variables, as well as with the market wage and union status. Overall, the human capital mode! performs very well, explaining 81 percent of the variation in the logarithm of the wage rates. Women in Iowa state government earned 22 percent less than men, on average, with- out using control variables. Women, how- ever, earned only 4 percent less than men after controlling for human capital and mar- ket wage variables, and 12 percent less than men after controlling for human capital vari- ables except the market wage (see column 1 of Table 8-6~. As found in other studies, women earned less primarily because they had less human capital on average and be- cause they tended to be employed in lower paying jobs relative to men. An identical mode! was estimated using cross-sectional data for both of the com- parable worth pay plans (columns 2 and 3 of Table 8-3~. The estimates are very similar in sign and significance. However, the R2 drops, inclicating that the human capital mode} is less useful in explaining the com- parable worth plans. This is not surprising in that variables such as market wages, education, and experience were given Tower weights under the comparable worth plans. It is easier to contrast the comparable 187 worth pay structures with the original 1983 estimates of Equation (21. These are presented in col- umns 4 and 5 of Table 8-3. The dependent variable is the difference between the pro- pose(1 comparable worth pay level and the original 1983 pay level. The coefficients of column 4, in essence, summarize whether the corresponding coefficients increase or decrease (and by how much) as we compare column 1 and column 2. Likewise, column 5 contrasts columns 1 and 3. Of most interest, women gain relative to the 1983 pay plan. As indicatecI by the dummy variable "female," the Arthur Young/ committee plan would have narrowed the underpayments by 2.8 percentage points while the 1985 compromise narrowed the gap by .7 points. The main explanation for the smaller relative gain for women in the compromise plan is un(loubte(lly the fact that no jobs (especially male jobs) suffered pay cuts on enactment of the plan. Table 8-3 also provides insights concern- ing the relationship between the control variables and the pay plans. We hypothe- sized that comparable worth would deem- phasize education anti work experience, and the results support that expectation. The Arthur Young plan, as well as the 1985 compromise plan, reduced the magnitude of the education, experience, and tenure coefficients. On the other hand, there is evidence that the plans placecl increased emphasis on credentials or diplomas as op- pose(1 to years of eclucation. This is consis- tent with our expectation that comparable worth plans would tend to move the pay structure closer to the form typically as- sumed in the job signaling literature (e.g., Spence, 1973~. This is particularly true for the Ph. D. and M. S. degrees and for licen- ses, and somewhat less so for vocational . . tralnlng. Both plans significantly (reemphasize the role of market wages as a determinant of government pay. The Arthur Young plan wouIc] have reduced the influence of market pay structure by examining ~ . . /~\ I'

OCR for page 179
188 TABLE 8-2 Average Annual Pay Change Relative to 1983 Pay PAY EQUITY: EMPIRICAL INQUIRIES Arthur Young 1985 Percent Subgroup Committee Compromise Female Females $1,682 $1,360 100.0 Males 170 707 o. 0 100 percent female job 1,822 1,491 98.4 81-g9 percent female job 2,047 1,,382 9s.3 41-80 percent female job 1,444 1,397 6L8 1-40 percent female job -244 380 15.1 O percent female job -145 723 o.o Whites 92s 1,037 49.8 Minorities 922 846 s5. 8 <12 yrs. school 1,653 1,328 3s.7 12 yrs. school 1,403 1,181 s7.0 13-15 yrs. school 833 1,022 53.6 16 yrs. school -170 694 41.2 >16 yrs. school -164 611 33.9 <2 yrs. experience 999 1,126 s8.2 2-s yrs. experience 1,064 1,092 ss.4 6-lo yrs. experience 817 967 s0.8 11-20 yrs. experience 9s8 1,olb 46.3 21-30 yrs. experience 6s3 883 29.3 >30 yrs. experience 64s 77s 11.7 <2 yrs. tenure 1,331 1,180 54.6 2-s yrs. tenure 973 1,039 s4.9 6-10 yrs. tenure 982 1,036 s6.s 11-20 yrs. tenure 563 896 42.6 21-30 yrs. tenure 39s sss 2s.3 >30 yrs. tenure 97 714 23.4 Part-time work 1,894 1,197 70.7 Full-time work 880 1,026 49.0 Health/medical 2,462 2,838 80.8 Social work 1,966 1,857 69.7 Clerical 2,146 1,128 96.7 Liquor stores 1,541 1,129 40.4 Education 4ss 894 67.1 Parks, agriculture 68s 832 ll.o Crafts 709 739 4 7 Service 1,694 671 67.4 Law enforcement -423 351 lo. 1 Employment services -2s3 328 43.2 Regulation -612 318 16.4 Transportation -1,115 241 12.3 Tax/finance -1,226 22s 36.6 Computation -2,398 160 34.4

OCR for page 179
THE IOWA CASE TABLE 8-2 Continuect ~9 Arthur Young 1985 Percent Subgroup Committee Compromise Female <$5 market wage 2,911 2,109 68.5 $5-8 market wage 1,244 933 68.3 $8-11 market wage 337 865 26.8 $11-14 market wage -1,078 541 25.7 >$14 market wage 105 438 14.9 Supervisors 387 1,323 26.0 Nonsupervisors 1, 007 989 52.2 Professionals 91 1,193 38.5 Nonprofessionals 1,140 992 52.9 Union contract 1,020 1,012 50.4 No contract 598 1,107 48.1 Dues payer 1,268 1,175 39.9 Association dues payer 809 1,082 49.6 No dues 865 1,005 51.8 wages by more than 50 percent relative to the 1983 plan. The compromise plan did reintroduce some of the market's influence, but the market wage coefficient still decliner] relative to the 1983 plan by one-third on implementation. However, the Arthur Young pay plan woul(l still have retained a statistically significant relationship to market wages, with a 10 percent increase in the market wage associated with a 1.5 percent increase in state pay. In 1983, before the comparable worth study, the same 10 percent increase in the market wage was associated with a 3.5 percent increase in state pay. We hypothesized that unions would at- tempt to use the comparable worth process to their favor. This is not as clear for the Arthur Young plan, in which we see positive but small and insignificant effects. It is notable, however, that the plan imple- mented in 1985, which was a compromise between the governor and the union, sig- nificantly boosted the pay of those covered by union contracts. Dues-paying union members received an even larger increase above that going to union contract workers as a whole. There was a statistically insig- nificant increase for dues-paying members of professional associations. Interpreted lit- erally, those both covered by union con- tracts ancT paying dues to the union gained up to 2.5 percentage points in pay as a result of the compromise. The Arthur Young plan proposed placing greater emphasis on supervisory tasks, and this was also true on implementation. Professionals also gained increased impor- tance on implementation, (respite negative recommendations by the Arthur Young con- sultants. Members of minority groups lost a little over 1 percent of their earnings relative to whites un(ler both plans, although this result is not statistically significant. Military ex- perience and full-time experience tended to be Reemphasized by the Arthur Young plan, although there was no impact on im- plementation. Single employees an(l those with more years out of the labor force also appear to have suffered some reductions in pay as a result of comparable worth, but the magnitudes of the reductions are small.

OCR for page 179
190 TABLE 8-3 Human Capital Mode! PAY EQUITY: EMPIRICAL INQUIRIES Pay Levels Pay Differences Variables 1983 Arthur Young 5.5628 (.0232 .0233 (.0011 .0119 (.0094 .1819 (.0179 .0760 (.0084 .0208 (.0039 .0098 (.0057) .0210 (.0007 - .0003 (.2 x 10-4)* .0006 (.0006) -.2 x 10-4 (.1 X 10-4) 1985 C. ompromlse .283 (.0276 .0298 (.0013 .0064 (.0111) .0770 (.0213 .0765 (.0100 .0292 (.0047 .0165 (.0067 .0213 (.0008 .0004 (.2 x 10-4)* .5 X 10-4 (.0007) .5 x 10-' (.2 x 10-4) .0013 (.0007) .0868 (.0118 .1841 (.0100 .1401 (.0199 .2274 (.0086 .0409 (.0079 .0208 (.0063 .0332 (.0122 .0067 (.0047) .0123 (.0152) .0319 (.0057 .763 Arthur 1985 Young Compromise Intercept Education M.S. degree Ph.D. degree License Vocational Military Tenure Tenure2 Experience Years out Supervisor Professional Full time Market wage Union contract Dues payer Assoc. dues payer Single Minority Female 9198 0266 .0353 0012 0097 0108) .0444 0205 .0112 0096) .0175 0045 .0169 0065 .0227 0008 0004 3 x 10-4)* .0005 0007) Experiences - .4 x 10 - 5 2 x 10-4) 0006 0006) .0547 0113 .1514 0096 .1367 0191 .3503 0083 .0270 0076 .0099 0061) .0229 0117) 0053 0045) .0006 .0147) .0389 .0055 .815 R2 .0008 (.0006) .0913 (.0099 .1483 (.0084 .1131 (.0167 .1497 (.0073 .0300 (.0067 .0133 (.0053 .0345 (.0103 .0093 (.0040 .0107 (.0128) .0113 (.0048 .760 .643 (.0169 .0119 (.0008 .0217 (.0068 .1375 (.0130 .0648 (.0061 .0033 (.0029) .0072 (.0041) .0016 (.0005 .2 x 10-4 (.1 X 10-4) .0001 (.0004) -.1 X 10-4 (.1 X 10-4) .0002 (.0004) .0366 (.0072 .0058 (.0061) .0237 (.0121) .2006 (.0053 .0030 (.0048) .0034 (.0039) .0116 (.0075) .0040 (.0029) .0113 (.0093) .0276 (.0035 .543 .3632 (.0131 .0054 (.0006 .0034 (.0053) .0326 (.0102 .0654 (.004)* .0117 (.0022 .0003 (.0032) .0013 (.0004 .2 x 10-4 (.1 X 10-4) - .0004 (.0003) —.1 x 10-0 (.1 X 10-4) .0007 (.0003 .0322 (.0056 .0300 (.0048 .0033 (.0095) .1229 (.0041 .0138 (.0037 .0108 (.0030 .0103 (.0058) .0014 (.0022) .0128 (.0073) .0069 (.0027 .373 *Significant at .05 level. It must be emphasized that here we are analyzing the partial impact of an indivicI- ual variable while holding other variables constant. Thus, when we refer to losses by more educated workers or by minority workers, for instance, we mean that those people received relatively smaller pay in- creases than did the less educated or whites.

OCR for page 179
THE IOWA CASE It must be stressed that since the 1985 compromise eliminated all pay cuts, no workers faced immediate Tosses in absolute terms. Comparable Worth Mode} Comparable worth advocates might ar- gue that a better specification of our earn- ings function would delete the market wage and add a variable that measures the proportion of each job classification that is female. According to this perspective, the market wage should be excluded be- cause it reflects existing sex discrimination in the private labor market. Since the purpose of comparable worth is to elim- inate sex discrimination, they argue that market wage information should be ig- nore(1 in evaluating pay. Comparable worth models use a percent female variable as a proxy for discrimination that results from the classification of a job as being "female." In other words, with human capital held constant, the hypothesis is that female- dominatec] jobs will be paid less than male- clominated jobs. It is interesting to observe the extent to which our conclusions are altered when we incorporate these two changes into the earnings equations. The results are pre- sented in Table 8-4. In virtually all cases, our conclusions are reaffirmed. A few com- ments are appropriate, however. The R2 for the 1983 pay structure drops from .815 to . 769, indicating that market wages were an important determinant of government pay. In this formulation, it appears that the Arthur Young plan would have sig- nificantly improved the pay of union em- ployees and dues payers. The earlier gains for professional workers resulting from the 1985 implementation no longer appear, and the gains for Ph.D.'s and the voca- tionally trainee] are no longer statistically significant. The main interest in comparing the earlier 191 human capital regressions with this speci- fication is in examining the implications for the pay of women relative to men. By excluding market wages and including the percentage of female incumbents on the job, the coefficient estimates on the female (rummy variable in the pay-level regressions become positive but not significantly clif- ferent from zero, and the comparable worth plans actually appear to lower the relative pay for inclividual women. However, the coefficients on percent female incumbents in the job classification are negative and quite large in absolute value. Taken literally, these coefficients in the 1983 pay structure imply that an increase of 10 percentage points in the percentage of women in an occupation lowers pay by 2.6 percent. This implies that women received lower pay be- cause they were in "female jobs" that paid less. There was no evidence of additional discrimination against women as indivicluals since the coefficient on the female dummy variable is positive. The Arthur Young plan would have cut the reduction in pay (as- sociated with a 10 point increase in the percentage of female incumbents in the job) to 1 percent, but the compromise plan cut the pay reduction to only 1.8 percent. The comparison of the results of Tables 8-3 and 8-4 clearly illustrates the nature of the (rebate regarding the validity of claims that female-dominate(l jobs are systemati- cally un(lervaluecl by the market. Propo- nents of human capital or market expla- nations could point to the results of Table 8-3 and claim that measured discrimination against women is very slight, controlling for human capital and labor market demand an(1 supply factors. Proponents of compa- rable worth could point to the results of Table 8-4 and claim that large discrepancies in pay exist between men and women be- cause women are concentrated in jobs that are pai(1 below the value placed on com- parable male jobs. The data support both claims, an(1 therefore, any assessment that

OCR for page 179
192 TABLE 8-4 Comparable Worth Mode} PAY EQUITY: EMPIRICAL INQUIRIES Pay Level Pay Differences Variables 1983 Arthur Young 5.77 (.023 .029 (.001 .007 (.010) .207 (.018 .097 (.009 .030 (.004 .007 (.006) .023 (.001 - .0004 1985 Compromise 5.609 (.027 .038 (.001 .014 (.012) .111 (.022 .109 (.010 .045 (.005 .011 (.007) .024 (.001 .0004 (.2 x .002 (.001 .4 x (.2 x 10-4) .002 (.001 .065 (.012 .230 (.010 .211 (.021 .032 (.008 .0006 (.007) .053 (.013 .006 (.005) .021 (.016) .003 (.007) .175 (.009 .742 Arthur Young .353 (.017)* .020 (.001)* .029 (.007)* .109 (.014)* .035 (.007)* .011 (.003)* .002 (.004) .004 (.001)* .6 x 10-4 (.2 x 10-4)* .001 (.0005)* .2 x 10-4 (.1 X 10-4) .0003 (.0005) .057 (.008 .046 (.006 .086 (.013 .012 (.005 .022 (.004 .006 (.008) .005 (.003) .004 (.010) .008 (.005) .164 (.006 .472 1985 Compromise .191 (.013)* - .010 (.001)* .008 (.006) .013 (.011) .048 (.005)* .004 (.002) .002 (.003) - .003 (.0004)* .5 X 10-4 (.1 X 10-4)* - .001 (.0004)* .2 x 10-4 (.1 X 10-4) .0004 (.0004) .044 (.006)* .004 (.005) .035 (.010)* .019 (.004)* .021 (.003)* .0008 (.006) .002 (.002) .009 (.008) .006 (.004) .083 (.005)* .284 Intercept Education M.S. degree Ph.D. degree License Vocational Military Tenure Tenure2 Experience Experience2 Years out S- upervlsor Professional Full time Union contract Dues payer 5.418 (.028 .048 (.001 .022 (.012) .098 (.023 .061 (.011 .041 (.005 .008 (.007) .026 (.001 .0005 (.3 x .003 (.001 .6 x (.2 x .001 (.001) .021 (.013) .226 (.010 .245 (.021 .014 (.009) .021 (.007 .054 (.013 Single —.004 (.005) .012 (.016) .009 (.008) Percent female job -.258 (.010 .769 Assoc. dues payer Minority Female R2 (.2 x 10-4)* .002 (.001 .4 x (.2 x .001 (.001) .078 (.010 .180 (.008 .159 (.018 .026 (.007 .001 (.006) .048 (.011 .009 (.004 .016 (.013) .0006 (.006) .094 (.008 .742 *Significant at .05 level. One view clominates the other must rest on assumption and not statistical analysis. We can, however, conclude that the standard market explanation performs at least as well as the comparable worth explanation. Human Capital-Job Attributes Model As an additional sensitivity test, we rees- timated the earnings equations by adding to our original model both the percent fe-

OCR for page 179
THE IOWA CASE male variable and the 13 job factors cle- veloped as part of the Arthur Young study. The job factors attempt to measure attri- butes of each job, such as working envi- ronment, responsibility, physical and men- tal demands of the job, and minimum requirements in terms of education and experience. Comparable worth advocates would tend to emphasize these variables in that they believe that salaries should be pegged to attributes of the job rather than only to attributes of the individuals. fob factors are also important to the extent that government bureaucracies and civil service agencies downplay individual characteristics except in relationship to job characteristics. By holding constant job attributes and per- cent female, we can determine whether our conclusions regarding the effects of com- parable worth on individuals are sensitive to the inclusion of these variables. With a few exceptions, our conclusions do not change. Some might question our strategy of re- gressing proposed wage rates on job attri- butes in the sense that Arthur Young gen- eratec] their pay proposals from these same job attributes. The 13 variables should ex- plain most of the variations in nav under their proposal. On the other hand, this shouIc] not be a problem for the 1983 pay structure since it existed prior to the Arthur Young study. Nor should it be as much of a problem for the 1985 implementation since political compromises partially clistorted the relationship between factor points and pay. Nevertheless, it is important to keep in mind that these regressions will yield in- formation on the changes in marginal re- turns to personal characteristics above and beyond the changes due directly to the formal comparable worth process of mea- suring and weighting job factors. Because personal characteristics are undoubtedly correlated with (and perhaps causally re- latec] to) the level and weights attached to the job characteristics, the earlier estimates are better measures of the gross changes in 193 marginal returns to personal characteristics. The equations in Tables 8-3 and 8-4 also avoid overparameterization and the intro- duction of spurious regressors that may ar- tificially explain the wage differential be- tween men and women. Our estimates are presented in Table 8-5. The regression for 1983 pay is in column 1 and may be contrasted with the compa- rable estimates of Tables 8-3 anti 8-4. First, it is clear that even in 1983, the job attributes acid a lot to the explanatory power. The R2 rises from .815 to .932, and 12 out of 13 job attributes are statistically significant. It is also interesting that all have positive coefficients except physical effort and pace of work. Interpreter! literally, the 11 positive job attributes appear to be measuring as- pects of the job that are undesirable or costly to acquire and that require compen- sation in order to attract applicants, whereas physical effort appears to be a desirable job attribute. Of course, we must modify this to the extent that political or other non- economic forces created the pay structure. We also note that the complexity and the scope of the job have the largest positive impact on pay. Many of the human capital and personal characteristic coefficients decline in abso- lute value while retaining their sign and significance. This is not surprising in that some ofthese variables compete with certain job attribute variables. For instance, the minimum education job attribute (NED) competes with the years of education ant] the credential variables. Further, the com- parable worth pay plans represent a con- scious effort to clownplay individual char- acteristics and emphasize job characteristics. Still, few significant sign reversals occur. The supervisor variable changes from pos- itive to negative when the supervisory job factors are ad(le(l. The association dues- paying member and the vocational training variables tend to change from positive sig- nificance to negative insignificance. The only other coefficients that change sign are those

OCR for page 179
194 TABLE 8-5 Human Capital-Job Attributes Mode} PAY EQUITY: EMPIRICAL INQUIRIES Pay Level Pay Differences Arthur 1985 Arthur 1985 Variables 1983 Young Compromise Young Compromise Intercept 5.4189 5.5016 5.354 .0827 - .0652 (.0255)* (.0234)* (.0230)* (.0204)* (.0143 Education .0073 .0065 .0057 - .0008 - .0017 (.0009)* (.0008)* (.0008)* (.0007) (.0005 M. S. degree .0054 .0150 .0033 .0096 - .0021 (.0066) (.0061)* (.0059) (.0053) (.0037) Ph. D. degree .0127 .1256 .0215 .1129 .0088 (.0127) (.0116)* (.0114) (.0102)* (.0071) License .0164 .0229 .0245 .0065 .0081 (.0061)* (.0056)* (.0055)* (.0049) (.0034 Vocational .0019 - .0001 .0052 - .0020 .0033 (.0028) (.0025) (.0025)* (.0022) (.0016 Military .0002 - .0008 .8 x 10-4 - .0011 - .0001 (.0039) (.0036) (.0035) (.0031) (.0022) Tenure .0179 .0183 .0177 .0004 - .0002 (.0005)* (.0005)* (.0004)* (.0004) (.0003) Tenures - .0003 - .0003 - 0003 - .7 x 10 - 5 - .2 x 10 - 5 (.2 x 10-4)* (.2 x 10-4)* (.1 X 10-4)* (.1 X 10-4) (.9 X 10-5) Experience .0012 .0014 .0010 .0002 - .0002 (.0004)* (.0004)* (.0004)* (.0003) (.0002) Experiences -.3 x 10-4 -.2 x 10-4 -.2 x 10-4 -.5 X 10-8 .8 x 10-5 (.1 X 10-4)* (-1 X 10-4)* (.1 X 10-4) (.9 X 10-5) (.7 x 10-5) Years out - .0004 .8 x 10-4 - .0002 .0005 .0002 (.0004) (.0004) (.0004) (.0003) (.0002) Supervisors —.0598 - .0374 - .0172 .0223 .0425 (.0086)* (.0079)* (.0078)* (.0069)* (.0048 Professionals .0612 .0294 .0358 - .0317 - .0254 (.0062)* (.0058)* (.0057)* (.0050)* (.0035 Full time .0942 .0830 .0854 - .0112 - .0087 (.0119)* (.0109)* (.0107)* (.0095) (.0066) Market wage .1248 .0391 .0404 - .0857 - .0844 (.0080)* (.0073)* (.0072)* (.0064)* (.0045 Union contract .0245 .0060 .0106 - .018 - .0139 (.0048)* (.0044) (.0043)* (.0039)* (.0027 Dues payer .0192 .0074 .0127 - .0118 - .0065 (.0039)* (.0035)* (.0035)* (.0031)* (.0022 Assoc. dues payer .0053 - .0013 - .0049 - .0066 - .0102 (.0073) (.0067) (.0066) (.0058) (.0041 Single —.0019 - .0048 - .0013 - .0029 .0006 (.0028) (.0025) (.0025) (.0022) (.0015) Minority —.0076 - .0025 - .0086 .0050 - .0010 (.0089) (.0082) (.0080) (.0071) (.0050) Female - .0086 - .0078 - .0109 .0008 - .0023 (.0041)* (.0038)* (.0037)* (.0033) (.0023) Percent female job - .1459 .0016 - .0582 .1475 .0877 (.0070)* (.0065) (.0063)* (.0056)* (.0039 JE D .0192 .0251 .0418 .0059 .0225 (.0018)* (.0016)* (.0016)* (.0014)* (.0010 JEX .0091 .0011 .0203 - .0080 .0112 (.0020)* (.0018) (.0018)* (.0016)* (.0011 JCPX .0567 .0354 .0372 - .0213 - .0195

OCR for page 179
THE IOWA CASE TABLE 8-5 Continued ]95 Pay Level Variables 1983 Arthur Young (.0028 .0193 (.0029 .0175 (.0014 .0355 (.0020 .0181 (.0027 .0311 (.0016 .0235 (.0029 .0326 (.0027 .0038 (.0028) .0101 (.0020 .0079 (.0021 .903 1985 Compromise (.0028 .0285 (.0029 .0221 (.0014 .0314 (.0020 .0181 (.0027 .0150 (.0016 .0323 (.0029 .0290 (.0026 .0094 (.0028 .0226 (.0019 .0066 (.0020 .935 Pay Differences Arthur 1985 Young Compromise (.0025 - .0026 (.0026) .0038 (.0012 .0592 (.0018 .0036 (.0024) .0116 (.0014 - .0087 (.0026 .0146 (.0023 —.0130 (.0024 .0035 (.0017 .0107 (.0017 .735 JGD ICON JPH JMV JSUP JSCP JIMP JENV JHAZ JPAC R2 (.0031 .0220 (.0032 .0137 (.0016 .0237 (.0022 .0144 (.0030 .0195 (.0018 .0322 (.0032 .0179 (.0029 .0167 (.0031 .0065 (.0022 .0029 (.0022) .932 (.0017 .0066 (.0017 .0084 (.0009 .0551 (.0012 .0036 (.0017 .0045 (.0010 .5 x 10- (.0018) .0111 (.0016 .0074 (.0017 .0161 (.0012 .0095 (.0013 .708 NOTE: The job factors are JED, the minimum education required for the job; JEX, the minimum time required to gain the knowledge necessary for the job; JCPX, the complexity of the job; JGD, the guidance or supervision available; ICON, the number of personal contacts; JPH, physical effort and fatigue; JMV, mental and visual coordination; JSUP, supervision exercised; JSCP, scope and elect of duties; JIMP, impact of errors; JENV, working environment; JHAZ, unavoidable hazards and risks; JPAC, ' ' work pace, pressures, and interruptions. *Significant at .05 level. for the M.S. degree, military, and years out of labor force variables. For the most percent impact m ladle 5-0J. part, these coefficients remain insignificant. Even controlling for the percent female job variable, the inclividual female variable is still negative and significant, albeit small in magnitude. The market wage response elasticities de- cTine in magnitude once the job factors are added. In 1983, a 10 percent increase in market wages increased state pay by only 1.2 percent (versus 3.5 percent from Table 8-3~. It is not surprising that once the in- stitutional factors that determine pay are included, the impact of the market pay variable declines. Nor is it surprising that each comparable worth plan reducer! the impact of market wages even further (a .4 . . ,~ . . ~ _\ It is clear that both the indiviclual attri- butes and the job attributes contribute to the explanation of variation in 1983 pay levels. Tests of the joint significance of the coefficients on individual attributes and on job characteristics easily rejected the null hypothesis of no effect. This strongly implies that both external competitive markets and internal pay determination are important. As before, estimates of the percent female coefficient imply major gains for female jobs under the comparable worth plans relative to 1983. Whereas a 10 percent increase in the percentage of women in the job reduced pay by 1.5 percent in 1983, this is narrowed

OCR for page 179
196 to .6 percent less on implementation. The Arthur Young plan would have totally elim- inated the pay gap between jobs, although inclividual females would have been paid .5 percent less, which is slightly smaller than the .9 percent estimate for 1983. Comparable worth still tends to reduce the importance of years of education and experience as measured by the individual characteristics, and to increase the impor- tance of credentials. Several of the coeffi- cients, however, are no longer statistically significant, probably due to competition from the job factors. Supervisors still gain from comparable worth, but professionals lose relative to nonprofessionals. A major change occurs for the union variables. In the earlier estimates, it ap- peared that those covered by a union con- tract and those paying union dues gained, at least from the 1985 compromise. Once the job factors are added, however, union workers appear to lose relative to nonunion workers. All six pay difference coefficients are negative in Table 8-5 and five are sig- nificant. This suggests that the union gains occurred primarily through the establish- ment and weighting of the job factors. We should emphasize that our earlier estimates obtained from the human capital earnings equation yield more meaningful estimates of the gross gains to union members as a result of the comparable worth process in that the coefficients on union status include gains that resulted from the factor point system. Table 8-5 provides estimates of the net impact of union status after holding job attributes constant. If one wants to ask what is the overall impact on unionized workers after controlling only for other personal an(3 human capital characteristics, the earlier tables are more appropriate. Overall, a(l(ling the job factors reaffirms our previous conclusions. The market wage is significantly reduced in importance by the comparable worth process. Years of ed- ucation and experience are also reduced but creclentials gain in importance. Female jobs gain a lot relative to male-dominated jobs. PAY EQUITY: EMPIRICAL INQUIRIES Predicted Pay Ratios As an alternative method of measuring the impact of comparable worth on the female-male pay gap, we used a stan(lar(l procedure introducer] by Oaxaca (1973) and others. The first step in this procedure is to estimate the human capital model sep- arately for males and separately for females. This generates an estimated pay equation (1) for each sex. The second step is to plug the individual personal and human char- acteristics of, say, women into the estimated male equation. This yields a prediction of the amount (WF) that women would have been paid given their actual characteristics but according to the mate earnings struc- ture. By comparing womens' actual wage rates (WF) with their predicte(l wage rate (WF), we can calculate a ratio (WF/WF) that one can interpret as the female to male wage ratio adjusted for differences in in- dividual characteristics between men and women. Alternatively, one could plug the male characteristics into the estimates] fe- male equation to predict male pay (WM) according to the female pay structure. Di- viding WM by actual male pay (WM) yields a ratio (WM/WM), which also is an estimate of the adjusted female to male wage ratio. In principle, either ratio is legitimate. How- ever, some have a preference for using the male equation to predict female pay because they believe, in part, that the male coef- ficients are more stable an(l precisely es- timated and less likely to reflect discrimi- nation. Our results are summarized in Table 8-6. As a base for comparison, row 1 presents the average female to male wage ratio as computed without controlling for any per- sonal or human capital characteristics. In 1983, women, on average, earne(1 78 per- cent of the average male earnings in Iowa state government. The uncorrected ratios indicate that the Arthur Young proposal would have increased the ratio to 86 percent and the actual implementation raised it to 82 percent.

OCR for page 179
THE IOWA CASE TABLE 8-6 Ratio of Female to Male Earnings Arthur 1985 Ratio 1983 Young Compromise Unadjusted WF W.\I .781 .862 .818 Adjusted for human capital model (Table 8-3) WF WA F WA! Wet .944 .998 .962 .987 .999 .989 Adjusted for comparable worth model (Table 8-4), excluding percent female job Woe ~ .877 .957 .913 WF `~ .895 .959 .926 Was Adjusted for human capitaljob attributes newel (Table 8-5), excluding percent female job We WF Was WE .918 1.007 .955 .965 .993 .986 Rows 2 and 3 of Table 8-6 show the female to male wage ratios controlling for the hu- man capital mode! using variables identical to those shown in Table 8-3 (with the dele- tion of the female variable since each equa- tion is estimated by sex). Controlling for these variables eliminates most of the pay gap even in 1983, before the comparable worth process began. In 1983, women were only paid 1 to 6 percent less than comparably skilled men, on average. The Arthur Young proposal would have eliminated all of the gap, and the comparable worth adjustments introduced in 1985 would have narrowed the gap to 1 to 4 percentage points. We also estimated the pay ratios while controlling for the comparable worth mode! and for the human capita~job attributes model. Both specifications excluded the fe- male dummy variable and the percent fe- maTe in the job. The results are reported in rows 4 and 5 and rows 6 and 7 of Table 8-6, respectively. Our conclusions are not 197 altered in principle. Use of the various specifications results in adjusted pay gaps of at most 12 percent and as little as 3.5 percent in 1983. As before, the Arthur Young plan would have eliminated most of the pay gap using the specification of Table 8-5, but would have left a 4 percent gap adjusting only for the variables in the com- parable worth specification. Implementa- tion in 1985 eliminated one-third to one- half of the 1983 pay gap. The results of this exercise strongly rein- force our previous conclusions. First, al- though women were paid less than men in 1983, the pay gap in Iowa state government was small, especially after human capital and other personal characteristics were held constant. Second, both comparable worth plans raise female pay relative to male pay, but the 1985 compromise plan was less generous to women than the Arthur Young plan. CONCLUSIONS Our results strongly support our expec- tations. concerning comparable worth. Al- though the 1985 political compromise mod- erated the size of the pay increases and eliminated any pay cuts, both the original Arthur Young plan and the actual plan im- plemented in 1985 increased the pay of women relative to that of men. This was accomplished by raising pay in the predom- inately female jobs. Although the Arthur Young plan would have eliminated virtually all of the underpayment gap, the compro- mise plan only eliminated a portion of that gap. Although the majority of state employees gained from the additional funds allocated in 1985 to implement comparable worth, certain groups gained more than others. Most notably, our uncontrollecl tabulations suggest that low wage earners and those with the least education and experience gained the most. These gains occurred in the health, clerical, and social service de- partments of state government, where wages

OCR for page 179
198 were low and women tended to preclomi- nate. This suggests that low-wage women tended to gain more than high-wage women. On the other hand, minorities may have lost slightly relative to higher paid whites, although it should be stresses] that Iowa has very few minority employees. Most of these conclusions are reinforced in the controlled regression models. For instance' the com- parable worth mode} Reemphasized edu- cation, experience, and market wages. A few higher paid employees tende(1 to gain relative to others. In particular, su- pervisors gained relative to nonsupervisors. In addition, those having creclentials, such as a license, vocational training, or a Ph. D., came out ahead. This suggests that com- parable worth evaluators tend to stress cre- clentials and Reemphasize years of education and experience. It is less clear whether union members gaine(l relative to nonunion employees. Our basic human capital model suggests that they did gain from comparable worth. This conclusion is also supporte(1 by the com- parable worth mode! in which the market wage is eliminated. Although the addition of job factors to the human capital variables reverses the sign on the union impact, we tend to (liscount the importance of this result. In our view, union gains came through the job factors, so it is inappropriate to control for those job factors if one wants to measure the total impact of the union. In total, unionized workers gained relative to nonunionized workers. Similar comments apply to such groups as supervisors and professionals. Both comparable worth plans tended to reduce the role played by market wages in the state pay plan. However, neither plan was so ra(lical as to eliminate totally a role for market wages, as measured in our regres- sions. Even after controlling for job factors and for the percent female, market wages and human capital variables stfl} significantly explain a proportion of the post comparable worth variations in pay rates. PAY EQUITY: EMPIRICAL INQUIRIES As we emphasized above, no state em- ployee lost pay in absolute terms in the short run as a result of comparable worth, even though some employees lost in relative terms. In the long run, however, the Tosses in relative pay may eventually become ab- solute losses. The reason is that the costs of implementing the comparable worth plan were substantial, particularly in the face of significant budget problems at the state level in Iowa. This would imply that future overall pay growth in the Iowa state merit system may be constrained due to the acIded costs of the comparable worth adjustments. If, indee(l, pay growth in the Iowa public sector slows relative to pay growth in the private sector, the real earnings for those suffering relative reductions in earnings may even- tually fall. As a final point, we can only speculate as to whether the conclusions obtained in the Iowa case can be generalizes! to other states or localities. There are several reasons why the Iowa case would seem to be typical of comparable worth plans in general. First, the metho(lology used by the Arthur Young consultants was quite standard in the area of job analysis, an(l in fact, Arthur Young has been quite active in performing such analyses in other states. Seconcl, AFSCME is the largest public sector union and has been quite active in the comparable worth (rebate in other states. It seems likely that some type of compromise would occur, such as that struck in Iowa, to prevent pay cuts. On the other hancI, it is clear that the comparable worth process in Iowa was open- ly, an(1 perhaps uniquely, influenced by economic and political factors and the sub- jectivity of committee-assigne(1 factor weights. We have no basis for judging whether this experience is common to other settings. We wouicI not be surprised, however, to clis- cover that economics, politics, and the val- ues of those involved in the evaluation pro- cess would be very important in shaping the outcomes of pay analyses done else- where. Governmental budget constraints,

OCR for page 179
THE IOWA CASE for example, would tend to cause pressure toward moderation on We parties. We would anticipate that the issue of pay cuts and reallocation of resources would arise and lead to opposition (especially from unions representing men, who might lose). This implies that in order to build support for a comparable worth plan, the scope of the plan (in terms of dollars, number and type of jobs analyzed, and potential size of pay increases or decreases may have to be limited ex ante. Clearly, such compromises need! not take the exact form as in Iowa, but the potential for pressure to compromise at some stage in the comparable worth process would exist in all states. ACKNOWLEDGMENTS The authors would like to thank the Na- tional Research Council for funding. Jeff Greig and Kyle Stevens provided able re- search assistance. Members of the Iowa State departments of Transportation, Per- sonnel, and Revenue and Finance provicled invaluable assistance in obtaining and in- terpreting the data. We would also like to thank our colleague km Prescott, workshop 199 discussant Cathy Schoen, and members of a September 1987 Pay Equity workshop for their comments and suggestions. REFERENCES Arthur Young and Company 1984 Study to Establish an Evaluation System for State of Iowa Merit Employment System Classifications on the Basis of Comparable Worth. Milwaukee, Wis.: Arthur Young and Co. Job Service of Iowa 1984 Wage Survey: Iowa Statewide 1983. Des Moines. Mincer, J. 1974 Schooling, Experience, and Earnings. New York: Columbia University Press. Oaxaca, R. 1973 Sex discrimination in wages. Pp. 124-151 in O. Ashenfelter and A. Rees, eds., Discrim- ination in Labor Markets. Princeton, N.J.: Princeton University Press. Spence, M. 1973 Job market signaling. Quarterly Journal of Economics 87(August):355-374. Willis, R. 1986 Wage determinants: A survey and reinter- pretation of human capital earnings func- tions. Pp. 525-602 in O. Ashenfelter and R. Layard, eds., Handbook of Labor Economics. Vol. 1. Amsterdam: North-Holland.