s. leonard: spatial and racial segregation

Upload: aparnett93

Post on 04-Jun-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    1/26

    STORhe Impact of ffirmative ction on EmploymentJonathan S. LeonardJournal of Labor Economics Vol. 2, No. 4. (Oct., 1984), pp. 439-463.

    Stable URL:http://links.jstor.org/sici?sici=0734-306X 28198410 292 3A4 3C439 3ATIOAA0 3E2.0.00 3B2-KJournal of Labor Economics is currently published by The University of Chicago Press.

    Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html . JSTOR's Terms and Conditions of Use provides, in part, that unless youhave obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, andyou may use content in the JSTOR archive only for your personal, non-commercial use.Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.j stor.org/journals/ucpress.html.Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen orprinted page of such transmission.

    JSTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive ofscholarly journals. For more information regarding JSTOR, please contact [email protected].

    http://www.j stor.org/Mon Nov 27 15:25:17 2006

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    2/26

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    3/26

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    4/26

    Impact of Affirmative Action 41most certain aspect of affirmative action is that it is controversial. Thisstudy brings new evidence to bear on that controversyevidence that canenlighten discussions both of discrimination in the United States and ofthe impact of government regulation.

    Past StudiesThe literature on affirmative action can be divided into studies of the

    regulatory process that find it mortally flawed and studies of impact thatfind it successful. The process studies by the U.S. Commission on CivilRights (USCCR), the General Accounting Office (GAO), and the Houseand Senate Committees on Labor and Public Welfare all conclude thataffirmative action has been ineffective and blame weak enforcement anda reluctance to apply sanctions. That this is not merely politics can bejudged from the fact that the Department of Labor has been sued withsome measure of success more than once for failure to enforce affirmativeaction properly. Debarment, the ultimate sanction, has been used only26 times; debarment of the first nonconstruction contractor did not occuruntil 1974. The GAO and USCCR have found that other forms of reg-ulatory pressure, such as pre-award reviews, delay of contract award, andwithholding of progress payments, have not been forcefully and consis-tently pursued. However, as evidenced by the increased incidence of de-barment and back-pay awards, enforcement did become more aggressiveafter 1973.

    In the light of the unanimity of these process studies in finding theaffirmative action regulatory mechanism seriously deficient, it is surprisingthat the few econometric studies of the impact of affirmative action in itsfirst years (Burman 1973; Ashenfelter and Heckman 1976; Goldstein andSmith 1976; Heckman and Wo1pin 1976), all based on a comparison ofEEO-1 forms by contractor status, have generally found significant evidencethat it has been effective for black males. These few studies of the initialyears of affirmative action (196673) are not directly comparable becauseof different specifications, samples, and periods. They do find, nevertheless,that despite weak enforcement in its early years, and despite the ineffec-tiveness of compliance reviews, affirmative action has been effective inincreasing black male employment share in the contractor sector, butgenerally ineffective for other protected groups. Of the four studies, Gold-stein and Smith (1976) find the weakest effects. Their results indicate a.0004 yearly increase in black males share of total employment and adecrease in the ratio of black to white males among nonreviewed con-tractors between 1970 and 1972. Heckman and Wolpin s (1976) resultsindicate an effect an order of magnitude greater, a .007 annual increase

    2 See, e.g., the case of Legal Aid Society of Alameda County v. Brennan, 608F.2d 1319 (9th Cir. 1979), cert. denied 10 0 S. Ct. 301 0 198 0).

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    5/26

    442 eonardin black male s share of total employment between 1972 and 1973 com-paring contractors and noncontractors. For comparison, Burman (1973)reports roughly a .003 annual increase in black males share of male em-ployment in the late sixties, and Ashenfelter and Heckman s (1976) resultsindicate a .0086 yearly increase in the ratio of black to white males between1966 and 1970. The present study will use a new set of highly disaggregatedpanel data to examine a more mature affirmative action program between1974 and 1980, which includes the beginning of substantial enforcementof regulations barring sex discrimination, the start of aggressive enforcementin the early seventies, and the period after the major reorganization ofthe contract compliance agencies into the OFCCP in 1978.

    II. A Tax Model of Affirmative ActionThis section develops a two-sector general equilibrium tax model that

    underlies most of the empirical work analyzing the effect of affirmativeaction. I model affirmative action as a tax on the employment of whitemales in the contractor sector. If these workers are immobile, they bearthe tax burden and relative white male wages fall.

    Assume the owner of the firm maximizes utility:max U = {T[F(m)] T(W i)m T(W )(1 m)

    t(m .iz) d(1 m)} (1)where

    T = total employment,m = proportion of white males in T,

    = average proportion of white males employed in a given industryand geographic area,

    Wm = wage of white males,WF = wage of other workers,

    = tax on proportion male employment,d = taste for discrimination against females and nonwhites,

    F(.) = a production function with F' > 0, F < 0.Now abstracting from the scale effect by fixing T = 1, the first-order

    condition isF' = W WF t d 2)

    from which we find:= g(W , WF , t, d). 3

    Taxes in this model of affirmative action are symmetrical with tastes

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    6/26

    Impact of Affirmative Action 43for discrimination in Becker's model. Intuitively, an increase in theaffirmative action tax shifts the demand curve for white male labordownward.I assume fixed tastes for discrimination and fixed technology or, lessrestrictively, technological change that is neither male nor female saving.Under these conditions, the change in demand is a function only of wagesand the tax. The contractor firms that are liable for the tax are distributedthroughout the economy, so all firms are assumed to be wage takers inthe same labor market. I also assume that the wage elasticity of labordemand is the same in the contractor and noncontractor sectors. Thedifference between the change in the employment of males at contractorfirms, Arn , and at noncontractor firms, AmNc , is then simply a functionof affirmative action pressure.

    m A mNC g t). 4)This is the central equation to be tested. I will compare shifts in the

    proportional employment of members of protected groups across con-tractor and noncontractor establishments across time. The hypothesis isthat if affirmative action has been ineffective, these employment shiftswill be the same for contractors and noncontractors. An effective affirmativeaction program is expected to shift the demand curve for blacks in thecontractor sector to the right, driving black wages up, increasing blackemployment in the contractor sector, and decreasing it in the noncontractorsector. Tests of the employment effects are presented next, and the wageeffects are discussed in Section IV. One might also expect affirmativeaction to lead to occupational upgrading, a possibility analyzed in otherwork (Leonard 19846). In reality, there are other policies, such as TitleVII, promoting the employment of blacks in the noncontractor sector,so I will measure only the differential impact of affirmative action overand above the effects of general policies or changes in tastes.'While this model has straightforward implications for changes in theemployment of males and females, one cannot fully explain changes in

    3 Throughout this study, one of the chief concerns is what the labor market forminorities and females would have been in the absence of Executive Order no.11246. A distinct question is, What would happen in the future if affirmativeaction were abolished? The tax models of regulation just presented are suitablefor the first question but prejudge the answer to the second. In the absence of theregulatory tax, the model assumes the demand for minorities and females willresume its former level. A more complex model along the lines of Aigner andCain (1977) is needed that allows for learning and changes in discriminatorybehavior. After all, one of the goals of affirmative action is to break down prejudice.If employers have falsely prejudged minorities and women to be less capable thanwhite males, a temporary affirmative action program might have permanent effectsby shocking them into correcting their mistake faster.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    7/26

    444 eonard

    relative wages without considering supply shifts. A finding that affirmativeaction has been effective in increasing female employment is consistentwith an unchanging ratio of female to male wages if female labor supplyhas increased at the same time. Since any such supply shift will affectcontractors and noncontractors alike, I isolate the impact of affirmativeaction on labor demand by comparing changes in employment acrosscontractors and noncontractors.

    In the discussion above, I abstracted from scale effects, and as we shallsee, these are usually unimportant. However, scale effects may lead to astriking reversal of changes in relative employment, obscuring the impactof affirmative action. The analysis in this case is analogous to that in two-sector general equilibrium models of taxation or unionization. Considerthe case in which the affirmative action tax is levied on the employmentof males only in the contractor sector, which is male intensive. Absentany scale effect, this leads to the substitution of females for males in thecontractor sector. At the same time, though, this tax increases costs in thecontractor sector, depending on the elasticity of substitution, and leadsto a decline in the size of the contractor sector, depending on productdemand elasticities. As the contractor sector shrinks it becomes even moremale intensive, under the usual assumptions. If this scale effect is large,we may observe an increase in the ratio of males to females in both thecontractor and noncontractor sectors because affirmative action has beeneffective in taxing male employment. This is an important paradox toconsider in theory. However, the differences in observed scale effects arein general negligible, so we may draw inferences concerning the effect ofaffirmative action by comparing what are in practice substitution effects.

    III The Impact of Affirmative Action on the Employmentof Minorities and Females

    The male share of employment has fallen steadily since 1960 as femaleshave flooded into the labor force (see table 1). In 1974, .384 of the employedwere female. By 1980 this had increased by 9.4 to .420. Over the sameperiod, the employment share of black males fell. While the proportionof nonwhite males in total employment remained stable over this sameperiod at .062, their proportion among males rose by 6 , from .101 to

    The contractor and review effects here are unlikely to be due in fact to anomitted Title VII variable. Across state by two-digit SIC industry cells, the cor-relation of Title VII class action cases decided in the federal district courts andcontractor employment is only .19. Of course, this is not to detract from theimportant role played by Title VII in the development and enforcement of affir-mative action. The direct effect of Title VII is the subject of other work.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    8/26

    Impact of Affirmative Action 45

    Table 1Changes in Black, Hispanic, and White Employmentby Sex in the CPS, 1974-80

    DemographicGroupEmployment Share

    Change PercentageChange974 19801 Black males .035 0.31 .004 11.42 Hispanic males .027 .031 .004 14.83 White males .554 .518 .036 6.54 Black females .036 .037 .001 2.85 Hispanic females .015 .020 .005 33.06 White females .333 .363 .030 9.07 Total employment 79 347 91,593 12 246 15.4

    NOTI .Employment data U.S. Bureau of Labor Statistics (1982), pp. 590-99, tables B-8, and B-9.Data refer to those at least 20 years old. In this table only, black includes Asians and other non-Hispanicminorities. Lines 1-6 refer to shares of total employment. Line 7 refers to the level of total employment.

    .107, due to Hispanic growth. On their face, these growth rates in rep-resentation are not strikingly higher after 1970 than before. At the sametime, both females and nonwhite males share of unemployment has beengrowing, along with their employment shares. The period between 1974and 1980 witnessed growth in females share of employment, and in non-white males share of male employment. What part has affirmative actionplayed in these increases?

    We have seen that affirmative action may usefully be thought of as atax on the employment of white males in the contractor sector, a tax thatshifts the demand for white male labor downward. If affirmative actionis effective, then the rate of change of protected groups employment sharewill be higher in contractor establishments than in noncontractor estab-lishments, ceteris paribus. Since affirmative action goals are similar withinindustry within region, the variance of employment share may be expectedto fall more and remain lower at contractor firms, when industry andregion are controlled for. In the long run, the levels of the employmentshares of protected groups will be higher in contractor firms, controllingfor industry and region, which determine skill requirements and locallabor supply and are not controlled for in the following cross-tabulations.The crucial tests are those on changes in levels of employment shares,which difference out unchanging variables. Since stocks are susceptibleto policy only through changes in flows, I expect the flows, or in otherwords the change in stocks, to be a more sensitive indicator of the impactof policy. It should also be noted that 1974 is an early year in the historyof affirmative action, especially for females. Although affirmative actionbecame effective in 1965, the provisions pertaining to females were a lateraddition, first enforced about 1974. For both nonwhites and females, theadjustment process was by no means over in 1974.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    9/26

    446 eonard

    Comparison of Mean Changes in Employment Shares in theContractor and Noncontractor Sectors

    Table 2 shows that between 1974 and 1980 black and female employmentshares increased significantly faster in contractor establishments than innoncontractor establishments. The appendix discusses the data on whichthe empirical tests discussed here are based. The t-tests in table 2 rejectthe equality of changes in means in all cases except that of nonblackminorities, which are the smallest groups. There is no striking evidencein the changes in variances in representation over time, but the variancein the contractor sector is always significantly less than in the noncontractorsector. Contractor establishments start with proportionately more nonwhitemales but fewer females in 1974, which in itself casts doubt on the argumentthat female-intensive establishments are being selected as contractors. Themost compelling evidence of the impact of affirmative action in table 2is the significantly greater increases in female and black male employmentshares in contractor establishments.'

    The interpretation of tests of affirmative action would be less straight-forward were scale effects to differ greatly across sectors. However, wesee in table 2 that there is only a small difference in the growth rates ofcontractor and noncontractor establishments; both are growing at between2 and 3 per year, so differences in scale effects across sectors are likelyto be negligible. On net, in these establishments members of protectedgroups are in part being substituted for white males over time. This alsosuggests that contractor firms are not growing fat on government largessethat allows them to expand total employment to take on relatively un-productive minorities and females. This is consistent with evidence pre-sented in other work that the productivity of members of protected groupsrelative to that of white males did not fall as their relative employmentshare increased. (Leonard 1984a).

    Reviewed versus NonreviewedGiven that contractor establishments have increased their employmentof nonwhites and females more than noncontractor firms have, whatadministrative tools have been useful? Do compliance reviews matter? Onealternative is that the threat of a compliance review is sufficient to obtainthe desired behavior, so that reviewed establishments do not differ sig-nificantly from nonreviewed contractor establishments. This would beplausible if the threatened penalty were severe enough to outweigh the

    5 Table 2 shows that affirmative action has been effective at the average estab-lishment, not for the average worker. To draw inferences regarding wages, changesin employment share must be weighted by establishment size. This comparisonshows smaller differences between sectors, suggesting that affirmative action hasproduced greater changes at smaller establishments.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    10/26

    Impact of Affirmative ActionTable 2Proportion of all Employees

    447

    LineDemo-graphicGroup

    Con-tractorStatus1974 1980

    Means Mean%Aean o Mean i1 Black N .053 .10 .059 .10 .006 282 Males Y .058 .10 .067 .10 .008 333 6.0) 9.4) 6.5) 3.6)4 Other N .034 .10 .046 .10 .012 525 Minority Y .035 .08 .048 .09 .013 586 Males 1.6) 2.1) 1.2) 2.1)7 White N .448 .27 .413 .26 -.034 28 Males Y .584 .26 .533 .25 -.047 49 66.7) 66.5) 16.4) (2.0)

    1 0 Black N .047 .10 .059 . 1 1 .012 471 1 Females Y .030 .07 .045 .08 .015 771 2 24.0) 19.2) 5.7) 10.8)1 3 Other N .024 .08 .036 .08 .012 651 4 Minority Y .016 .05 .028 .06 .012 771 5 Females 14.8) 13.0) 1.1) 3.2)1 6 White N .394 .27 .400 .26 .006 1 71 7 Females Y .276 .23 .288 .23 .012 301 8 59.7) 57.8) 7.8) 11.9)1 9 Total N 186 286 209 341 23 1 720 Y 271 728 276 720 5 2121 21.2) 16.2) 10.7) 3.3)

    NOTE.-T-tests across means in parentheses, on every third line. In every case, F-tests reject equalityof variances across contractors and noncontractors, with more than 99% confidence. The last col. is themean of percentage changes, not the percentage of change in means. N noncontractor in 1974 (27,432establishments); Y = contractor in 1974 (41,258 establishments).

    small probability of being caught. In fact, both the probability of reviewand the penalties imposed are not great. The OFCCP has records of atleast 27,000 reviews completed between 1973 and 1981 at 11,000 differentidentifiable establishments. In 1980, roughly 115,000 establishments weregovernment contractors, so at least 10% of all contractor establishmentshad been reviewed between 1973 and 1981. Twenty-six of these reviewsresulted in the ultimate penalty of debarment. Between 1969 and 1976,compliance reviews produced 331 conciliation agreements, according toa Department of Labor memorandum (Ferrand 1978). These agreementsawarded 61,279,000 in back pay, or 185,133 per company in a heavilyskewed distribution. Short of debarment or back-pay awards, affirmativeaction extracts only promises, although, as we have seen, these promisesare not empty.

    Compliance reviews have been targeted at large establishments thatalready employ proportionately more nonwhites. The economics of tar-geting enforcement and detailed empirical tests of actual enforcement

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    11/26

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    12/26

    Impact of Affirmative Action 49lations of a positive relationship between contractor status and growthrates of female and minority employment share might be spurious; it mightbe picking up the correlation between protected group share and theomitted establishment size. Similar arguments of a more tenuous naturecan be made about industry and region.

    lnIn this section I present the results of regressions of the formNt a f 3 CONTRACT 13 REVIEWN,4-1

    N,P3 14t , + XX + c, 5)I t

    whereN,,,= employment of workers of demographic group i in year

    t;N, = total employment in year t;

    CONTRACT = one if the establishment is part of a federal contractorcompany in 1974;

    REVIEW = one if the establishment experienced a compliance reviewin the years 1975 to 1979;

    X = a vector of variables including controls for establishmentsize, industry, region, occupational and corporate struc-ture.

    These are regressions across a longitudinal sample of establishments ofemployment growth for a particular group on contractor and review status,total employment growth, size, corporate structure, percentage nonclericalwhite collar, industry, and region. To reduce heteroscedasticity, all regres-sions are weighted by 1974 establishment size. The dependent variable isthe natural logarithm of the growth rate of employment of the givendemographic group/ If the employment of all demographic groups growsat the same rate, then f3 = 1. If affirmative action has not changed em-ployment growth rates differentially among contractors or the reviewed,then f3 = 0 and 132 = 0, respectively. The characteristics of the contractorand noncontractor subsamples are presented in table 3. The contractorestablishments are on average bigger, faster growing, more likely to bepart of a multiplant company, and slightly less nonclerical, white-collarintensive. In terms of regional distribution, they are relatively more likelyto be in the West, and less likely to be in the South. They are also much

    To keep establishments with no employees of a given demographic group inthe sample, 0.1 has been added to all employment subtotals.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    13/26

    450 eonard

    Table 3Characteristics of the Contractor and Noncontractor SubsamplesVariable Description

    MeanContractor Noncontractor

    Total number of employees, 1974 270.8 186.5Rate of growth of total employment,1974-80 .212 .174Proportion of establishments in single-establishment companies .078 .341Proportion of all employees who areofficials, managers, professionals,technicians, and salespersons .373 .391Industry SIC nos.):1-9 Agriculture, forestry and fishing .003 .00610-14 Mining .019 .00715-17 Construction .020 .01020 Food and kindred .048 .02921 Tobacco .002 .00022 Textiles .014 .02423 Apparel .009 .03724 Lumber and wood .013 .01225 Furniture and fixtures .004 .01526 Paper .023 .01027 Printing and publishing .012 .02428 Chemicals .031 .00829 Petroleum and coal .006 .00130 Rubber and plastics .014 .00931 Leather .003 .00732 Stone, clay, and glass .018 .01133 Primary metal .020 .01134 Fabricated metal .032 .02335 Machinery except electrical .036 .01836 Electrical and electronic machinery .031 .01037 Transportation equipment .021 .00838 Instruments .012 .00439 Miscellaneous manufacturing .005 .00740-47 Transportation .058 .04548-49 Utilities .056 .02550-51 Wholesale trade .091 .04252-59 Retail trade .206 .34060-69 Finance, insurance, real estate .096 .07070-89 Services .096 .186Region U.S. census geographic division):New England .056 .059Middle Atlantic and East North Central .354 .358West North Central .079 .080South .320 .353West .190 .149Sample size 41 258 27 432

    less likely than noncontractors to be in the service or retail trade sectors,and relatively more likely to be in manufacturing, utilities, construction,and mining. The interpretation of contractor effects in this section assumesthat both contractor and noncontractor establishments are wage takersin the same labor markets, and that both sectors have the same wageelasticity of demand.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    14/26

    Impact of Affirmative Action 51Table 4 (odd cols.) presents the primary regression results and shows

    a consistent pattern across demographic groups of effective affirmativeaction. Over a 6 year period the employment of members of protectedgroups grew significantly faster in contractor than in noncontractor es-tablishments. The growth rate is 3.8 faster for black males, 7.9 forother minority males, 2.8 for white females, and 12.3 for black females.A summary measure, white male employment, grew 1.2 slower in thecontractor sector. All of these effects are highly significant at the 99confidence level or better, and the effects for blacks and for white malesare robust across a number of specifications.'

    The demand shift for black males relative to white males estimated herefor contractor status is similar to that previously estimated by Ashenfeiterand Heckman (1976) and by Heckman and Wolpin (1976). The growthrate of black male employment over 6 years in the contractor sector is3.8 greater than among noncontractors. Taking the sixth root yields anannual growth rate that is 0.62 greater in the contractor sector. Forwhite males, the annual growth rate is 0.2 slower among contractors,so contract status appears to shift the demand for black males relative towhite males by 0.82 per year. For comparison, using a different spec-ification in a sample of integrated establishments for the earlier period1966-70, Ashenfelter and Heckman report an annual shift correspondingto 0.86 per year.

    Compliance reviews have played a significant role over and above thatof contractor status, advancing black males by 7.9 , other minority malesby 15.2 , and black females by 6.1 among reviewed establishments.Compliance reviews have retarded the employment growth of whites.The effect is significantly negative in the case of white females but smalland insignificant in the case of white maleswhom one would haveexpected to bear the brunt of the adjustment. The anomalous result forwhite females is sensitive to specification. It is also difficult to reconcilewith the positive impact of contractor status on white females, but maybe influenced by a review process that asks for more than last year, ratherthan more than average, in a time of sharply increasing female laborsupply. For black and other minority males, the impact of undergoing acompliance review is roughly twice that of being a contractor. With theexception of white females, compliance reviews have an additional positiveimpact on protected group employment beyond the contractor effect.Direct pressure does make a difference.

    The estimate in equation (3) is that the growth rate of black males'employment increased 3.8 more in contractor establishments, not count-ing the direct effect of reviews: 6.8 of all contractor establishments,

    Other regressions show som e evidence that affirmative action for black maleshas been m ore effective at m ale intensive establishmen ts.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    15/26

    O

    Q.OO

    o0

    C J

    0

    O

    L.

    OUOU

    CL

    >

    -

    I - 1 4JCI X

    kFme

    WheFme

    OhMe

    74

    7 ,2

    452 eonard

    . .

    u, u, N. r M.00 q7 W1 en .1% f . ) 0 CN . NIO - Un SC 0 ry D r. O N N O 0W I , D0 0 WI 0 0 CD 0 0 C) 0 C) 0 CD 0 0 01 0 0 01MM - f.0007D nrAr so C r CD 0

    C N00 0 OND O rn 7 I. , .cr N A ,7. 000 00N )O 0, ,t , c, c, c, ,0 CD CD Cr. 0 CD 0 C)

    0 0 I D O 0 0 , C ) u 1 u n rgr,NI O rn D I. O rg - 0O O0 0O0O C) O0OI 2 I : I 1.

    N.000 ulNy . , C ) f , ) U- 0 0 f rA .0 0,u, CN u1NNO, ,1 4 0 r - )0 MCF 00). . - : . CN ,N 00 N , 0 NC NC C 00 WI O I - ) 0 )

    5C) 00 Cr, 00 n 00 O n rn M N . 04 .0 rA 00 N. n 0 0f 00r, r.00--00 A 04 NOOu, y 0, u,O 0 0 CD CD 0 O 0 OrI '-

    .00 - 0U . s f C I M 1 . .- NI N J 0 0 0 - - 00 00 0 rI 0 0 0 0 0 0 0-:- ry . 0, ry N 7 .0 r. .0 rn rn 00 Cs rnrn - .0 NI CO O O O ry ID 0 0 0 0 0 7 CD 0 0 CD 0 O CN O 0 O 0 O 0 O C) O 0 O 0CD 0 O r.

    U n....-.N 0-N n D t WD .--. . . - - , - . . .0 0 , t - NO P. --.0 0 0 r,00 0 NI 0f. 0 0 0 0 0 CN00 0 C) 0 0 0I. - - : . . . 1 : .U

    0. ) '1 L , U DU ,

    gum.O xi--O < 94 6 g 7v 1.-

    z47 s E 6 .r. g xO C CE o

    1c5 7 L cg .0-Ex

    uEIr N4.50O0 oUt ' L E - 91-o 'C r.

    g oh r .O

    0

    2

    C:6gt5gbu 0 0=6 O oE,rkl 9 a .

    O 1 1C . ' - : 4 O D

    E 2.5, ced6' 5 -F 4 - 2 -E E c

    . E Ac,4 5,

    E,

    0C ?, L L ,o 4 E a

    0-0 .0 `

    N

    c z , h:

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    16/26

    Impact of Affirmative Action 53accounting for 17.4 of all contractor employment, were reviewed insubsequent years. In these establishments the black male growth rate wasan additional 7.9 faster than in nonreviewed contractors, so 12 fasterthan noncontractors. The total impact of affirmative action among con-tractors is then the weighted average of the annual 0.62 shift amongnonreviewed contractors and the 1.91 shift among reviewed contractors,or 0.84 per year. The demand shifts for other minority males, whitefemales, and black females are 1.69 , 0.37 , and 2.13 , respectively. AsSection IV will show, these are not small demand shifts. The shift is largestfor black females, although the ranking of these effects is sensitive tospecification. It has been commonly assumed that employers can kill twobirds with one stone by hiring black females. The evidence here is thatblack females not only have gained relatively more under affirmative actionthan white females, but have also gained more than black males, as onemight have expected if they were relatively favored.

    Employment opportunities depend critically on growth. Table 4 alsoindicates that minorities and females experienced significantly greater in-creases in representation in establishments that were growing and so hadmany job openings. The elasticity of white male employment growth withrespect to total employment growth is .976, significantly less than one.This indicates that members of protected groups dominate the net incomingflows in both contractor and noncontractor establishments. The respectiveelasticities for black males, other males, white females, and black femalesare 1.22, 1.09, 1.02, and 1.19, all significantly greater than one. Particularlyin the case of blacks, of whom the quantity supplied has not greatlyincreased, this suggests the importance of Title VII, which applies to allestablishments in the sample, in expanding employment opportunities.Establishments that are not part of multiplant corporations have signif-icantly lower growth rates of employment of members of protected groups.Corporate size is probably of greater consequence than establishment size,with larger corporations showing greater increases in minority and femaleemployment. Establishment size itself has insignificant effects on whiteand black males, but other males and black females grow significantlyfaster at larger establishments, while white females grow significantlyslower. It is also important to note that the tests here also control for theskill requirements of each establishment. Establishments that are nonclericalwhite-collar intensive exhibit faster employment growth for both maleand female blacks and significantly slower growth for white males.

    The efficacy of affirmative action depends critically on employmentgrowth. The even-numbered equations in table 4 include interactions ofcontractor and review status with establishment size and growth rate. In

    9 Other tests show evidence of a trade o ff between the em ployment growth offemales and m inority m ales.

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    17/26

    454 eonard

    every case, being a contractor or undergoing a compliance review havesignificantly greater effects if the establishment is growing. The evidencewith respect to interactions of affirmative action with establishment sizeis mixed. To illustrate, equation (4) indicates that while black male em-ployment grows 5.6 faster at contractor establishments than at non-contractor establishments with stable employment, it grows 6.7 fasterat the mean total employment growth rate of 5.1 , and 7.4 faster iftotal employment grows by 15 . Affirmative action has been far moresuccessful at establishments that are growing and have room to accom-modate federal pressure.

    There is significant variation in the growth of minority and femaleemployment across industries and regions. To determine the within-in-dustry, within-region impact of affirmative action, all of the equations intable 4 include 27 industry and four region dummy variables. The omittedgroups were trade and New England. Most of these variables had significantand strong effects, although these vary across specifications. White males'employment growth, a summary measure, is significantly at least 6higher in tobacco, furniture, paper, instruments, and utilities. It is at least6 slower in finance, insurance, and real estate. White females employmentgrowth is significantly slower in tobacco and textiles and remarkably fasterin mining, construction, lumber, petroleum, primary metals, and otherindustries where females have long been rare. Black males' growth issignificantly at least 10 slower in agriculture, construction, primarymetals, and services, and at least 10 faster in tobacco, textiles, miscel-laneous manufacturing, utilities, finance, insurance, and real estate. Sinceregion, growth rate, and percentage nonclerical white collar are controlledfor, these appear to reflect real differences across sectors in the growthof minority and female employment. Across regions, black females increasetheir employment significantly faster in the South, although for blacksand whites F-tests do not reject the equality of coefficients on contractorand review status in the South and nationally. In contrast, F-tests doindicate a significantly less effective program for other minorities in theSouth than elsewhere.

    Members of protected groups have enjoyed improved employment op-portunities at contractor establishments, and compliance reviews appearto have been an effective tool in changing employment patterns. Theevidence here is that a process that has been frequently criticized as largelyan exercise in paper pushing has actually been of material importance inprompting companies to increase their employment of minorities andfemales.

    Changes in Contractor StatusThe impact of establishments that change contractor status between

    1974 and 1980 is explored in table 5. One would expect protected groups

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    18/26

    Impact of Affirmative Action 55

    Table 5The Effect of Changes in Contractor Status on Employment Growthby Demographic Group N = 68,690)WhiteMales BlackMales OtherMales WhiteFemales BlackFemales1) 2) 3) 4) 5)

    STAYC -.016 .042 .084 .038 .136.004) .013) .016) .007) .015)LEAVEC -.017 -.007 -.004 .011 -.0002.0060) .019) .023) .010) .022)ENTERC -.015 -.008 -.024 .022 -.020.005) .018) .0220) .010) .0210)REVIEW -.0008 .073 .142 -.033 .045.004) .015) .018) .8) .017)GRO WTH .976 1.225 1.090 1.017 1.194.0030) .9) .10) .005) .010)SIZE .0019 .0022 .050 -.0048 .021.0011) .0035) .004) .0019) .004)SINGLE .026 -.008 .065 -.073 -.072.004) .013) .0160) .007) .015)PWC -.066 .091 -.010 -.031 .0009.007) .022) .027) .012) .025)R .70 .25 .15 .45 .20

    Nom.-All equations include 7 industry and four region dichotomous variables. STAYC is a di-chotomous variable set to one if the establishment was part of federal contractor company in 1974 and1980. LEAVEC = contractor in 1974, not 1980. ENTERC = contractor in 1980, not 1974.

    to experience the fastest employment growth at establishments that re-mained contractors, followed in order by establishments that left contractorstatus, those that became contractors, and, finally, those that never werecontractors. This assumes no cohort effect and a lagged adjustment toaffirmative action over time because of inertia in employment stocks andpersonnel policies. For all protected groups, employment gains were sig-nificantly greater at establishments that were contractors in both 1974and 1980 than in other establishments. As one might expect, the impactof affirmative action is greater at establishments that remain under theaffirmative action obligation for longer periods of time. For most protectedgroups, establishments that pass through contractor status are insignifi-cantly different from those that remain noncontractors. On the otherhand, the growth rate of white males is significantly slower at establish-ments that were contractors at any time than at noncontractors. In relationto white males, then, members of protected groups have experienced em-ployment gains at establishments that were contractors at some time.While the status-change variables are usually individually significant, theydo not generally contribute to a significant reduction in the standard errorof the estimate.

    The statistical tests shown in this section give evidence of a contractcompliance program that works. Executive Order 11246 has led to sig-nificant employment gains for females and for blacks in contractor es-

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    19/26

    456 eonardtablishments, and compliance reviews have played an important role inthis process.

    Selection or Changed Behavior?It is not implausible to suppose, along the lines of Heckman and Wolpin

    1976), that establishments with a relative overabundance of white maleswould avoid being federal contractors. Some of the findings presentedabove might be qualified if there were evidence of such selection: if es-tablishments with high protected group employment were more likely tobe contractors. Because of the recursive nature of the system, this prop-osition is tested in table 6 in logit estimates of the probability of being acontractor in 1980 as a function of 1974 demographics, the change indemographics between 1974 and 1980, and establishment size, growthrate, corporate status, industry, and region. There is little evidence hereto support the proposition that establishments with a high or growinglevel of minority or female employment are more likely to be contractors.This leads one to speculate that perhaps the costs of affirmative actionare not great on average, or that they are balanced by lump-sum trans-fers from the government in a contracting process that does not turn onprice alone.

    The evidence in table 6 suggests that the establishments that were morelikely to be contractors in 1980 were actually those with the greatestproportion of white males and the least proportion of minorities and

    Table 6Logit Estimates of Simultaneity: The Effect of EstablishmentDemographics on Contractor Status in 1980 N = 68,690P) /5)x

    AsymptoticStandard ErrorProportion white male, 1974 .079 .344 .13Proportion black male, 1974 .032 .138 .15Proportion nonblack minority male,

    1974 .086 .376 .18Proportion white female, 1974 .205 .891 .13Proportion black female, 1974 .200 .871 .17A Proportion white male, 1980-74 .552 2.40 .18A Proportion black male, 1980-74 .543 2.36 .23A Proportion nonblack minority,1980-74 .727 3.16 .25A Proportion white female, 1980-74 .499 2.17 .17A Proportion black female, 1980-74 .511 2.22 .23SIZ .0011 .0047 .0043GROWTH .453 1.97 .021SINGL .00009 .00039 .00013Twenty-seven industry and fourregion dummies YesMean squared error .189Mean of dependent variable .641

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    20/26

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    21/26

    458 eonardpart of this improvement in relative black male earnings may be explainedby affirmative action?

    To frame the implications of the demand shifts found here for thechange in relative earnings of black males, consider the following simplemodel of the labor market, where all variables are in logarithmic form:

    -= TIW 6)Ns = b + W, 7)

    where

    ND = the logarithm of the demand for black male labor relative to whitemale labor,Ns = the logarithm of relative labor supply, andW = the logarithm of the ratio of black to white male wages.

    In equilibrium:b E

    The logarithmic derivative of relative wages with respect to a demandshift is then

    ws (9)

    In table 4, I estimated that the relative demand for black male to whitemale workers increased by 6.5 in the contractor sector between 1974and 1980. To derive an estimate of the effect of affirmative action onmarket demands, assume that affirmative action has not directly alteredlabor demand schedules outside the contractor sector, that the demandelasticities are equal in both sectors, and the supply curve identical, sothat the differential between employment shifts in the contractor and non-contractor sectors can be identified as a demand shift. The market shiftmay then be taken as the weighted average of sectoral shifts. Though68.6 of employment in the sample is in contractor establishments, thiswill tend to overestimate the market shift because many small employerswho are not contractors are not included in the sample. Assuming thatroughly half of all employment is in the contractor sector, the impliedrelative demand shift overall is 3.25 . For demand and supply elasticitiesthat sum to less than 1.4, all of the improvement in black relative earningsamong men may be explained by affirmative action. At least half of thereduction in racial inequality can be accounted for if the relevant elasticities

    8 )

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    22/26

    Impact of Affirmative Action 59sum to 2.8 or less. Since the actual percentage increases in the ratio ofmedian earnings or in the mean or median earnings of all workers, wereall less than 2.3 , affirmative action may well have played an even greaterrole. The same would be true the lower the elasticity of supply or demand.While it is not implausible to think of the elasticity of relative labordemand as being greater than one, in other work I estimate this elasticityof substitution of nonwhite for white male labor to be on the order of .7to 1.1 and to have fallen over time, perhaps in response to affirmativeaction and Title VII. While other factors on both the demand and thesupply sides of these markets have likely also played a role, the increasein the demand for black male labor relative to white induced by affirmativeaction can help account for a significant part of the increase in the relativeearnings of black males. 2

    V. ConclusionThe tests presented here suggest that while generating tremendous public

    criticism and resistance and while undergoing frequent regulatory reor-ganization, affirmative action has actually been successful in promotingthe employment of minorities and females, though less so in the case ofwhite females. In the contractor sector affirmative action has increasedthe demand relative to white males for black males by 6.5 , for nonblackminority males by 11.9 , and for white females by 3.5 . Among females,it has increased the demand for blacks relative to whites by 11.0 . Fora program lacking public consensus and vigorous enforcement, this is asurprisingly strong showing. While the gains of white females are smallerthan those of blacks, it is important to keep in mind that the employmentof females and minorities has been increasing in both sectors. Indeed, ifthe OFCCP pressured establishments to hire more females and minoritiesrelative their own past records rather than to industry and region averages,the observed pattern is just what we would expect to see during a periodwhen female labor supply had been growing. Females' share would increaseat all establishments because of the supply shift, and contractor estab-lishments would be under little pressure to employ more females thannoncontractors. The relatively short history of affirmative action for fe-males, as well as the demographic composition of the bureaucracies thatenforce affirmative action, may also help explain the differential impactof affirmative action across protected groups.

    This paper has presented significant large-sample evidence with detailedcontrols at the establishment level that minority and female employmenthas increased faster at contractor establishments that bear the affirmativeaction obligation. It has also shown that compliance reviews have notbeen well targeted against discrimination but have been an effective en-

    12 For a discussion of these issues see Brown 1984 ).

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    23/26

    46 eonard

    forcement tool. In sum, despite vigorous contention and weak enforcement,affirmative action appears to have played a major role in improving theeconomic position of minorities and females.AppendixData

    Two rich, detailed, and disaggregated data sets are used in the empiricaltests: establishment-level EEO-1 reports on more than 16 million employeesfor 1974 and 1980, and establishment level affirmative action compliancereview reports for the period 1973-81, made available by the Office ofthe Assistant Secretary for Policy Evaluation of the U.S. Department ofLabor and by the OFCCP's Division of Program Analysis.

    Title VII of the Civil Rights Act of 1964 requires annual reports onwork force demographics from all private employers with 100 or moreemployees, or 50 or more employers and a federal contract or first-tiersubcontract worth $50,000 or more. In the case of multiplant employers,all establishments with more than 24 employees that belong to firmsfulfilling the above conditions must report individually. In 1978, 39,000employers with more than 165,000 establishments filed reports covering36 million employees, more than half of all private nonfarm employees.The EEO sample is extensive, covering three-quarters of all manufacturingemployment as reported by the Bureau of Labor Statistics. Employerswith small work force establishments, such as construction, trade, andagriculture, are underrepresented. Construction and agriculture are alsounderrepresented because temporary or casual employees are not countedas employees for the purposes of reporting requirements (U.S. Equal Em-ployment Opportunity Commission 1980, p. xi).

    Firms may have an incentive to discover that their white male employeesare really minority females when it comes to filling out their EEO-1 forms.For example, in 1978 the EEO sample, in which small establishments areunderrepresented, reported 18% nonwhite employment in manufacturing,while the CPS showed 12%. This need raise no problem for the currentstudy if such inflation is uniform across contractors and noncontractors,as is likely if liability to Title VII litigation is perceived as the primarythreat. Under Title VII pressure, moreover, firms may be more concernedwith inflating minority and female employment levels than growth rates.However, if this strategic reporting does extend subtly to growth rates,then to the extent that such behavior is greater among contractor estab-lishments, all studies that rely on comparisons of contractor and non-contractor EEO-1 forms, including the one at hand, will overestimate thetrue impact of affirmative action.From samples of roughly 160,000 establishments in 1980 and 100,000establishments in 1974, I found 68,690 establishments, with more than 16

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    24/26

    Impact of Affirmative Action 6million employees, that filed identifiable reports in both years on whichthe tests reported here are based.A company is considered a contractor if it or any of its establishmentsare prime government contractors or first-tier subcontractors with a con-tract, subcontract, or purchase order of 50,000 or more. This study iden-tifies any establishment that is part of a contractor company as a contractor,whether or not the individual establishment so identified itself. Note thatthe subcontractor clause vastly extends the compass of affirmative actionregulation. To the extent that I label as contractors some establishmentsthat are not aware that they are under the affirmative action obligation,I mix the sin of ignorance together with the greater sin of volition. Ifaffirmative action were found to be ineffective, it would be useful toseparate these two causes.Contractor status changers, particularly entrants, between 1974 and1980 are surprisingly common. Eleven percent of all 1974 contractorsestablishments were noncontractors in 1980, whereas 27% of all 1974noncontractors were i entifie as contractors in 1980, constituting 17%of all 1980 contractors. This suggests contractors have become betterlabeled over time. Whether these status changes are true, or just an artifactof more accurate reporting, my results will be biased against finding anyaffirmative action effect when I test according to 1974 status only. Inother words, I underestimate the effect of being a contractor because Iinclude among the noncontractors some establishments that became orreally were contractors, and I include among the contractors some estab-lishments that became or really were noncontractors.

    To compare demographic changes across reviewed and nonreviewedestablishments, I merged the matched 1974 and 1980 EEO-1 establishmentdemographic data with data on OFCCP compliance reviews. The OFCCP'sadministrative records contain data on 27,000 compliance reviews at 11,000identifiable establishments between 1973 and 1981. Before 1978, I havedata almost exclusively on reviews conducted by the Department of De-fense, but these accounted for nearly half of all compliance reviews. Reviewscompleted prior to 1973 or after 1979 are underrepresented, and becauseof general underreporting, some establishments that were reviewed willbe included among the nonreviewed, biasing my tests against finding animpact of compliance reviews. I labeled as reviewed any establishmentsthat had a record of at least one compliance review between 1975 and1979 inclusive. Multiple reviews are not rare, but are not controlled forin my tests. Since I expect decreasing returns to multiple reviews, thiswill bias against finding any review effect in the case of establishmentsreviewed prior to 1974. In other cases I will simply be measuring thecumulative effect of reviews. Since the mode year of review completionin the sample is 1975, while demographic changes are measured between

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    25/26

    46 eonard

    1974 and 1980, there is little potential for underestimating review effectsdue to lags in response.ReferencesAigner, Dennis, and Cain, Glen. Statistical Theories of Discrimination

    in Labor Markets. Industrial and Labor Relations Review 30 (January1977): 175-87.Ashenfelter, Orley, and Heckman, James. Measuring the Effect of anAntidiscrimination Program. In Evaluating the Labor Market Effects ofSocial Programs, edited by Orley Ashenfelter and James Blum. Princeton,N.J.: Princeton University, Industrial Relations Section, 1976.

    Brown, Charles. Black/White Earnings Ratios since the Civil Rights Actof 1964: The Importance of Labor Market Dropouts. Quarterly Journalof Economics 99 (February 1984): 31-44.Burman, George. The Economics of Discrimination: The Impact of PublicPolicy. Ph.D. dissertation, University of Chicago, 1973.Ferrand, Louis G., Jr. Compilation of Lists of Back Pay AgreementsObtained by Compliance Agencies under Executive Order 11246. Af-fadavit of the Counsel for Civil Rights, U.S. Department of Labor, inthe matter of Warner and Swasey Company and Defense LogisticsAgency, September 14, 1978.

    Freeman, Richard B. Changes in the Labor Market for Black Americans,1948-1972. Brookings Papers on Economic Activity no. 1 (1973), pp.67-120.

    Time-Series Evidence on Black Economic Progress: Shifts inDemand or in Supply. Unpublished paper. Cambridge, Mass.: HarvardUniversity, Department of Economics, May 1978.Goldstein, Morris, and Smith, Robert S. The Estimated Impact of theAntidiscrimination Program Aimed at Federal Contractors. Industrialand Labor Relations Review 29 (July 1976): 523-43.

    Heckman, James J., and Butler, Richard. The Government's Impact onthe Labor Market Status of Black Americans: A Critical Review. InEqual Rights and Industrial Relations edited by Leonard Hausman etal. Madison, Wis.: Industrial Relations Research Association, 1977.Heckman, James J., and Wolpin, Kenneth I. Does the Contract Com-pliance Program Work? An Analysis of Chicago Data. Industrial andLabor Relations Review 29 (July 1976): 544-64.

    Leonard, Jonathan S. Affirmative Action as Earnings Redistribution: TheTargeting of Compliance Reviews. Unpublished manuscript. Berkeley:University of California, School of Business, 1983.

    Anti-Discrimination or Reverse Discrimination: The Impact ofTitle VII, Affirmative Action and Changing Demographics on Pro-ductivity. Journal of Human Resources 19 (Spring 1984): 145-74. a)Employment and Occupational Advance under Affirmative Ac-tion. Review of Economics and Statistics (1984), in press b)U.S. Bureau of Labor Statistics. Labor Force Statistics Derived from the

  • 8/13/2019 S. Leonard: Spatial and Racial Segregation

    26/26

    Impact of Affirmative Action 63

    Current Population Survey: A Databook Vol. 1. Bulletin 2096. Wash-ington, D.C.: Government Printing Office, September 1982.

    U.S. Bureau of the Census. Money Income in 1974 of Families andPersons in the U.S. Current Population Reports, ser. P-60, no. 101.Washington, D.C.: Government Printing Office, January 1976.

    Money Income in 1974 of Households, Families, and Personsin the U.S. Current Population Reports, ser. P-60, no. 132. Washington,D.C.: Government Printing Office, July 1982.

    U.S. Equal Employment Opportunity Commission. 1978 Report: Minoritiesand Women in Private Industry Washington, D.C.: Government PrintingOffice, February 1980.