Journal of Economic and Social Measurement 22 (1996) 43-55
IOS Press
43
Discrimination due to race and gender in the youth
labor market: is it a double jeopardy?
Deanna L. Sharpe*
Consumer and Family Economics Department, College of Human Environmental Sciences, University
of Missouri, USA
Mohatned Abdel-Ghany
Department of Consumer Sciences, College of Human Environmental Sciences, The University of
Alabama, USA
A wage decomposition method is used to examine components of average wage differentials in the
youth labor market, letting both race and gender vary. Issues of selection bias and computation of the
nondiscriminatory wage are addressed. Findings indicate that bias in the youth labor market focuses
more on gender than race. However, until researchers devise consistent methods of estimating the
source, amount, and direction of discrimination, policy makers face difficulty devising procedures to
correct discriminatory wage differences.
1. Introduction
Racial and gender bias in wages has long interested labor economists (see, for
example, [5, 7, 8, 10, 14, 16, 18, 22, 23]). Explanations for differing average
wages among whites and blacks, males and females have focused on both sides of
the labor market. According to the human capital approach, increased education
and training improves the quality of labor supply, resulting in higher wages. Thus,
attempts to equalize the quality of human capital endowments such as education,
on-the-job training, etc., should result in a narrowing of wage differentials among
different groups of individuals. Wage differentials which persist among those with
equal education and training suggests a differential demand exists in the labor
market for certain racial or gender groups, pointing to a need for public policy
initiatives to decrease the incidence of overt racial or gender discrimination.
Wage decomposition methods offer a means of examining the sources of wage
differentials and identifying the impact of both human capital endowments and
discrimination on average wages earned by different racial and gender groups.
Most previous studies of wage decomposition, however, have either focused on
'Correspondence to: Deanna L. Sharpe, Consumer and Family Economics Department, College
of Human Environmental Sciences, University of Missouri, Columbia, MO 65211, USA. E-mail;
cfedls@mizzou I .missouri.edu.
0747-9662/96/$8.00 © 1996 - IOS Press. All rights reserved
44
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
racial wage differentials, holding gender constant, or have examined gender wage
differentials within a given racial group (examples include [3, 4, 11, 17, 20]). Few
studies have allowed both race and gender to vary [7, 8].
This paper uses a wage decomposition method introduced by Cotton [8] to
examine the components of average wage differentials in the youth labor market,
letting both race and gender vary. Cotton's efforts, however, did not address the
influence of sample-selection bias on the size of these differentials and the impact
of varying the assumptions of the nondiscriminatory wage structure. Both of these
issues are examined in this study.
2. Theoretical model
The human capital model of wage determination developed by Becker [1,2] and
Mincer [19] provides the basis for the general theoretical model used in this study:
\nWi = a + bxi-\-ei.
(1)
In this model, the natural logarithm of the hourly wage rate of the i-th worker
(ln Wi) is regressed on a set of variables representing human capital endowments
and factors that influence labor demand (a;,). The portion of the log wage not
explained by either human capital endowments or labor demand factors is attributed
to discrimination. Thus, discrimination is captured and quantified in the error
term, ei, which is assumed to follow a normal distribution. Clearly, the size of
this component will depend on the quality of model specification and level of
measurement error [15].
Without discrimination, whites and blacks (or males and females) would be
perfect substitutes in production and any wage differential would be attributed to
skill differential alone [9]. Thus, nonzero differences in log wages between two
different groups of wage earners which remain after controlling for human capital
endowments and labor demand factors imply one group has a relative advantage
over the other in the labor market. Differences in log wages between two separate
groups of wage earners can be partitioned into three distinct components [9]:
K
K
j=0
The left-hand side of this equation is the difference in wages between the advantaged group and the disadvantaged group. This difference is composed of
a skill differential between males (or whites) and females (or blacks) valued in
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
45
absence of discrimination Z^^o^^'C^T ~ ^ j ) ' ^ favored group treatment advantage YLj=Q^3{^J
^ i ^
^)
~ ^j)'
3"'^ 3 disfavored group treatment disadvantage
^* represents the nondiscriminatory wage structure.
3. Empirical model
Four wage equations were estimated using data from the 1984 Youth Cohort of
the National Longitudinal Survey (NLS) of Labor Market Experience. There were
1527 white males, 1128 white females, 477 black males, and 351 black females.
All respondents in this study were between 19 and 26 years of age and employed
at least part time. Analysis focused on those in this group that were employed full
time.
White young men received the highest mean hourly wage (geometric mean)
at $6.37, while black young women received the lowest at $4.71. The mean
hourly wage for black young men and white young women was $5.46 and $5.18,
respectively.
The empirical model corresponding to equation (1) was:
\nW = a + bxi
( z = 1,2,3,...,31),
(3)
where:
Xi is educational attainment measured as years of schooling completed;
X2 is job tenure measured as experience with current employer in terms of
number of weeks employed;
x-j is knowledge of the world of work measures as respondent's score on a test
of knowledge of different occupations;
X4 is the size of the labor force in the respondent's area of residence (rural
versus urban);
0:5 is employment in the public sector;
xe is wage determination by collective bargaining;
xy is a dummy variable indicating student status;
xg... x\Q are dummy variables indicating region of residence;
x\i...xig
and X20 •ic.ii are dummy variables representing specific occupations and industries of employment, respectively.
Human capital endowment is measured by educational attainment, time on the job,
and knowledge of the world of work. Remaining independent variables measure
aspects of the structure of labor demand.
Using the same set of explanatory variables, equation (3) was estimated separately for white males, black males, white females, and black females. Chow tests
[6] confirmed wage structures were significantly different for these comparisons:
white male to black male, white male to white female, wbite male to black female.
46
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
Table 1
Mean values and regression coefficients for selected human capital and labor market
characteristics on hourly wage rates of white and black male youth
White
Variables
Educational attainment
Job tenure
Knowledge of world of work
Rural residence
Govemment employment
Collective bargaining
Student
Region:
Northeast
North Central
West
South
Occupation:
Managerial/professional
Sales
Craft
Technical
Operative (assemblers)
Operative (transp.)
Laborer (nonfarm)
Farm laborer
Service laborer
Clerical
Industry:
Wholesale trade
Public administration
Entertainment and
recreation serv.
Construction
Agriculture
Personal services
Professional and
related serv.
Business and repair
services
Financial, insurance.
real estate
Transp., communication.
utilities
Mining
Manufacturing
' Retail trade
Black
Mean
values
12.6950
224.3200
6.8513
0.1919
0.2502
0.1794
0.7924
Regression
coefficients
0.0359'"
-0.0015
0.0268'"
-0.2037"
-0.0305'
0.2807'"
0.0852'"
Mean
values
12.1890
197.0800
5.0943
0.1216
0.1719
0.3270
0.7107
Regression
coefficients
0.0634"
0.0008
0.0247"
-0.0903
-0.0118
0.2659"'
-0.0897'
0.1991
0.3124
0.1709
0.0109
-0.0548'
0.1392'"
0.1824
0.1656
0.0191
0.0552
-0.0149
0.3111'"
-
-
-
-
0.1310
0.8055
0.2528
0.0602
0.1035
0.0661
0.0982
0.0426
0.0891
-
0.1479'"
0.0693
0.0817'
0.3234*
0.0347
-0.0541
-0.0617
-0.1243
-0.0472
-
0.0818
0.0419
0.1845
0.0210
0.1195
0.0713
0.1614
0.0189
0.2159
-
0.1697'
-0.0547
0.0853
0.1535
0.0583
0.0616
-0.0436
0.1203
-0.1250
-
0.0419
0.0393
0.1184'
0.1264*
0.0273
0.0524
0.0148
0.0689
0.0164
0.1054
0.0426
0.0282
0.0670
0.2633***
-0.1243
-0.2055'"
0.0105
0.1111
0.0189
0.0461
0.4188**
0.0111
0.1203
0.0404
0.0596
0.0726
0.0881
0.0283
0.0838
0.1461***
0.1132
0.0116
0.0255
0.2087*"
0.0273
0.1812
0.0576
0.0295
0.3222
0.2434***
0.4153***
0.2106***
-
0.0587
0.0147
0.2977
-
-0.0210
0.2652
0.1509"
-
-
D.L. Sharpe and M. Abdel-Ghany /Discrimination due to race and gender
Table 1
(Continued)
White
Variables
\
Constant
B}
Sample size
Ln hourly wage
Mean hourly wage
(geometric mean)
Mean
values
-0.4902
I.O
0.2650' •*
1527
6.4572
$6.37
Black
Regression
coefficients
1.7716*
6.7480* •*
Mean
values
-0.5387
1.0
Regression
coefficients
0.1483
5.1940*"
0.2760* ••
477
6.3031
$5.46
* Significant at the .10 level.
**Significant at the .05 level.
***Significant at the .01 level.
Table 2
Mean values and regression coefficients for selected human capital and labor market
characteristics on hourly wage rates of white and black female youth
White
Variables
Black
Mean
values
13.1340
217.4700
6.7952
0.1551
0.2083
0.1268
0.8502
Regression
coefficients
0.0584***
-0.0017
0.0235***
-0.2645**
-0.0608**
0.0722*
-0.0399
Mean
values
12.94
186.94
5.3732
0.0912
0.2707
0.2222
0.8889
Region:
Northeast
North Central
West
South
0.1747
0.2748
0.1640
-
0.0851**
-0.0208
0.1439***
-
0.1282
0.1225
0.0798
-
0.0042
0.0135
0.1043
-
Occupation:
Managerial/professional
Sales
Craft
Technical
Operative (assemblers)
Operative (transp.)
Laborer (nonfarm)
Farm laborer
Service laborer
Clerical
0.2012
0.0860
0.0443
0.0602
0.0629
0.0035
0.0284
0.0080
0.1764
-
0.1128** .
0.0647
0.0460
0.1624**
0.0105
0.3368
-0.0002
0.1416
-0.0378
-
0.1396
0.0769
0.0456
0.0684
0.1083
0.0142
0.0313
0.0030
0.2023
-
0.1374*
-0.0532
0.0311
0.2626**
0.0282
0.2451
0.0839
-0.4910
-0.0728
-
Educational Attainment
Job tenure
Knowledge of world of work
Rural residence
Govemment employment
Collective bargaining
Student
Regression
coefficients
0.0348
0.0030
0.0361***
-0.0088
-0.0583
0.1510**
-0.0720
47
48
D.L. Sharpe and M. Abdel-Ghany /Discrimination due to race and gender
Table 2
(Continued)
White
Industry:
Wholesale trade
Public administration
Entertainment and
recreation serv.
Construction
Agriculture
Personal services
Professional and
related serv.
Business and repair
services
Financial, insurance.
real estate
Transp., communication.
utilities
Mining
Manufacturing
Retail trade
A
Constant
Sample size
Ln hourly wage
Mean hourly wage
(geometric mean)
Black
Mean
values
Regression
coefficients
Mean
values
0.0284
0.0505
0.1119
0.1757"
0.1709
0.0940
-0.1266
0.2613'
0.0168
0.0142
0.0133
0.0745
-0.0460
0.2168'
-0.4767
-0.6319***
0.0114
0.0030
0.0057
0.0883
0.1808
0.3861
0.2288
-0.2968"
Variables
Regression
coefficients
0.2598
0.0706
0.2792
0.1192
0.0665
0.1181*
0.0456
0.0654
0.0851
0.0798
0.0570
0.1487
0.0417
0.0071
0.2012
-
0.2100*
0.3677'
0.1018'
-
0.0598
0.0028
0.2308
-
0.3615"'
0.3129
0.1140
-
-0.5076
1.0
0.3315* »•
1128
6.2492
$5.18
1.7684
6.5649'"
-0.5654
1.0
-0.9511
4.3217"
0.3343*
351
6.1541
$4.71
'Significant at the .10 level.
"Significant at the .05 level.
***Significant at the .01 level.
black male to white female and black male to black female. Wage structures for
white females and black females were not significantly different. The mean values
of these variables and their regression coefficients for the young men and for the
young women are reported in Table 1 and Table 2, respectively.
Cotton [9] points out that variations in female employment patterns due to family
responsibilities suggest it is important to consider the influence of marital status on
labor market participation. Data limitations did not permit inclusion of this variable
in this study. However, the impact of this omission is, perhaps, not as great among
the age group included in this study as it would be among older workers.
For each wage structure compared, four different models were estimated. Two
models focused on the issue of selection bias, while the other two models fo-
D.L Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
49
cused on the impact of assumptions made regarding the nondiscriminatory wage
structure.
Adjustment for selection bias. Discrimination, along with government welfare
and transfer programs, can affect the number of youth securing full time employment [22]. Assuming labor supply schedules are upward sloping, employment of
a group facing discrimination would fall. Workers in this group who would accept
nondiscriminatory wages will not accept employment at the lower, discriminatory
offer wage, especially when government welfare and transfer programs can provide
access to at least a minimal level of consumption. It is hypothesized that neglecting to consider sample-selection bias would result in an underestimation of wage
differentials when an advantaged group is contrasted with a disadvantaged group.
To test tbis hypothesis, A was introduced as a regressor in the wage equations for
full time workers (equation (3)) to correct for selection bias. This correction was
calculated according to a procedure developed by Heckman [12, 13].
Impact of nondiscriminatory wage structure assumption. The nondiscriminatory
wage structure, identified as B* in equation (2), is not observed but may be estimated given one is willing to make certain assumptions about tbe nature of B*.
Cotton [9] assumes, given discrimination, B* is bounded above by the wage structure of the advantaged group (who have an overstated wage) and bounded below
by the wage structure of the disadvantaged group (who have an understated wage).
The nondiscriminatory wage structure is assumed to be close to the wage structure
prevailing for white males. He proposes B* is a linear function of the wage structure of the groups included in tbe analysis weighted by the proportions of each
group employed full time in the labor market:
B' = /wmB™ + /bmS*"" -f /wfB-f + /bfB''f.
(4)
In 1984, wbite male youth comprised 68.9% of the workers in this category, black
male youth 3.8%, white female youth 23.4%, and black female youth 1.8% [23].
Reimers [21] has argued that the nondiscriminatory wage should be assumed
to be midway between tbe highest and lowest wages received. In this study, the
nondiscriminatory wage falls midway between the mean hourly wages received by
white males ($6.37) and black females ($4.71). Rather than using labor market proportions, B* is weighted proportionate to the difference between the wage received
and the wage that would have been received in the absence of discrimination. The
resulting estimate of B* is:
B* = (.3952 X Bwm) + (.1714 x .Btm) + (.0381 x
(5)
-{-(.3952 X
)
50
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
Table 3
Decomposition of wage differentials of white and black male youth (percentage of wage differentials
in parentheses)
Model
Wage differential
Skill differential
1
0.1540
(100%)
2
0.1040
(67.5%)
White male
treatment advantage
0.1778
(115.5%)
Black male
treatment disadvantage
-0.1278
(-83.0%)
0.1183
(100%)
0.0485
(41.0%)
0.1770
(149.6%)
3
0.1540
(100%)
0.1104
(71.7%)
0.0726
(47.1%)
4
0.1183
(100%)
0.0590
(49.9%)
0.0742
(62.7%)
-0.1072
(-90.6%)
-0.0290
(-18.8%)
-0.0149
(-12.6%)
Models used in this study. For each gender and race comparison that was statistically significant according to the Chow test, four separate models were constructed.
All four models follow Cotton's general model outlined in equation (2). In the first
two models. Cotton's assumptions are utilized in the calculation of B*. The latter
two models, in contrast, use Reimers' assumption regarding the nondiscriminatory
wage to calculate B*. Model 1 and model 3 incorporate a correction for selection
bias, while model 2 and 4 exclude this correction.
4. Results
The decomposition results for the 91 cent average wage differential between
white men and black men are reported in Table 3. When selection bias is taken
into account, approximately two-thirds of the differential can be attributed to white
male skill advantage evaluated in absence of discrimination. Ignoring selection bias
drops the percentage of wage differential attributed to skill differences to almost
50% in model 4 and 41% in model 2. In dollar terms, between 37 cents and
65 cents of the 91 cent wage differential is explained by nondiscriminatory skill
differences. The remaining difference is attributed to discrimination.
Examination of the white male treatment advantage and black male treatment
disadvantage across all four models reveals a wage premium paid to black males.
In the absence of discrimination, the average wage for blacks would have been
approximately 1% to 12% lower than it actually was. Note this premium is much
more pronounced for models 1 and 2 than for models 3 and 4, suggesting wage
decomposition results are very sensitive to the assumptions underlying calculation
of B* for these two groups.
Table 4 presents the decomposition results for white males compared to white
females. For each of the four models, approximately one-forth of the $1.19 wage
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
51
Table 4
Decomposition of wage differentials of white male and female youth (percentage of wage differentials
in parentheses)
Model
Wage differential
Skill differential
1
0.2079
(100%)
2
0.0534
(25.7%)
White male
treatment advantage
0.1778
(85.5%)
White female
treatment disadvantage
-0.0233
(-11.2%)
0.2078
(100%)
0.0550
(26.5%)
0.1770
(85.2%)
-0.0242
(-11.7%)
3
0.2079
(100%)
0.0726
(34.9%)
0.0842
(40.5%)
4
0.2079
(100%)
0.0511
(24.6%)
0.0512
(24.6%)
0.0742
(35.7%)
0.0825
(39.7%)
Table 5
Decomposition of wage differentials of white male and black female youth (percentage of wage differentials in parentheses)
Model
Wage differential
Skill differential
1
0.3225
(100%)
2
0.1129
(35.0%)
White male
treatment advantage
0.1778
(55.1%)
Black female
treatment disadvantage
0.0318
(9.9%)
0.3029
(100%)
0.1275
(42.1%)
0.1770
(58.4%)
-0.0016
(-0.53%)
3
0.3225
(100%)
0.1342
(41.6%)
0.0726
(22.5%)
0.1157
(35.9%)
4
0.3030
(100%)
0.1322
(43.6%)
0.0742
(24.5%)
0.0966
(31.9%)
differential is explained by nondiscriminatory skill differences of wbite males. In
models 1 and 2, it would appear wbite females receive 2% bigher wages than tbey
would have received in absence of discrimination, a very slight wage premium
indeed. In these models, white male treatment advantage explains about $1.01 of
the $1.19 wage differential. Models 3 and 4, in contrast, suggest white women
receive almost 8% lower wages than they would have received in absence of
discrimination. Here, white male treatment advantage explains only about 42 cents
of the $1.19 differential.
The wage difference between white males and black females reported in Table 5
was the largest of all groups compared ($1.66). This comparison is one of two
possible cross race and cross gender comparisons. For model 1 and 2, skill differences and white treatment advantage explained over 90% of this wage difference.
Black female treatment disadvantage was substantially higher in models 3 and 4
at 36% and 32%, respectively. In dollar terms, black female wages were 53 to
52
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
Table 6
Decomposition of wage differentials of black male and white female youth (percentage of wage differentials in parentheses)
Model
Wage differential
Skill differential
1
0.0896
(100%)
2
0.0066
(7.4%)
Black male
treatment advantage
0.1072
(119.6%)
White female
treatment disadvantage
-0.0242
(-27.0%)
0.0539
(100%)
-0.0506
(-93.9%)
0.1278
(237.1%)
-0.0233
(-43.2%)
3
0.0895
(100%)
-0.0079
(-8.8%)
0.0149
(16.7%)
0.0825
(92.2%)
4
0.0539
(100%)
-0.0593
(-110.0%)
0.0290
(53.8%)
0.0842
(156.22%)
Table 7
Decomposition of wage differentials of black male and female youth (percentage of wage differentials
in parentheses)
Model
Wage differential
Skill differential
1
0.1685
(100%)
2
0.0089
(5.3%)
Black male
treatment advantage
0.1278
(75.9%)
White female
treatment disadvantage
0.0318
(18.9%)
0.1846
(100%)
0.0790
(42.8%)
0.1072
(58.1%)
(-0.87%)
3
0.1686
(100%)
0.0239
(14.2%)
0.0290
(17.2%)
0.1157
(68.6%)
4
0.1847
(100%)
0.0732
(39.6%)
0.0149
(8.1%)
0.0966
(52.3%)
-0.0016
60 cents lower than they would otherwise have been in absence of discrimination.
White male treatment advantage accounted for almost twice as much of the wage
differential in models 1 and 2 (55% and 58%) compared to models 3 and 4 (23%
and 25%). However, across all four models, skill differences accounted for close
to 40% of the wage differential.
The other cross race and cross gender comparison in this study, black males
and white females, had the smallest wage differential ($0.28) of all comparisons
made. Results of the wage decomposition are reported in Table 6. For models 2,
3, and 4, skill differences were negative. This result suggests gender discrimination predominates over racial discrimination. For, although mean human capital
endowments of black males generally lag behind white females, on average, black
males receive a higher wage than white females. Note, the magnitude of the negative skill differences are more pronounced for the two models (model 2 and 4)
which omitted the correction for selection bias. The size of the skill differential is
D.L. Sharpe and M. Abdel-Ghany /Discrimination due to race and gender
53
considerably smaller in model 3 where selection bias bas been taken into account.
In models 1 and 2, the black male treatment advantage was much larger than in
models 3 and 4, again indicating the impact of varying the assumptions underlying
calculation of B*. White females appear to receive a sizable wage premium in
models 1 and 2, getting 27% (model 1) to 43% (model 2) more in wages tban they
would have received in the absence of discrimination. This picture is, however,
markedly different in models 3 and 4 which show the wage differential between
black males and white females dominated by treatment disadvantage experienced
by white females.
The wage differential between black men and black women reported in Table 7
is next to the smallest of all the groups compared at 75 cents. Unlike other comparisons, no consistent pattern emerged from the four models. For this comparison
more so than any other, each of the models used to decompose the wage differential tell very different stories. Skill differential accounted for a low of 5% (model
1) to a high of 43% (model 2) of the wage difference. Model 2 suggests black
females have an almost negligible treatment advantage compared to black males
in the labor market. But, according to model 3, average wages received by black
females are approximately 69% less than they would have been in the absence of
discrimination. The advantage of being a black man compared to a black woman in
the labor market ranged from a high of approximately 75% of tbe wage differential
(model 1) to a low of around 8% of the wage differential (model 4).
5. Discussion
During the past several decades in America, desegregation laws have been put
into effect and equal opportunity regardless of race or gender has been stressed.
These social forces should have the effect of narrowing skill differentials and eliminating treatment advantages awarded to certain racial or gender groups, especially
for youth just beginning their participation in the labor force.
Results of this study indicate differences in average log wages persist even after
controlling for differences in human capital. Differences in average log wages
were decomposed for four separate racial and gender comparisons. For every
comparison except white male to black male, skill differentials accounted for less
than half of the difference in average log wage. This result suggests differential
treatment by race and gender persists in the labor market, accounting, in general,
for over half of the differences in average log wage.
Comparisons across race, holding gender constant, indicate wbites receive a
sizable treatment advantage, regardless of the model used to decompose tbe wage
differential. Comparisons across gender, holding race constant, reveals a male
treatment advantage for both wbite and black males. This advantage was over
twice as large for models 1 and 2 as for models 3 and 4. The size and direction
of female treatment disadvantage varied. White females received a wage premium
54
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
for models 1 and 2, black females received the same only for model 2 and only
for a negligible amount. Female treatment disadvantage was markedly greater
for models 3 and 4 than for models 1 and 2. Examination of results of cross
race and cross gender comparisons suggests gender discrimination exceeds racial
discrimination in the youth labor market.
Selection of a given model as a basis for the decomposition of wage differentials did made a difference. Across virtually all comparisons, correcting for
selection bias widens the discriminatory wage differential. This finding supports the assertion in previous research that neglecting to correct for selection
bias would lead to misleading wage comparisons [13, 22]. For some comparisons, the change was negligible. When comparing average wages of white males
with white females, for example, without correcting for selection bias, the white
female treatment disadvantage accounted for 39.7% of the wage differential in
model 4. When the selection bias correction is present in model 3, white female treatment disadvantage increased to 40.5%. In other comparisons, the change
was more pronounced. For instance, comparing average wages between black
men and black women, accounting for selection bias increased the black female treatment disadvantage from 52.3% to 68.6% (in model 4 and 3, respectively).
Changing assumptions underlying B* also influenced the final percentages of
the wage decomposition attributed to skill differential, treatment advantage, and
treatment disadvantage. When the assumption is made that the nondiscriminatory wage structure resembles the wage structure of white males, the treatment advantage is generally over twice as large as it is when the nondiscriminatory wage is assumed to be midway between the highest and lowest wages
received.
6. Conclusion
Two conclusions may be drawn from this study. First, the results indicate that
wage discrimination in the youth labor market is due more to gender differences
than to racial differences. No difference was found between the wage structures
for black and white females, indicating no racial discrimination exits between these
two groups. Comparing females to males, males received higher wages than they
would have in absence of discrimination, regardless of race. In fact, although white
females had better human capital endowments than black males, they received less
wages to compensate them for their work in the labor market. Single jeopardy
rather than double jeopardy is at work here.
Second, current methods of decomposing wage differentials yield inconsistent
results. Agreement is needed among researchers regarding calculation of the
nondiscriminatory wage rate {B*). In absence of such agreement, policy makers
face difficulty in devising procedures to correct discriminatory wage differences
as estimates of the source, amount, and direction of the discrimination can vary
widely.
D.L. Sharpe and M. Abdel-Ghany / Discrimination due to race and gender
55
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
G. Becker, Human Capital, Free Press, New York, 1964.
G. Becker, The Economics of Discrimination, University of Chicago Press, Chicago, 1971.
A.S. Blinder, Wage discrimination: reduced form and structural estimates, J. Hum. Re.-:. 8 (1973)
436-455.
F. Bloch and S.P Smith, Human capital and labor market employment, J. Hum. Res 12 (1977)
550-559.
C. Brown, Black-white earnings ratios since the civil rights act of 1964: the importance of labor
market dropouts. Quart. J. Econ. 99 (1984), 31-44.
G.C. Chow, Test of equality between sets of coefficients in two linear regressions, Econometrica
28 (1960), 591-606.
M. Corcoran and G. Duncan, Work history, labor force attachment, and earnings differences
between the races and sexes, / Hum. Res. 14 (1979), 3-20.
J. Cotton, Decomposing income, earnings, and wage differentials: a reformulation of method.
Sociological Methods and Research 14 (1985), 2.
J. Cotton, On the decomposition of wage differentials. Rev. Econ. Stat. 70 (1988), 236-243.
R. Freeman, Decline of labor market discrimination and economic analysis. Am. Econ Rev 63
(1973), 280-286.
J. Gwartney and J.E. Long, The relative earnings of blacks and other minorities, Ind. Labor Rel.
Rev 31 (1978), 336-346.
J. Heckman, Sample selection bias as a specification error, Econometrica 47 (1979), 153-163.
J. Heckman, Sample selection bias as a specification error with an application to the estimated
labor supply functions, in: Female Labor Supply, J.P. Smith, ed., Princeton University Press,
Princeton, 1980, pp. 206-248.
S. Hoffman, Black-white life cycle earnings differences and the vintage hypothesis: a longitudinal
analysis. Am. Econ. Rev. 69 (1979), 855-867.
F.L. Jones and J. Kelley, Decomposing differences between groups: a cautionary note on measuring discrimination. Sociological Methods and Research 12 (1984), 323-324.
G.D. Jud and J. Walker, Racial differences in returns to schooling and pact and wage discrimination by region, J. Hum. Res. 12 (1982), 521-534.
E. Lazear, The narrowing of black-white wage differentials is illusory. Am. Econ Rev. 69 (1979)
553-564.
J. Mincer, Schooling, Experience, and Earnings, Columbia University Press, New York, 1974.
R. Oaxaca, Male-female wage differentials in urban labor markets. Int. Ecan. Rev. 14 (1973)
693-709.
C. Reimers, Labor market discrimination against hispanic and black men. Rev. Econ. Stat. 65
(1983), 570-579.
D. Shapiro, Wage differentials among black, hispanic, and white young men, Ind. Labor Rel.
Rev. 37 (1984), 570-581.
J. Smith and F. Welch, Black-white male wage ratios: 1960-70, Amer. Econ. Rev. 67 (1977),
323-338.
U.S. Department of Labor, Employment and Earnings, Annual Averages 32 (1985), 192.