Professional Documents
Culture Documents
Capital Humano
Capital Humano
South
Author(s): Finis Welch
Source: Journal of Political Economy , Jun., 1967, Vol. 75, No. 3 (Jun., 1967), pp. 225-
240
Published by: The University of Chicago Press
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend
access to Journal of Political Economy
THE JOURNAL OF
POLITICAL ECONOMY
EDITED BY ROBERT A. MUNDELL
IN CO-OPERATION WITH THE OTHER MEMBERS OF THE DEPARTMENT OF ECONOMICS
OF THE UNIVERSITY Of CHICAGO
BOOK REVIEWS:
The Journal of ?olltlcal Ecosomy Is publIed bimonthly in Februar. rl. June, Augus and December by
The Univerity of Press, 57SO Ellis Avenue icag llinoa 6W7. S cdption rates for U.A. and po I ya
$10. 9.00; years, $27.50. Canada a PanAmercan Patal Union: 1 year, $10.30; 2 years, $20.00;
All other countries: 1Iyer,100;2years, $21.00; 3 r# yers 30.50. Si co, $2.50. S e $0 per yer0 Su
start with the Xtlsue publised after oder is rolved. Make remittance plyable toThe University of Chio Ei United
States curr rit eqivlt.
Claims, for missing aumber should be made within the month following the regular mont
erect to supply mixing numbers free only when lose have been sustained in transit and when
Business correspowleno. should be addressed to The University of Chag Pres, Chicago, Illinois 607.
Communcatons for the adltorsand m should e addressed to the Editorof Tnt Jo iMoCAL BMW
oy, 1126 East 5th Stret, Chicago Illinois 6WV7. Mnets ll be ssesedforpo t y co nav l form
but, if accepted, an original t rtte cop, on no ble paper, double spcd, gwit genrmas (3-Imc top), will he
required, and footnotes should double spaced ad placed consecutively at the end.
The articles bn thi journal ae minded in the SociaW &4iewsd H uiesWldes, New York, New York. Book
are Indexed in Beak Review lmex, Detroit, Michigan.
A vftid forp to quote fom this Journal should be addressed to The University of Chicago Pres
Microfilms of Journal are now available to regular subscribers only and may be obtained at the end of the volume
yea. Orders and inquires shd be addrsed to Universi , 313 North First Street, Ann Arbor, Michigan.
Noe to subecrihers I you change your address, please notify u and your local postmaster lmediately, giving W your
old and your new address. Alke'o fa kfor #A. ehaxge.
PRIXTZD IN 'U.S.A.
POLITICAL ECONOMY
Volume 75 JUNE 1967 Number 3
FINIS WELCH*
Southern Methodist University
225
The wage of a laborer in the ith group Thus, so long as r1 5-? r2,
will then be increase the average wage of each group
dY (Y (Y (recall that &2Y/&N&E > 0), and, there-
Wi = dN= -N + riE (3) fore, the individuals in different groups
are complementary. In fact, in this world
That is, a laborer's wage will be the price
of constant returns to scale, the addition
of one unit of physical labor plus the unit
of any given person to the labor force
price of education times the number of
will not affect the wage of those persons
units of education the individual pos-
having the same quantity of education
sesses. Thus without any form of dis-
and will raise the wage of all other
crimination, laborers' wages would vary
laborers. It is this complementarity
only according to the quantities of
which, it seems to me, helps to contradict
education they "own."
the segregation tendencies implicit in
If all firms have the same production
the Becker model. Of course, segregation
function, then linear homogeneity im-
can still result. Psychic income dis-
plies that firm size and the number of
counts may exceed gains, and/or external
firms are indeterminant, but marginal
inefficiencies may accompany integra-
productivities (factor prices) are de-
tion.
termined as though there were a single
Let us now assume a form of dis-
producer who equates marginal cost
crimination in which laborers of group 1
with product price. When considering a
become less efficient when they work in
market in which labor is segregated, two
the presence of laborers of group 2 and,
firms will be considered, one consisting
symmetrically, members of group 2 may
of N1 laborers and the other of N2. For
become less efficient in the presence of
an integrated market, only one firm is
members of group 1. This assumption of
considered. Assume that r1 and r2 rep-
factor externality is basic to the re-
resent the per capita quantities of educa-
mainder of the paper and should be ex-
tion of the respective groups. And per-
plored. Presumably, the efficiency of
sons who possess, respectively, ri and r2
integrated laborers could decline simply
units of education are defined as being
because they have disutility for inte-
representative of their group, that is,
gration, and, under integrated conditions,
their income will be the average income
some of a laborer's time may be wasted
of their group.
because of his disgruntlement. However,
Now, consider the average wage of
this is undoubtedly an oversimplifica-
each group when production is integrated
tion. The incentive to integrate is derived
and no external inefficiencies exist.
from the assumption that laborers with
if, l=ON
N+ ri unequal quantities of education are com-
OE
plementary in production. In turn, any
and
resultant complementarity must be de-
1FT2 = A)N+ r2 Y rived from an exchange between the
ON TiE'
laborers which results in specialization
Since Y is homogeneous of degree 1, of function. (Perhaps, the more educated
it follows that person assumes a supervisory role.) It is
dl T dIT-'2 2 y therefore possible that the full amount of
d=2 dONO
dN2 d~l ,-EaN~E'
2 the complementarities which would be
derived from combining two laborers of the full advantages of the comple-
the same group is not derived when the mentarities are not realized with integra-
laborers of the two groups are integrated. tion. Rather than redefine the production
An obvious example of this sort of function to allow for the reduction in
"friction" occurs when the two groups of complementarities, I have redefined the
laborers speak different languages. quantities of labor in terms of segregated
Another example would be a case in equivalent efficiency units.
which intergroup association results in The impact of the induced inefficiency
overspecialization of labor. Combining is expressed as though a portion of a
laborers of the two groups might result laborer's working time is lost when
in a precise division of tasks such that if working under integration. Thus for a
a given task were in the domain of firm employing laborers of the two
group 1 it must always be performed by a groups, the apparent quantities of physi-
member of that group even in circum- cal labor and education would exceed
stances in which a member of group 2 their efjective quantities because of effi-
could perform the task with less resource ciency losses, that is, N = N1 + N2 -
cost. This case would be particularly type 1 loss and E = E1 + E2- type 2
likely if members of one of the groups loss. The losses are assumed to be a
considered themselves superior to the function of association between laborers
other group and consequently believed of the two groups. However, the incen-
that certain tasks (perhaps menial) were tive to combine laborers must also de-
"beneath" them. Of course, this condi- rive from association to gain the benefits
tion may also exist when members of of complementarity between physical
different trade unions are associated labor and education. An entrepreneur is
because of the rigidly defined domain of therefore faced with a dilemma: associa-
activities for each trade. tion of the groups results in both
In each of these examples, association advantages and disadvantages. Presuma-
has an implicit cost, so that the increase bly, optimal association is determined
in production from combining laborers such that the marginal efficiency gains
of different groups is less than it would resulting from the complementarity be-
be if the combination consisted only of tween education and physical labor are
members of a particular group.6 That is, offset by equal marginal losses resulting
from inefficiencies of association between
8 J. K. Chadwick-Jones (1964) reports a case
study in which the effects were similar to those hy- the groups. For simplicity, I will assume
pothesized here. Chadwick-Jones tells of the recruit- that the optimal association is achieved
ment of unskilled laborers in Italy who joined skilled
by associating laborers of the two groups
British laborers in the steel industry. Since the
laborers were to perform complementary functions, in a one-to-one correspondence. That is,
the British labor union supported the recruitment if a laborer of group 1 is hired by a firm
policy. A discussion of the problems arising due to
in which laborers of group 2 are pre-
association between the laborers is provided al-
though there is no empirical support. "It was ob- ponderant, he will, at each point in time,
served, not only that the extent of verbal communi- be associated with one laborer of group 2.
cation possible with the Italian recruits was,
When he is isolated, the complementari-
naturally, very limited, but also that their cultural-
ly-derived attitudes and expectations were inap- ties are assumed to be lost, and when he
propriate" (p. 194). Although we may question the associates with more than one member of
use of the word "inappropriate," it is evident that
group 2, there may be additional losses
the anticipated complementarity was not fully
realized. without compensating gains. Assume
also that when a member of group 1 increase the amount of association be-
associates with a member of group 2, tween the groups. In this case a laborer's
wage is given by
proportion pi of the first laborer's effec-
tive time is lost and that the correspond-
dYaY O Ay N Oy aE
ing proportion for the second laborer is
idN aN aN7+ aE'yr; (4)
P2. In this case the effective quantities
of labor are given by assuming that N1 represents the minority
group
N= N1 + N2 -min {N1, N2} (P1 + P2),
ayva
and W1- f-l) a-N + (I 1-f2
E = riN1 + r2N2 where the discrimination coefficients, fi
andf2, are defined as:
- min {N1, N2} rip, + r2p2)
fl = Pi + P2,
The forms of the type 1 and type 2
losses are important since they determine f2= p1+p2 r
how the cost of the inefficiency is to be ri
and
distributed between the groups. In gen-
eral, the form is simply the product of W2= a + r2 -ay
aN E
two quantities: (1) the number of as-
sociating units and (2) the effective input
W1 and W2 indicate the average wage of
loss for each associating unit. Let us persons in groups 1 and 2, respectively.
assume that optimal intergroup associa- Thus, to the extent that the above in-
tion occurs when an associating unit terpretation is descriptive of reality, we
consists of K members of group 1 for would expect discrimination against
each member of group 2. Now the num- members of the economic minority.
ber of associating units is given as K-1 Furthermore, if the majority possessed
min {N1, KN2J. For physical labor, the greater per capita education, we would
loss accompanying each unit is Kpi + P2 expect discrimination against education
and for education is Krip, + r2p2. If, as to exceed discrimination against physical
I assume, K is independent of the respec- labor.
tive numbers of persons in each group, If a laborer derives disutility from
then min {N1, KN2} defines an economic associating with persons of other groups,
minority and majority in a very special a necessary condition for him to inte-
sense. As in this paper, let K = 1 so grate is that doing so increases his wage.
that the concepts of the economic Although this condition is necessary, it
majority and minority are consistent is not sufficient, since, as in Becker's
with their numeric definitions. Thus, example of employee discrimination,
members of the minority group must laborers discount the value of wages
absorb the full effects of the decline in earned under integration. For his model,
labor efficiency because the addition of the increase in wages would have to be
a minority laborer will reduce his effec- sufficient to compensate laborers for the
tive working time and that of one psychic costs of integrating.
laborer in the majority group with It is important that the analysis has
whom he is associated, whereas the considered only average wages. Thus,
addition of a majority laborer will not to the extent that there is variation in
non-whites may have generated differ- differentials would reflect the compound-
ences in tastes for market as opposed to ed effects of market discrimination and
non-market sources of income.7 We can the rational response to this discrimina-
observe only the earning capacity of in- tion. Since derived discrimination is ob-
cremental schooling, without an adjust- served simultaneously with market dis-
ment for non-pecuniary reward, and crimination, a problem of identification
would therefore expect the measured arises. Of course, the effects of derived
productivity of schooling to be lower for discrimination are partially observed
persons who place relatively high value since non-whites characteristically attend
on sources of satisfaction that are exter- school for fewer years and for fewer days
nal to the market. each year. Nevertheless, the incidence of
It is also important to note that the leisure while attending school is indis-
quantity of schooling a person possesses tinguishable from current discrimination.
is affected by his ejfective learning time, the estimates of market discrimina-
Thus
for which years of attendance nmay be a tion against the education of non-whites
rather poor approximation. The number provided here are biased upward.
of days in attendance may vary, and it For the base population, the average
is obvious that time passed in school is income of the whites is $2,690 and for
not necessarily productive. Idleness is a non-whites $1,045, the ratio being .39.
student's prerogative. During any given Although a slightly higher proportion of
period a person can choose between non-whites had no income, an adjust-
school attendance, work, and leisure. ment for those without income increases
Presumably, he will be idle as long as the the relative income of non-whites by less
marginal value of leisure exceeds the than .01.9
marginal value of present or future in- In order to adjust income for differ-
come foregone. For a student, the higher ences in age and the ownership of farm
the reward for increased effort, the great-
capital, I have estimated an income-gen-
er will be the effort forthcoming.8 Thus,erating function for the rural farm popu-
market discrimination against the serv- lation in which the logarithm of income
ices provided by the education of non- was regressed upon (1) state of residence,
whites can affect their behavior as stu- (2) years of school completed, (3) age,
dents. For example, non-white students and (4) farm capital.'0 Table 1 provides
who anticipate discrimination against estimates of average income by years of
their education might be absent a higher school completed for whites and non-
proportion of the time, and, while in whites in the ten states, holding the age
school, a higher proportion of their time distribution and the ownership of farm
may be devoted to leisure activities. capital constant at their national average
Once, the effort expended in learning levels. Having adjusted for age and farm
may be less because of the lower value capital, the relative income ratio of non-
of schooling. This is an instance of de- whites rises to .48-an increase of .09.
rived discrimination in which income The most striking feature demonstrated
7 This would be expected if there is discrimina- 9 The average income of whites with income was
tion against purchased goods since there can be no $2,865 and for non-whites, $1,130.
discrimination against leisure.
10 The estimates and the estimation procedure are
8 Here I assume a positively sloped supply func- discussed in detail in my dissertation (Welch,
tion of labor. 1966a).
in Table 1 is that the increase in income, capital constant at their national average
in both relative and absolute terms, per levels. An individual's wage is the sum
year of school completed is greater for of the marginal product of physical
whites than for non-whites. For example, labor, of which he has one unit, and the
marginal
if the distribution of schooling among the product of education times the
non-white population were the same as number of units of education he pos-
the white distribution, the relative in- sesses. Assume that the marginal prod-
come of non-whites would increase to ucts of physical labor and education are
only .56. That an additional year of constant within states but vary between
school is less valuable for a non-white states. Then,
may be interpreted as follows. First,
Wii = MP(N)j + eijMP(E)j,
quality of schooling may have been an
important source of discrimination. Sec- where MP(N)j and MP(E)j denote the
ond, the market may discriminate more marginal products of physical labor and
heavily against the productive services education, respectively, in state -j, and
provided by the education of non-whites eij represents the quantity of education
TABLE 1
ESTIMATED AVERAGE INCOME FOR RURAL FARM MALES BY YEARS OF SCHOOL COMPLETED
TABLE 2
Note.-These estimates were derived from observations of fifty-seven "states" for the eight schooling classes in each state. The
states were ten southern white and ten southern non-white plus thirty-seven states in which no separation by color was possible.
Alaska, Hawaii, and Rhode Island were omitted. The dependent variable for the regression was log (Wii - W0o), and the independent
variables were (1) a set of seven dummy variables to indicate years of school completed, the zero attendance class being omitted,
and (2) a set of fifty-seven dummy variables to indicate state of residence. R2 = .89.
resulting estimate for the profile of time- only 28 per cent of the corresponding in-
in-school attendance into schooling and crease for whites. Thus, school attendance
the average marginal product of schooling is simply a better investment for whites.
for southern whites and non-whites are In short, the opportunity cost of school
provided in Table 2. Since the units in attendance, relative to income gains from
which schooling is measured are arbi- attendance, is much higher for non-
trary, the estimate is scaled so that one whites.
unit of schooling represents eight years Although discrimination against school-
of attendance. The marginal product of ing is the relevant parameter to consider
schooling, therefore, corresponds to an for analyzing investments in school at-
estimate of the annual increment in in- tendance by non-whites, it is useful to
come derived from eight years in attend- separate this factor into two components:
ance. differential quality of schooling and cur-
Thus, from Table 1, the discrimination rent market discrimination.
coefficient against non-white physical Value added by the rural farm popula-
labor is .19 (1 minus the ratio of the tion is assumed to be a Cobb-Douglas
TABLE 3
TABLE 4
VARIABLES
The above estimates can be considered school systems (relative quality for Negro
as (1) an overestimate of discrimination schools is estimated as .73), the relative
against education and (2) an underesti- quantity of education per year of school
mate of discriminatory quality of school- enrolment is only .66 for Negroes. This
ing. These biases arise largely because I alteration requires an adjustment in the
have not explicitly considered the effects estimated discrimination coefficient for
of historic discrimination.'7 For example, education which reduces it from .68 to
average annual days of attendance can .58.19
be observed, yet for statistical purposes'8
SUMMARY
this variable was not included, and the
effect of differences in days of attendance My approach to the explanation of
is included in the estimated discrimina- market discrimination entails using a
tion coefficient for education. In 1945, model in which the integration of labor-
southern white students who were en- ers generates external effects which move
rolled in school averaged 149 days in in opposite directions. First, integration
attendance, whereas Negro students av- increases the productivity of laborers as
eraged 136 days (U.S. Office of Educa- a consequence of complementarity be-
tion, 1950). If we assume that the skill tween workers of different education.
acquired per day of attendance is con- Second, it generates a class of external
stant, differences in attendance rates inefficiencies in which the productivity
imply that Negroes acquire only 91 per of a laborer may decline if he is required
cent as much schooling per year of en- to work alongside a laborer of a different
rolment, and, allowing for differences in race. The advantage of this interpreta-
the quality of the services offered by tion is that it does not rely upon imper-
fections in capital markets and product
increases by 7 per cent and the white's declines by differentiation by consumers or upon an
26 per cent. But, when the discrimination co- assumption of specialized factors such as
efficients are adjusted for differences in school-
attendance rates, the estimate of Pi becomes 0 and "entrepreneurship." It is not that these
of P2, .20. considerations are irrelevant; rather, it
17 There are also statistical reasons for anticipat- is suggested that labor externalities
ing these biases. The school systems' inputs are ob- should also be considered in a complete
served subject to rather large errors of estimation
specification of discrimination. Discrimi-
which bias the quality coefficients downward and un-
derstate differences in quality of schooling. Also, the nation and segregation should be dis-
residual variance of the logarithmic estimating equa- tinguished, since, as the analysis indi-
tion is likely to be greater for non-whites than for
whites, implying that by taking the antilog of the
cates, discrimination does not necessarily
average predicted log, the relative understatement result in segregation. Other advantages
of average values is greater for non-whites than are that the model provides for measure-
whites.
ment of the direct social cost of discrimi-
18 There is apparently no systematic within race
nation in terms of product foregone and
relationship between days of attendance and the
value of schooling. The only significant variation oc- that the model is operational since it
curs between whites and non-whites. Thus annual provides an aggregation criterion for
days in attendance is highly collinear with the
combining labor inputs for persons of
white-non-white dummy variable which identifies
the market-discrimination coefficient for education. different races. The most critical features
The separate effects of these variables could not be are: (1) the assumed constancy of the
observed, and the estimated discrimination co-
efficient for education contains the effects of differ- 19 This adjustment alters the implicit estimates
ences in attendance rates. of pi and P2 (see n. 16) to pi = 0; P2 = .2.
TABLE 5
5-7 8 12
Income:
White .............................. $2,090 $2,340 $3,790
Non-white .................. 1,300 1,480 1,840
Difference .......................... 790 860 1,950
1. Impact of market discrimination against
physical labor . ........... 250 250 250
2. Impact of discrimination against school-
ing* .... .......................... 540 610 1,700
a) Inferior quality of schooling ... . 200 230 630
b) Market discrimination against ed-
ucation ........................ 340 380 1,070
* The adjustment for interaction between quality of schooling and market discrimination
against education is prorated according to the proportion of the total (difference in the return to
schooling) accounted for by each. Actually, interaction represents 14 per cent of the total dis-
crimination against schooling.
of the underlying sources, social, psycho- for non-whites. Furthermore, the esti-
logical, and economic, of market dis- mates of differences in quality of school-
crimination. ing, although substantial, account for
To illustrate the essential characteris- only 37 per cent of the discrimination
tics of the model, I have provided esti- against schooling-implying that market
mates for males in the rural South. These discrimination against education is a
estimates are best summarized by con- more important source of income differ-
sidering the impact on earnings of both entials. It would seem that discrimina-
inferior quality of schooling and market tory quality of schooling is more easily
discrimination for persons who have eliminated than market discrimination,
completed five to seven and eight years because legislative authorities have rela-
of school since these schooling classes are tively little control over such markets.
representative of rural areas. One feature In fact, to the extent that market dis-
of the findings which is of particular im- crimination is determined largely by
portance is that as a non-white increases sociological phenomena, we cannot ex-
his investment in schooling, the impact pect these factors to be eliminated either
quickly or easily. Nevertheless, the elimi- creased investment in schooling, and (3)
nation of discrimination in quality of induce greater effort while in school,
schooling may be an important vehicle which will increase the quantity of edu-
for removing income differences; for an cation per unit of attendance time. In
improvement in the quality of schooling addition, the reduction of differences in
will: (1) reduce the observed discrimina- education may reduce associational fric-
tion against schooling, (2) induce an in- tion, which then reduces discrimination.
APPENDIX
REFERENCES
Becker, Gary S. The Economics of Discrimina- Education, 1944-46. Washington: U.S. Gov-
lion. Chicago: Univ. of Chicago Press, 1957. ernment Printing Office, 1950.
Chadwick-Jones, J. K. "Italian Workers in a Biennial Survey of Education, 1954-56.
British Factory: A Study of Informal Selec- Washington: U.S. Government Printing
tion and Training," Race, J. Institute Race Office, 1959.
Welch, F. "Determinants of the Return to
Relations (July, 1964).
Schooling in Rural Farm Areas, 1959."
Kislev, Yoav. "Estimating a Production Func-
Unpublished Ph.D. dissertation, Dept. of
tion from 1959 to U.S. Census of Agriculture
Econ., Univ. of Chicago, 1966 (a).
Data." Unpublished Ph.D. dissertation, '. "Measurement of the Quality of
Dept. of Econ., Univ. of Chicago, 1965. Schooling," A.E.R., LVI, No. 2 (May, 1966),
U.S. Office of Education. Biennial Survey of 379-92. (b)