Investment in Human Capital - Becker

Investment in Human Capital: A Theoretical Analysis
Author(s): Gary S. Becker

Source: Journal of Political Economy, Vol. 70, No. 5, Part 2: Investment in Human Beings
(Oct., 1962), pp. 9-49
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/1829103
Accessed: 24-03-2017 13:10 UTC
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
http://about.jstor.org/terms
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Political Economy
This content downloaded from 128.103.149.52 on Fri, 24 Mar 2017 13:10:07 UTC
All use subject to http://about.jstor.org/terms
INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS'
GARY S. BECKER
Columbia University and National Bureau of Economic Research
I. INTRODUCTION prove the physical and mental abilities of

people and thereby raise real income
SOME activities primarily affect fu-
prospects.
ture well-being, while others have
their main impact in the present. People differ substantially in their
Dining is an example of the latter, while economic well-being, both among coun-
purchase of a car exemplifies the former. tries and among families within a given
Both earnings and consumption can be country. For a while economists were re-
affected: on-the-job training primarily lating these differences primarily to dif-

ferences in the amount of physical capital
affects earnings, a new sail boat primari-
ly affects consumption, and a college since richer people had more physical
education is said to affect both. The ef- capital than others. It has become in-
fects may operate either through physical creasingly evident, however, from studies
resources, such as a sail boat, or through of income growth2 that factors other
human resources, such as a college edu- than physical resources play a larger
role than formerly believed, thus focus-
cation. This paper is concerned with
ing attention on less tangible resources,
activities that influence future real in-
like the knowledge possessed. A concern
come through the imbedding of resources
in people. This is called investing in with investment in human capital,
human capital. therefore, ties in closely with the new
The many ways to invest include emphasis on intangible resources and
schooling, on-the-job training, medical may be useful in attempts to understand

care, vitamin consumption, and acquir- the inequality in income among people.
The original aim of my study was to
ing information about the economic
system. They differ in the relative effects estimate the money rate of return to
on earnings and consumption, in the college and high-school education in the

United States. In order to set these esti-
amount of resources typically invested,
mates in proper context I undertook a
in the size of returns, and in the extent to
brief formulation of the theory of invest-
which the connection between invest-
ment in human capital. It soon became
ment and return is perceived. But all im-
clear to me, however, that more than a
1 I am greatly indebted to the Carnegie Corpora-restatement was called for: while im-
tion of New York for the support given to the Na- portant and pioneering work had been
tional Bureau of Economic Research to study invest- done on the economic return to various
ment in education and other kinds of human capital.
I benefited greatly from many discussions with my 2 The evidence for the United States appears to
colleague Jacob Mincer, and also with other partici- show that the growth in capital per capita explains
pants in the Labor Workshop of Columbia Univer- only a small part of the growth in per capita income
sity. Although many persons offered valuable com- and that the growth in "technology" explains most
ments on the draft prepared for the conference, I am of it. On this see S. Fabricant, Economic Progress and
especially indebted to the detailed comments of Economic Change: 34th Annual Report of tIe National
Theodore Schultz, George Stigler, and Shirley John- Bureau of Economic Research (New York: National
son. Bureau of Economic Research, 1954).
1() GARY S. BECKER
occupations and education classes,3 there The typical investor in human capital is
have been few, if any, attempts to treat more impetuous and thus more likely to
the process of investing in people from err than is the typical investor in tangible
a general viewpoint or to work out a capital. What a diverse and possibly even
broad set of empirical implications. I confusing array! Yet all these as well as
began then to prepare a general analysis many other important empirical impli-
of investment in human capital. cations can be derived from very simple
As the work progressed, it became theoretical arguments. The purpose of
clearer and clearer that much more than this paper is to set out these arguments
a gap in formal economic analysis would in some generality, with the emphasis
be filled, for the analysis of human in- placed on empirical implications, al-
vestment offered a unified explanation of though little empirical material is pre-
a wide range of empirical phenomena sented. My own empirical work will ap-
which had either been given ad hoc inter- pear in a later study.
pretations or had baffled investigators. First, a lengthy discussion of on-the-
Among these are the following: (1) Earn- job training is presented and then, much
ings typically increase with age at a de- more briefly, discussions of investment in
creasing rate. Both the rate of increase schooling, information, and health. On-
and the rate of retardation tend to be the-job training is dealt with so elabo-
positively related to the level of skill. (2) rately not because it is more important
Unemployment rates tend to be nega- than other kinds of investment in human
tively related to the level of skill. (3) capital-although its importance is often
Firms in underdeveloped countries ap- underrated-but because it clearly illus-
pear to be more "paternalistic" toward trates the effect of human capital on
employees than those in developed coun- earnings, employment, and other eco-
tries. (4) Younger persons change jobs nomic variables. For example, the close
more frequently and receive more school- connection between foregone and direct
ing and on-the-job training than older costs or the effect of human capital on
persons do. (5) The distribution of earn- earnings at different ages is vividly
ings is positively skewed, especially brought out. The extended discussion of
among professional and other skilled on-the-job training paves the way for
workers. (6) Abler persons receive more much briefer discussions of other kinds of
education and other kinds of training investment in human beings.
than others. (7) The division of labor is II. DIFFERENT KINDS OF INVESTMENT
limited by the extent of the market. (8)
A. ON THE JOB
I In addition to the earlier works of Smith, Mill, Theories of firm behavior, no matter
and Marshall, see H. Clark, Life Earnings in Selected how they differ in other respects, almost
Occupations in the U.S. (New York: Harper & Bros.,
1937); J. R. Walsh, "Capital Concept Applied to
invariably ignore the effect of the produc-
Man," Quarterly Journal of Economics, February, tive process itself on worker productiv-
1935; M. Friedman and S. Kuznets, Incomefrom In- ity. This is not to say that no one
dependent Professional Practice (New York: National
Bureau of Economic Research, 1945); G. Stigler and
recognizes that productivity is affected
D. Blank, The Demand and Supply of Scientific Per- by the job itself; but the recognition has
sonnel (New York: National Bureau of Economic not been fonnalized, incorporated into
Research, 1957); and T. W. Schultz, "Investment in
Man: An Economist's View," Social Service Review,
economic analysis, and its implications
June, 1959. worked out. We now intend to do just
INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 11
that, placing special emphasis on the where W equals wages or expenditures

broader economic implications. and MP equals the marginal product or
Many workers increase their produc- receipts. Firms would not worry too
tivity by learning new skills and perfect- much about the relation between labor
ing old ones while on the job.. For ex- conditions in the present and future
ample, the apprentice usually learns a partly because workers were only hired
completely new skill while the intern de- for one period, and partly because wages
velops skills acquired in medical school, and marginal products in future periods
and both are more productive afterward. would be independent of a firm's current
On-the-job training, therefore, is a behavior. It can therefore legitimately be
process that raises future productivity assumed that workers have unique mar-
and differs from school training in that ginal products (for given amounts of
an investment is made on the job rather other inputs) and wages in each period,
than in an institution that specializes in which are, respectively, the maximum
teaching. Presumably, future productiv- productivity in all possible uses and the
ity can be improved only at a cost, for market wage rate. A more complete set
otherwise there would be an unlimited of equilibrium conditions would be the set
demand for training. Included in cost are
MPt= Wt, (2)
a value placed on the time and effort of
trainees, the "teaching" provided by
where t refers to the tth period. The
others, and the equipment and materials
equilibrium position for each period would
used. These are costs in the sense that
depend only on the flows during that
they could have been used in producing
period.
current output if they were not used in
These conditions are altered when ac-
raising future output. The amount spent
count is taken of on-the-job training and
and the duration of the training period
the connection thereby created between
depend partly on the type of training-
present and future receipts and expendi-
more is spent for a longer time on an
tures. Training might lower current
intern than on an operative-partly on
receipts and raise current expenditures,
production possibilities, and partly on
yet firms could profitably provide this
the demand for different skills.
training if future receipts were sufficient-
Each employee is assumed to be hired
ly raised or future expenditures suf-
for a specified time period (in the limiting
ficiently lowered. Expenditures during
case this period approaches zero), and for
each period need not equal wages,
the moment both labor and product
receipts need not equal the maximum
markets are assumed to be perfectly
possible productivity, and expenditures
competitive. If there were no on-the-job
and receipts during all periods would be
training, wage rates would be given to
interrelated. The set of equilibrium
the firm and would be independent of its
conditions summarized in equation (2)
actions. A profit-maximizing firm would
would be replaced by an equality be-
be in equilibrium when marginal products
tween the present values of receipts and
equaled wages, that is, when marginal
expenditures. If Et and Rt represent ex-
receipts equaled marginal expenditures.
penditures and receipts during period t,
In symbols
and i the market discount rate, then the
MP=W, (1) equilibrium condition can be written as
12 GARY S. BECKER
Rn -I t n -I E t as the sum of opportunity costs and out-

v (3)
lays on training, (6) becomes
0 (I +i) t=o +
MP'+G=Wo+C. (7)
where n represents the number of
periods, and R, and E, depend on all The term G, the excess of future
other receipts and expenditures. The receipts over future outlays, is a measure
equilibrium condition of equation (2) of the return to the firm from providing
has been generalized, for if marginal training; and, therefore, the difference
product equals wages in each period, the between G and C measures the difference
present value of the marginal product between the return from, and the cost of,
stream would have to equal the present training. Equation (7) shows that mar-
value of the wage stream. Obviously, ginal product would equal wages in the
however, the converse need not hold. initial period only when the return
If training were given only during the equals costs, or G = C; it would be
initial period, expenditures during the greater or less than wages as the return
initial period would equal wages plus the was smaller or greater than costs. Those
outlay on training, expenditures during familiar with capital theory might argue
other periods would equal wages alone, that this generalization of the simple
and receipts during all periods would equality between marginal product and
equal marginal products. Equation (3) wages is spurious because a full equi-
becomes librium would require equality between
the return from an investment-in this
n-i MRt
MPO + _ - case, made on the job and costs. If this
O +1 ( 1 +0Z t implied that G = C, marginal product
(4) would equal wages in the initial period.
=WO + k +E t)tX There is much to be said for the rele-
vance of a condition equating the return
from an investment with costs, but such
where k measures the outlay on training.
a condition does not imply that G = C or
If a new term is defined,
that marginal product equals wages.
The following discussion demonstrates
G= E M1-Wt (5)
t=__ (1 +iW that great care is required in the applica-
tion of this condition to on-the-job
equation (4) can be written as investment.
1. General. Our treatment of on-the-
APo P+G=Wo+k. (6)
job training produced some general re-
Since the term k only measures the actual sults summarized in equations (3) and
outlay on training it does not entirely (7) of wide applicability, but more con-
measure training costs, for excluded is crete results require more specific as-
the time that a person spends on this sumptions. In this and the following sec-
training, time that could have been used tion two types of on-the-job training are
to produce current output. The differ- discussed in turn: general and specific.
ence between what could have been pro- General training is useful in many firms
duced, call this MPo and what is pro- in addition to the firm providing it, as a
duced, MPo, is the opportunity cost of machinist trained in the army finds his
the time spent in training. If C is defined skills of value in steel and aircraft firms,
or a doctor trained (interned) at one Equation (7) is reduced to

hospital finds his skills useful at other
MP,=W WO+C, (9)
hospitals. Most on-the-job training pre-
or
sumably increases the future marginal
product of workers in the firm providing VO=MP, -C. (10)
it, but general training would also in-
In terms of actual marginal product
crease their marginal product in many
other firms as well. Since in a competitive MPo=Wo+ k, (9')
labor market the wage rates paid by any or
firm are determined by marginal produc- Wo=MPo-k. (10')
tivities in other firms, future wage rates
as well as marginal products would in- The wage of trainees would not equal
crease to firms providing general train- their opportunity marginal product but
ing. These firms could capture some of would be less by the total cost of train-
the return from training only if their ing. In other words, employees would pay
marginal product rose by more than for general training by receiving wages
their wages. "Perfectly general" training below their current (opportunity) pro-
would be equally useful in many firms ductivity. Equation (10) has many other
and marginal products would rise by the implications, and the rest of this section
same extent in all of them. Consequent- is devoted to developing the more im-
ly, wage rates would rise by exactly the portant ones.
same amount as the marginal product Some might argue that a really "net"
and the firms providing such training definition of marginal product obtained
could not capture any of the return. by subtracting training costs from
Why, then, do rational firms in com- "gross" marginal product must equal
petitive labor markets provide general wages even for trainees. Such an inter-
training, for why provide training that pretation of net productivity could for-
brings no return? The answer is that mally save the equality between marginal
firms would provide general training product and wages here, but later I show
only if they did not have to pay any of (pp. 18-25) that it cannot always be
the costs. Persons receiving general saved. Moreover, regardless of which in-
training would be willing to pay these terpretation is used, training costs would
costs since training raises their future have to be included in any study of the
wages. Hence the cost as well as the re- relation between wages and productivity.
turn from general training would be Employees pay for general on-the-job
borne by trainees, not by firms. training by receiving wages below what
These and other implications of gen- could be received elsewhere. "Earnings"
eral training can be more formally during the training period would be the
demonstrated with equation (7). Since difference between an income or flow
wages and marginal products are raised term, potential marginal product, and a
by the same amount, MP, must equal capital or stock term, training costs, so
W, for all t = 1, .. .n-1 ,and there- that the capital and income accounts
fore would be closely intermixed, with
__ =__ 0MP- t=?. ( 8)
changes in either affecting wages. In
(1+it - t other words, earnings of persons receiv-
t=1 ( 8 ing on-the-job training would be net of
14 GARY S. BECKER
investment costs and would correspond signs of these correlations might even
to the definition of net earnings used differ.6
throughout this paper, which subtracts Doubt has been cast on the frequent
all investment costs from "gross" earn- assertion that no allowance is made in
ings. Therefore, our departure with this the income accounts for depreciation on
definition of earnings from the account- human capital.7 A depreciation-type
ing conventions used for transactions in item is deducted, at least from the earn-
material goods-which separate income ings due to on-the-job training, for the
from capital accounts to prevent a cost would be deducted during the train-
transaction in capital from ipso factor4 ing period. Depreciation on tangible
affecting the income side is not capri- capital does not bulk so large in any one
cious but is grounded in a fundamental period because it is usually "written off"
difference between the way investment inor depreciated during a period of time
material and human capital are "written designed to approximate its economic
off." The underlying cause of this differ- life. Hence human and tangible capital
ence undoubtedly is the widespread re- appear to differ more in the time pattern
luctance to treat people as capital and of depreciation than in its existence,8
the accompanying tendency to treat all and the effect on wage income of a rapid
wage receipts as earnings. "write-off" of human capital is what
Intermixing the capital and income should often be emphasized and studied.
accounts could make the reported "in- Our point can be put differently and
comes" of trainees unusually low and per- more rigorously. The ideal depreciation
haps negative, even though their long- on a capital asset during any period
run or lifetime incomes were well above would equal its change in value during
average. Since a considerable fraction of the period. In particular, if value rose, a
young persons receive some training, and negative depreciation term would have
since trainees would tend to have lower
6 A difference in signs is impossible in Friedman's
current and higher subsequent earnings analysis of consumer behavior because he assumes
than other youth, the correlation be- that transitory and long-run (that is, permanent)
incomes are uncorrelated (see his A Theory of the
tween current consumption and current
Consumption Function [Princeton, N.J.: Princeton
earnings of young people' would not University Press, 1959]); we are suggesting that they
only be much weaker than the correla- may be negatively correlated for young persons.
tion with long-run earnings, but the I See, for example, A. Marshall, Principles of Eco-
nomics (8th ed.; New York: Macmillan Co., 1949); C.
Christ, "Patinkin on Money, Interest, and Prices,"
4 Of course, Eshift between assets having differ- Journal of Political Economy, August, 1957, p. 352;
ent productivities would affect the income account and W. Hamburger, "The Relation of Consumption
on material goods even with current accounting to Wealth and the Wage Rate," Econometrica, Janu-
practices. ary, 1955.
I I say "young people" rather than "young fam- 8 In a recent paper, R. Goode has argued (see
ilies" because as J. Mincer has shown (in a paper to "Educational Expenditures and the Income Tax," in
be published in a National Bureau of Economic Re- Selma J. Mushkin [ed.], Economics of Higher Educa-
search conference volume on labor economics), the tion [Washington: United States Department of
labor-force participation of wives is positively cor- Health, Education, and Welfare (forthcoming)])
related with the difference between husbands' long- that educated persons should be permitted to sub-
run and current income. Participation of wives, tract from income a depreciation allowance on tui-
therefore, makes the correlation between a family's tion payments. Such an allowance is apparently not
current and a husband's long-run income greater required for on-the-job training costs; indeed, one
than that between a husband's current and long-run might argue, on the contrary, that too much or too
income. rapid depreciation is permitted on such investment.
INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANAILYSIS 15
to be subtracted or a positive apprecia- period and equal afterwards, they would

tion term added to the income from the rise sharply at the end of the training
asset. Since training costs would be de- period and then level off (as shown by the
ducted from earnings during the training dashed line T'T' in Fig. 1), imparting a
period, the economic "value" of a trainee concave appearance to the curve as a
would at first increase rather than de- whole. In this extreme case an extreme
crease with age, and only later would it concavity appears; in less extreme cases
begin to decrease.9 the principle would be the same and the
Training has an important effect on concavity more continuous.
the relation between earnings and age. Foregone earnings are an important,
Suppose that untrained persons received although neglected, cost of much human
the same earnings regardless of age, as capital and should be treated on the same
shown by the horizontal line UU in footing as direct outlays. Indeed, all costs
Figure 1. Trained persons would receive appear as foregone earnings to workers
lower earnings during the training period
T
because training is paid for then, and
higher earnings at later ages because the
return is collected then. The combined
effect of paying for and collecting the re-
zc/ I,_ U
turn from training in this way would be
to make the age earnings curve of trained
persons, shown by TT in Figure 1,
steeper than that of untrained persons,
the difference being greater the greater AGE
the cost of, and return from, the invest- FIG. 1

ment.
Not only does training make the receiving on-the-job training; that is, all
curve steeper but, as indicated by costs appear as lower earnings than could
Figure 1, also more concave; that is, the be received elsewhere, although direct
rate of increase in earnings is affected outlays, C, may really be an important
more at younger than at older ages. Sup- part of costs. The arbitrariness of the
pose, to take an extreme case, that train- division between foregone and direct
ing raised the level of marginal produc- costs and the resulting advantage of
tivity but had no effect on the slope, so treating total costs as a whole" can be
that the marginal productivity of trained
10 The equivalence between foregone and direct
persons was also independent of age. If costs applies to consumption as well as to investment
earnings equaled marginal product, TT decisions. A household can be assumed to maximize a
utility function
would merely be parallel to and higher
than UU, showing neither slope nor con- U(K1, X 2, * X)
cavity. Since, however, earnings of X .i . X being consumption goods, subject to the
constraint
trained persons would be below mar-
ginal productivity during the training
P-ix=wV (h - E lX1)? + Y
9 In my study for the National Bureau of Eco-
nomic Research I try to measure the relation be-
tween depreciation and age for several education where pi is the market price of the ith good, I14
classes. average wage rate, y non-wage income, hI the total
16 GARY S. BECKER
further demonstrated by contrasting training in a wide variety of skills and

school and on-the-job training. Usually many-such as piloting and machine re-
only the direct cost of school training is pair-are very useful in the civilian sec-
emphasized, even though the foregone tor. Training is provided during part or
cost is sometimes (as with college educa- all of the first enlistment period and used
tion) an important part of the total. A during the remainder of the first period
shift of training from schools to on the and hopefully during subsequent periods.
job would, however, reverse the emphasis This hope, however, is thwarted by the
and make all costs appear as foregone fact that re-enlistment rates tend to be
earnings, even when direct outlays were inversely related to the amount of
important. civilian-type skills provided by the
Income maximizing firms in competi- military.11 Persons with these skills leave
tive labor markets would not pay the the military more readily because they
cost of general training and would pay can receive much higher wages in the
trained persons the market wage. If, civilian sector. Net military wages for
however, training costs were paid, many those receiving training are higher rela-
persons would seek training, few would tive to civilian wages during the first than
quit during the training period, and labor during subsequent enlistment periods be-
costs would be relatively high. Firms cause training costs are largely paid by
that did not pay trained persons the the military. Not surprisingly, therefore,
market wage would have difficulty satis- first-term enlistments for skilled jobs are
fying their skill requirements and would obtained much more easily than are re-
also tend to be less profitable than other enlistments.
firms. Firms that both paid for training The military is a conspicuous example
and less than the market wage for trained of an organization that both pays at
persons would have the worst of both least part of training costs and does not
worlds, for they would attract too many pay market wages to skilled personnel. It
trainees and too few trained persons. has had, in consequence, relatively easy
These principles have been clearly access to "students" and heavy losses of
demonstrated during the last few years "graduates." Indeed, its graduates make
in discussions of problems in recruiting up the predominate part of the supply in
military personnel. The military offers several civilian occupations. For ex-
ample, well over 90 per cent of United
number of hours available for either consumption or States commercial airline pilots received
work, and hj the number of hours required to con- much of their training in the armed
sume a unit of the jth good. By transposing terms
forces. The military, of course, is not a
the constraint can be written as
commercial organization judged by
2(p?+Whti))X =Wh+y . profits and losses and has had no diffi-
The total cost or price of consuming a unit of the ith culty surviving and even thriving.
good is the sum of two components: the market price What about the old argument that
or direct outlay per unit, pi, and the foregone earn-
ings per unit, Whi. I expect to show in another paper " See Manpower Management and Compensation
that this formulation of household decisions gives (Washington: Government Printing Office, 1957),
extremely useful insights into a number of important Vol. I, Chart 3, and the accompanying discussion.
economic problems, such as the choice between la- The military not only wants to eliminate the inverse
bor and "leisure," the effect of price control on relation but apparently would like to create a strong
prices, the role of queues, and the cause of differences positive relation because they have such a large in-
among income classes in price elasticities of demand. vestment in heavily trained personnel (see ibid.).
firms in competitive labor markets have amount in firms providing the training
no incentive to provide on-the-job train- than in other firms. Training that in-
ing because trained workers would be bid creases productivity more in firms pro-
away by other firms? Firms that train viding it will be called specific training.
workers are supposed to impart external Completely specific training can be de-
economies to other firms because the fined as training that has no effect on the
latter can use these workers free of any productivity of trainees that would be
training charge. An analogy with re- useful in in other firms. Much on-the-job
search and development is often drawn training is neither completely specific
since a firm developing a process that not completely general but increases
cannot be patented or kept secret would productivity more in firms providing it
impart external economies to competi- and falls within the definition of specific
tors."2 This argument and analogy would training. The rest increases productivity
apply if firms were to pay training costs, by at least as much in other firms and
for they would suffer a "capital loss" falls within a definition of general train-
whenever trained workers were bid away ing. The previous section discussed gen-
by other firms. Firms can, however, shift eral training and this one will cover
training costs to trainees and have an in- specific training. A few illustrations of
centive to do so when faced with compe- the scope of specific training are pre-
tition for their services. sented before a formal analysis is de-
The difference between investment in veloped.
training and in research and develop- The military offers some forms of
ment can be put very simply. Without training that are extremely useful in the
patents or secrecy, firms in competitive civilian sector, as already noted. Train-
industries cannot establish property ing is also offered that is only of minor
rights in innovations, and these innova- use to civilians: astronauts, fighter pilots,
tions become fair game for all comers. and missile men all illustrate this to a
Patent systems try to establish these greater or lesser extent. Such training
rights so that incentives can be provided falls within the scope of specific training
to invest in research. Property rights in because productivity is raised in the
skills, on the other hand, are automati- military but not (much) elsewhere.
cally vested, for a skill cannot be used Resources are usually spent by firms
without permission of the person pos- in familiarizing new employees with their
sessing it. This property right in skills organization, and the knowledge so
is the source of the incentive to invest in acquired is a form of specific training
training and explains why an analogy because productivity is raised more in
with unowneq innovations is misleading. the firms acquiring the knowledge than
2. Specific.-Completely general train- in other firms. Other kinds of hiring
ing increases the marginal productivity costs, such as employment agency fees,
of trainees by exactly the same amount the expenses incurred by new employees
in firms providing the training as in other in finding jobs (what Stigler calls in his
firms. Clearly some kinds of training in- paper in this Supplement the "costs of
crease productivity by a different 13 To judge by a sample of firms recently ana-
lyzed, formal orientation courses are quite common,
12 These arguments can be found in Marshall, op. at least in large firms (see H. F. Clark and H. S.
cit., pp. 565-66, although he compares training to Sloan, Classrooms in the Factories [New York: New
land-tenure systems. York University Press, 19551, chap. iv).
18 GARY S. BECKER
search"), or the time employed in inter- that come under the rubric of specific in-
viewing, testing, checking references, and vestment. This set is now treated ab-
in bookkeeping do not so obviously raise stractly in order that a general formal
the knowledge of new employees, but analysis can be developed. Empirical
they too are a form of specific investment situations are brought in again after
in human capital, although not training. several major implications of the formal
They are an investment because outlays analysis have been developed.
over a short period create distributed If all training were completely specific,
effects on productivity; they are specific the wage that an employee could get
because productivity is raised primarily elsewhere would be independent of the
in the firms making the outlays; they are amount of training he had received. One
in human capital because they lose their might plausibly argue, then, that the
value whenever employees leave. In the wage paid by firms would also be inde-
rest of this section I usually refer only to pendent of training. If so, firms would
on-the-job specific training even though have to pay training costs, for no ra-
the analysis applies to all on-the-job tional employee would pay for training
specific investment. that did not benefit him. Firms would
Even after hiring costs are incurred, collect the return from such training in
firms usually know only a limited amount the form of larger profits resulting from
about the ability and potential of new higher productivity, and training would
employees. They try to increase their be provided whenever the return-dis-
knowledge in various ways-testing, counted at an appropriate rate-was at
rotation among departments, trial and least as large as the cost. Long-run
error, etc.-for greater knowledge per- competitive equilibrium requires that the
mits a more efficient utilization of man- present value of the return exactly
power. Expenditures on acquiring knowl- equals costs.
edge of employee talents would be a These propositions can be stated more
specific investment if the knowledge formally with the equations developed
could be kept from other firms, for then earlier. According to equations (5) and
productivity would be raised more in the (7) the equilibrium of a firm providing
firms making the expenditures than else- training in competitive markets can be
where. written as
The effect of investment in employees
on their productivity elsewhere depends
on market conditions as well as on the 0~ + = G [E +0-t ( 1 1 )
nature of the investment. Very strong = lvo+C
monopsonists might be completely insu-
lated from competition by other firms, where C is the cost of training given only
and practically all investments in their in the initial period, MPo is the oppor-
labor force would be specific. On the tunity marginal product of trainees, Wo
other hand, firms in extremely competi- is the wage paid to trainees, and Wt
tive labor markets would face a constant and MPt are the wage and marginal
threat of raiding and would have fewer product in period t. If the analysis of
specific investments available. completely specific training given in the
These examples convey some of the preceding paragraph was correct, W
surprisingly large variety of situations would always equal the wage that could
be received elsewhere, MPt - Wt would paid for specific training would be un-
be the full return in t from training given able to collect any further return and
would also suffer a capital loss. The
in 0, and G would be the present value
of these returns. Since MPo measures the willingness of workers or firms to pay for
marginal product elsewhere and WO specific training should, therefore, closely
would measure the wage elsewhere of depend on the likelihood of labor turn-
trainees, MP' = Wo. As a consequence
over.
G = C, or, in full equilibrium, the return To bring in turnover at this point may
from training equals costs. seem like a deus ex machine since it is al-
Before claiming that the usual equal- most always ignored in traditional
ity between marginal product and wages theory. In the usual analysis of competi-
holds when completely specific training tive firms, wages equal marginal prod-
is considered, the reader should bear in uct, and since wages and marginal
mind two points. The first is that the product are assumed to be the same in
equality between wages and marginal many firms, no one suffers from turn-
product in the initial period involves op- over. It would not matter whether a
portunity, not actual marginal product. firm's labor force always contained the
Wages would be greater than actual same persons or a rapidly changing
marginal product if some productivity group. Any person leaving one firm could
was foregone as part of the training pro- do equally well in other firms, and his
gram. The second is that, even if wages employer could replace him without any
equaled marginal product initially, they change in profits. In other words, turn-
would be less in the future because the over is ignored in traditional theory be-
differences between future marginal cause it plays no important role within
products and wages constitute the return the framework of the theory.
to training and are collected by the firm. Turnover becomes important when
All of this follows from the assumption costs are imposed on workers or firms,
that firms pay all costs and collect all which are precisely the effects of specific
returns. But could not one equally well training. Suppose a firm paid all the
argue that workers pay all specific train- specific training costs of a worker who
ing costs by receiving appropriately quit after completing it. According to our
lower wages initially and collect all re- earlier analysis he would have been re-
turns by receiving wages equal to mar- ceiving the market wage and a new em-
ginal product later? In terms of equation ployee could be hired at the same wage.
(11), Wt would equal MPt, G would equalIf the new employee were not given
zero, and Wo=MP'-C, just as with training, his marginal product would be
general training. Is it more plausible less than that of the one who quit since
that firms rather than workers pay for presumably training raised the latter's
and collect and return from training? productivity. Training could raise the
An answer can be found by reasoning new employee's productivity but would
along the following lines. If a firm had require additional expenditures by the
paid for the specific training of a worker firm. In other words, a firm is hurt by the
who quit to take another job, its capital departure of a trained employee because
expenditure would be partly wasted, for an equally profitable new employee
no further return could be collected. could not be obtained. In the same way
Likewise, a worker fired after he had an employee who pays for specific train-
20 GARY S. BECKER
ing would suffer a loss from being laid off crease. Such training can be looked upon
because he could not find an equally as the sum of two components, one com-
good job elsewhere. To bring turnover pletely general, the other completely spe-
into the analysis of specific training is cific, with the former being relatively
not, therefore, a deus ex machine but is larger the greater the effect on wages
made necessary by the important link in other firms relative to the firms pro-
between them. viding the training. Since firms do not
Firms paying for specific training pay any of completely general costs and
might take account of turnover merely only part of completely specific costs, the
by obtaining a sufficiently large return fraction of costs paid by firms would be
from those remaining to counterbal- negatively related to the importance of
ance the loss from those leaving. (The re- the general component, or positively re-
turn on "successes"-those remaining- lated to the specificity of the training.
would, of course, overestimate the aver- Our conclusions can be stated formal-
age return on all training expenditures.) ly in terms of the equations developed
Firms could do even better, however, by earlier. If G is the present value of the re-
recognizing that the likelihood of a quit turn from training collected by firms, the
is not fixed but depends on wages. In- fundamental equation is
stead of merely recouping on successes
MP' + G = W + C. (12)
what is lost on failures, they might re-
duce the likelihood of failure itself by If G' measures the return collected by
offering higher wages after training than employees, the total return, G", would
could be received elsewhere. In effect, be the sum of G and G'. In full equi-
they would offer employees some of the librium the total return would equal total
return from training. Matters would be costs, or G" = C. Let a represent the
improved in some respects but worsened fraction of the total return collected by
in others, for the higher wage would firms. Since G = aG" and G" = C, equa-
make the supply of trainees greater than tion (12) can be written as
the demand, and rationing would be re-
quired. The final step would be to shift 14 Marshall was clearly aware of specific talents
and their effect on wages and productivity: "Thus
some training costs as well as returns to
the head clerk in a business has an acquaintance
employees, thereby bringing supply more with men and things, the use of which he could in
in line with demand. When the final step some cases sell at a high price to rival firms. But in
other cases it is of a kind to be of no value save to the
is completed firms no longer pay all
business in which he already is; and then his depar-
training costs nor do they collect all the ture would perhaps injure it by several times the value
return but they share both with em- of his salary, while probably he could not get half
that salary elsewhere" (op. cit., p. 626). (My italics.)
ployees."4 The shares of each depend on
However, he overstressed the element of indeter-
the relation between quit rates and minacy in these wages ("their earnings are deter-
wages, layoff rates and profits, and on mined ... by a bargain between them and their em-
ployers, the terms of which are theoretically ar-
other factors not discussed here, such as
bitrary" [ibid., fn.]) because he ignored the effect of
the cost of funds, attitudes toward risk, wages on turnover.
and desires for liquidity.'5 15 The rate used to discount costs and returns is
If training were not completely spe- the sum of a (positive) rate measuring the cost of
funds, a (positive or negative) risk premium, and a
cific, productivity would increase in other
liquidity premium that is presumably positive since
firms as well, and the wage that could capital invested in specific training is very illiquid
be received elsewhere would also in- (see the discussion in Sec. IV, C).
MP'+ aC = W + C, (13) of such training is borne entirely by em-

or ployees. Firms are concerned about the
I == P'- (1 - a)C.16 (14) turnover of employees with specific
training, and a premium is offered to re-
Employees pay the same fraction of
duce their turnover because firms pay
costs, 1 - a, as they collect in returns,
part of their training costs.
which generalizes the results obtained
The part of specific training paid by
earlier. For if training were completely
employees has effects similar to those
general, a = o, and equation (14) re-
discussed earlier for general training: it
duces to equation (10); if firms collected
is also paid by a reduction in wages dur-
all the return from training, a = 1, and
ing the training period, tends to make
(14) reduces to MP' = Wo; if 0 < a <
age-earnings profiles steeper and more
1, none of the earlier equations are satis-
concave, etc. The part paid by firms has
factory.
none of these implications, since current
A few major implications of this
or future wages would not be affected.
analysis of specific training are now
Specific, unlike general, training would
developed.
produce certain "external" effects, for
Rational firms pay generally trained
quits would prevent firms from capturing
employees the same wage and specifically
the full return on costs paid by them, and
trained employees a higher wage than
layoffs would do the same to employees.
they could get elsewhere. A reader might
Note, however, that these are external
easily believe the contrary, namely, that
diseconomies imposed on the employees
general training would command a higher
or employers of firms providing the train-
wage relative to alternatives than specific
ing, not external economies accruing to
training does, since, after all, competition
other firms.
for persons with the latter is apt to be
Employees with specific training have
weaker than for those with the former.
less incentive to quit, and firms have less
This view, however, overlooks the fact
incentive to fire them, than employees
that general training raises the wages
with no or general training, which im-
that could be received elsewhere while
plies that quit and layoff rates would be
(completely) specific training does not,
inversely related to the amount of
so a comparison with alternative wages
specific training. Turnover would be least
gives a misleading impression of the
for employees with extremely specific
absolute effect on wages of different types
training and most for those receiving
of training. Moreover, firms are not too
such general training that productivity
concerned about the turnover of em-
was raised less in firms providing the
ployees with general training and have
training than elsewhere. These proposi-
no incentive to offer them a premium
tions are as applicable to the large
above wages elsewhere because the cost
amount of irregular quits and layoffs
16 If G" did not equal C, these equations would that
be continually occur as to the more
slightly more complicated. Suppose, for example,
regular cyclical and secular movements
G" = G + G' = C + -n n > 0 so that the present
value of the total return would be greater than total in turnover; in this section, however,
costs. Then G = aG" = aC + an, and only the more regular movements are
discussed.
MIP'+ aC+ an = W+C,
or Consider a firm that experiences an
W = AP'- [ (1-a )C-anl]. unexpected decline in demand for its
22 GARY S. BECKER
output, the rest of the economy being training might be lost forever. If spe-
unaffected. The marginal product of cifically trained workers were not laid
employees without specific training- off, the firm would lose now because
such as untrained or generally trained marginal product would be less than
employees presumably initially equaled wages but would gain in the future if the
wages, and their employment would be decline in demand proved temporary.
reduced to prevent their marginal pro- There is an incentive, therefore, not to
ductivity from falling below wages. The lay off workers with specific training
marginal product of specifically trained when their marginal product is only
employees initially would have been temporarily below wages, and the larger
greater than wages. A decline in demand a firm's investment the greater the in-
would reduce these marginal products centive not to lay off such workers.
too, but as long as they were reduced by A worker collecting some of the return
less than the initial difference with wages, from specific training would have less in-
firms have no incentive to lay off such centive to find a new job when tempo-
employees. For sunk costs are sunk, and rarily laid off than others would: he does
there is no incentive to lay off employees not want to lose his investment. His be-
whose marginal product is greater than havior while laid off in turn affects his
wages, no matter how unwise it was, in chances of being laid off, for if it were
retrospect, to invest in their training. known that he would not readily take
Thus workers with specific training seem another job, the firm could lay him off
less likely to be laid off as a consequence without much fear of losing its invest-
of a decline in demand than are untrained ment.
or even generally trained workers.'7 The conclusion here can be briefly
If the decline in demand were suf- summarized. When one firm alone ex-
ficiently great so that even the marginal periences an unexpected decline in de-
product of specifically trained workers mand, relatively few workers with spe-
was pushed below wages, would the firm cific training would be laid off, if only
just proceed to lay them off until the because their marginal product were
marginal product was brought into initially greater than their wage. If the de-
equality with wages? To show the danger cline were permanent, all workers would
here, assume that all the cost and return be laid off when their marginal product
from specific training was paid and col- became less than their wage and all those
lected by the firm. Any worker laid off laid off would have to find jobs else-
would try to find a new job, since nothing where. If the decline were temporary,
would bind him to the old one.'8 The specifically trained workers might not be
firm might be hurt if a new job was laid off even though their marginal
found, for the firm's investment in his product were less than their wage be-
cause the firm would suffer if they took
17 A very similar argument is developed by Wal-
other jobs. The likelihood of their taking
ter Oi in "Labor as a Quasi-fixed Factor of Produc-
tion" (unpublished Ph.D. dissertation, University of other jobs would be inversely related,
Chicago). and therefore the likelihood of their
18 Actually one need only assume that the quit being laid off would be positively related,
rate of laid-off workers tends to be significantly to the extent of their own investment in
greater than that of employed workers, if only be-
training.
cause the cost of searching for another job is less for
laid-off workers. The analysis can easily be extended to
cover general declines in demand; sup- -often an extremely powerful one not
pose, for example, a general cyclical de- to quit. At the same time pension plans
cline occurred. Let me assume that wages "insure" firms against quits for they are
are sticky and remain at the initial level. given a lump sum-the non-vested por-
If the decline in business activity were tion of payments-whenever a worker
not sufficient to reduce the marginal quits. Insurance is needed for specifically
product below the wage, workers with trained employees because their turnover
specific training would not be laid off would impose capital losses on firms.
even though others would be, just as be- Firms can discourage such quits by shar-
fore. If the decline reduced marginal ing training costs and the return with
product below wages, only one modifica- employees, but they have less need to
tion in the previous analysis is required. discourage them and would be more
A firm would have a greater incentive to willing to pay for training costs if insur-
lay off specifically trained workers than ance was provided. The effects on the in-
when it alone experiences a decline be- centive to invest in one's employees may
cause laid-off workers would be less likely have been a major stimulus to the de-
to find other jobs when unemployment velopment of pension plans.20
was widespread. In other respects the An effective long-term contract would
implications of a general decline with insure firms against quits, just as pen-
wage rigidity are the same as those of a sions do, and also insure employees
decline in one firm alone. against layoffs. Firms would be more
The discussion has concentrated on willing to pay for all kinds of training-
layoff rates, but the same kind of reason- assuming future wages were set at an ap-
ing shows that a rise in wages elsewhere propriate level-since a contract, in
would cause fewer quits among specifical- effect, converts all training into com-
ly trained workers than among others. pletely specific training. A casual reading
For specifically trained workers initially of history suggests that long-term con-
receive higher wages than are available tracts have, indeed, primarily been a
elsewhere and the wage rise elsewhere means of inducing firms to undertake
would have to be greater than the initial large investments in employees. These
difference before they would consider contracts are seldom used today in the
quitting. Thus both the quit and layoff United States,2' and while they have de-
rate of specifically trained workers would clined in importance over time, they
be relatively low and fluctuate relatively were probably always the exception here
less during business cycles. These are im- largely because courts have considered
portant implications than can be tested them a form of involuntary servitude.
with the data available.
Although quits and layoffs are influ- 19 According to the National Bureau of Economic
Research study of pensions, most plans still have in-
enced by considerations other than in-
complete vesting (see D. Holland's report in A Re-
vestment costs, some of these, such as thespect for Facts: National Bureau of Economic Re-
presence of pension plans, are more search Annual Report [New York: National Bureau
of Economic Research, 1960], pp. 44-46).
strongly related to investments than may
20 In recent years pensions have also been an im-
appear at first blush. A pension plan
portant tax-saving device, which certainly has been
with incomplete vesting privileges'9 a crucial factor in their mushrooming growth.
penalizes employees quitting before re- 21 The military and entertainment industry are
tirement and thus provides an incentive the major exceptions.
24 GARY S. BECKER
Moreover, any enforcible contract could training. The investment approach pro-
at best specify the hours required on a vides a very different interpretation of
job, not the quality of performance. some common phenomena, as can be seen
Since performance can vary widely, un- from the following examples.
happy workers could usually "sabotage" A positive difference between mar-
operations to induce employers to release ginal product and wages is usually said
them from contracts. to be evidence of monopsony power, and
Some training may be useful neither just as the ratio of product price to
in most nor only in a single firm but in a marginal cost has been suggested as a
set of firms defined by product, type of measure of monopoly power, so has the
work, or geographical location. For ex- ratio of marginal product to wages been
ample, carpentry training would raise suggested as a measure of monopsony
productivity primarily in the construc- power. But specific training would also
tion industry, and French legal training make this ratio greater than one. Does
would be ineffective in the United States, the difference between the marginal
with its different language and legal product and the earnings of major-
institutions. Such training would tend league baseball players, for example,
to be paid by trainees, since a single firm measure monopsony power or the return
could not readily collect the return,22 and on a team's investment? Since teams do
in this respect would be the same as gen- spend a great deal on developing players,
eral training. In one respect, however, it some and perhaps most of the difference
is similar to specific training. Workers must be considered a return on invest-
with training "specific" to an industry, ment even were there no uncertainty
occupation, or country are less likely to about the abilities of different players.23
leave that industry, occupation, or coun- Earnings might differ greatly among
try (via migration) than other workers, firms, industries, and countries and yet
so their industrial, occupational, or there may be relatively little worker
country "turnover" would be less than mobility. The usual explanation would
average. The same result is obtained for be that workers were either irrational or
specific training, except that a firm faced with formidable obstacles in mov-
rather than an industry, occupation, or ing. However, if specific24 training were
country is used as the unit of observa- important, differences in earnings would
tion in measuring turnover. An analysis be a misleading estimate of what "mi-
of specific training, therefore, is helpful grants" could receive, and it might be
also in understanding the effects of cer- perfectly rational not to move. For ex-
tain types of "general" training. ample, although French lawyers earn less
Although a discrepancy between mar- than American lawyers, the average
ginal product and wages is frequently French lawyer could not earn the average
taken as evidence of imperfections in the American legal income simply by migrat-
competitive system, it would occur even 23 S. Rottenberg ("The Baseball Players' Labor
in a perfectly competitive environment Market," Journal of Political Economy, June, 1956,
where there is investment in specific p. 254) argues that the strong restrictions on entry of
teams into the major leagues is prima facie evidence
22 Sometimes firms co-operate in paying trainingthat monopsony power is important, but the entry or
costs, especially when training apprentices (see A threat of new leagues, such as have occurred in pro-
Look at Industrial Training in Mercer County, N.J. fessional basketball and football, is a real possibility.
[Washington Bureau of Apprenticeship and Train- 24 Specific, that is, to the firms, industries, or
ing, 19591, p. 3). countries in question.
ing to the United States, for he would offers training in conjunction with the
have to invest in learning English and production of goods. Some schools, like
American law and procedures.25 those for barbers, specialize in one skill,
In extreme types of monopsony, ex- while others, like universities, offer a
emplified by an isolated company town, large and diverse set. Schools and firms
job alternatives for both trained and un- are often substitute sources of particular
trained workers are nil, and all training, skills. The shift that has occurred over
no matter what the nature, would be time in both law and engineering is a
specific to the firm. Monopsony com- measure of this substitution. In acquir-
bined with control of a product or an ing legal skills the shift has been from
occupation (due, say, to anti-pirating apprenticeships in law firms to law
agreements) converts training specific to schools, and in engineering skills from
that product or occupation into firm- on-the-job experience to engineering
specific training. These kinds of monop- schools.27
sony increase the importance of specific Some types of knowledge can be
training and thus the inventive to invest mastered better if simultaneously related
in employees.26 The effect on training of to a practical problem; others require
less extreme monopsony positions is prolonged specialization. That is, there
more difficult to assess. Consider the are complementarities between learning
monopsonist who pays his workers the and work and between learning and time.
best wage available elsewhere. I see no Most training in the construction indus-
reason why training should have a try is apparently still best given on the
systematically different effect on the job, while the training of physicists re-
foregone earnings of his employees than quires a long period of specialized effort.
of those in competitive firms and, there- The development of certain skills re-
fore, no reason why specific training quires both specialization and experience
should be more (or less) important to and can be had partly from firms and
him. But monopsony power as a whole, partly from schools. Physicians receive
including the more extreme manifesta- apprenticeship training as interns and
tions, would appear to increase the im- residents after several years of concen-
portance of specific training and the in- trated instruction in medical schools. Or
centive for firms to invest in human to take an example closer to home, a re-
capital. search economist not only spends many
years in school but also a rather extensive
B. SCHOOLING
apprenticeship in mastering the "art" of
A school can be defined as an institu- empirical and theoretical research. The
tion specializing in the production of complementarity with firms and schools
training, as distinct from a firm that depends in part on the amount of formal-
ized knowledge available- price theory
25 Of course, persons who have not yet invested in
themselves would have an incentive to migrate, and can be formally presented in a course,
this partly explains why young persons migrate while a formal statement of the principles
more than older ones. For a further explanation see
my discussion on p. 38; also see the paper in this
27 State occupational licensing requirements often
Supplement by L. Sjaastad.
permit on-the-job training to be substituted for
26 A relatively large difference between marginal
school training (see S. Rottenberg, "The Economics
product and wages in monopsonies might measure, of Occupational Licensing" [paper given at the Na-
therefore, the combined effect of economic power tional Bureau of Economic Research Conference on
and a relatively large investment in employees. Labor Economics, April, 1960]).
26 GARY S. BECKER
used in gathering and handling empirical haps this is most apparent when a stu-
materials is lacking. dent works in an enterprise controlled by
Training in a new industrial skill is his school, which frequently occurs at
usually first given on the job, since firms many colleges.
tend to be the first to be aware of its Our definition of student net earnings
value, but as demand develops, some of may seem strange since tuition and other
the training shifts to schools. For ex- direct costs are not usually subtracted
ample, engineering skills were initially from "gross" earnings. Note, however,
acquired on the job, and over time that indirect school costs are implicitly
engineering schools have been developed. subtracted, for otherwise earnings would
A student does not work for pay while have to be defined as the sum of observed
in school but may do so "after" or "be- and foregoine earnings, and foregone
fore" school, or during "vacations." His earnings are a major cost of high school,
earnings are usually less than if he were college, and adult schooling. Moreover,
not in school since he cannot work as earnings of on-the-job trainees would be
much or as regularly. The difference be- net of all their costs, including direct
tween what could have been and is "tuition' costs. Consistent accounting,
earned is an important and indirect cost which is particularly important when
of schooling. Tuition, fees, books and comparing earnings of persons trained in
supplies, unusual transportation and school and on the job, would require that
lodging expenses are other, more direct, earnings of students be defined in the
costs. Net earnings can be defined as the same way.28
difference between actual earnings and Regardless of whether all costs or
direct school costs. In symbols, merely indirect costs are subtracted
from potential earnings, schooling would
WT= MP-k, (15)
have the same kind of implications as
where MP is actual marginal product general on-the-job training. Thus school-
(assumed equal to earnings) and k is di- ing would steepen the age-earnings pro-
rect costs. If MPo is the marginal productfile, mix together the income and capital
that could have been received, equation accounts, introduce a negative relative
(15) can be written as between the permanent and current
earnings of young persons, and allow
W = MPo - (MPo - MP + k) for depreciation on human capital. This
(16)
supports our earlier assertion that an
= MPO-C -
analysis of on-the-job training leads to
where C is the sum of direct and foregone general results that apply to other kinds
costs and where net earnings are the of investment in human capital as well.
difference between potential earnings
C. OTHER KNOWLEDGE
and total costs. These relations should be
On-the-job and school training are not
familiar since they are the same as those
the only activities that raise real income
derived for general on-the-job training,
primarily by increasing the knowledge
which suggests that a sharp distinction
at a person's command. Information
between schools and firms is not always
necessary: for some purposes schools can 28 Students often have negative net earnings and
in this respect differ from most on-the-job trainees,
be treated as a special kind of firm and
although at one time many apprentices also had
students as a special kind of trainee. Per- negative earnings.
about the prices charged by different that the direct costs of search, like the
sellers would enable a person to buy from direct costs of schooling, are usually
the cheapest, thereby raising his com- added to consumption rather than de-
mand over resources, or information ducted from earnings. If firms paid costs
about the wages offered by different and collected the return, search would
firms would enable him to work for the have the same implications as on-the-job
firm paying the highest (see Stigler's specific training.
paper in this Supplement, pp. 94-105). Whether workers or firms pay for
In both examples information about the search depends on the effect of a job
economic system, of consumption and change on alternatives: the larger the
production possibilities, is increased as number of alternatives made available
distinct from knowledge of a particular by a change, the larger, not the smaller,
skill. Information about the political or the fraction of costs that have to be paid
social system-the effect of different by workers. Consider a few examples.
parties or social arrangements-could Immigrants to the United States usually
also significantly raise real incomes.29 found many firms that could use their
Let us consider in more detail invest- talents, and these firms should have been
ment in information about employment reluctant to pay the large cost of trans-
opportunities. A better job might be porting workers to the United States. In
found by spending money on employ- fact, immigrants almost always had to
ment agencies and situation-wanted ads, pay their own way. Even the system of
using one's time to examine want ads, contract labor, which we have seen is a
talking to friends and visiting firms, or in means of protecting firms against turn-
Stigler's language by "search." When the over, was singularly unsuccessful in the
new job requires geographical movement, United States and has been infrequently
additional time and resources would be used.3" Firms that are relatively insu-
spent in moving.30 These expenditures lated from competition in the labor
constitute an investment in information market have an incentive to pay the costs
about job opportunities that would yield of workers coming from elsewhere since
a return in the form of higher earnings they have little to worry about in the
than would otherwise have been received. way of competing neighboring firms. In
If workers paid costs and collected the addition, firms would be willing partly to
return, an investment in search would pay for search within a geographical area
have the same implications about age- because some costs-such as an employ-
earnings profiles, depreciation, and the ment agency's fee-would be specific to
like as general on-the-job training and the firm doing the hiring since they must
schooling, although it must be noted be repeated at each job change.
29 The role of political knowledge is systemati-
D. PRODUCTIVE WAGE INCREASES
cally discussed in A. Downs, An Economic Theory of
Democracy (New York: Harper & Bros., 1957), and One way to invest in human capital
more briefly in my "Competition and Democracy,"
Journal of Law and Economics, Vol. I (Fall, 1958). is to improve emotional and physical
30 Studies of large geographical moves-thosehealth.
re- In Western countries today earn-
quiring both a change in employment and consump-
tion-have tended to emphasize the job change more 31 For a careful discussion of the contract-labor
than the consumption change. Presumably money system see C. Erickson, American Industry and the
wages are considered to be more dispersed geographi- European Immigrant, 1860-1885 (Cambridge, Mass.:
cally than prices. Harvard University Press, 1957).
28 GARY S. BECKER
ings are much more closely geared to ductive, some on-the-job investments
knowledge than to strength, but in an would not be undertaken even though
earlier day, and elsewhere still, strength they were very productive by "absolute"
had a significant influence on earnings. standards.
Moreover, emotional health increasingly Before I proceed further, one point
is considered an important determinant needs to be made. The amount invested
of earnings in all parts of the world. outside the job would be related to cur-
Health, like knowledge, can be improved rent earnings only if the capital market
in many ways. A decline in the death rate was very imperfect, for otherwise any
at working ages may improve earning amount of "outside" investment could
prospects by extending the period during be financed with borrowed funds. The
which earnings are received; a better diet analysis assumes, therefore, that the
adds strength and stamina, and thus capital market is extremely imperfect,
earning capacity; or an improvement in earnings and other income being a
working conditions-higher wages, coffee major source of funds.32
breaks, and so on-might affect morale A firm would be willing to pay for in-
and productivity. vestment in human capital made by em-
Firms can invest in the health of employees outside the firm if it could bene-
ployees through medical examinations, fit from the resulting increase in produc-
luncheons, or steering them away from tivity. The only way to pay, however,
activities with high accident and death would be to offer higher wages during
rates. An investment in health that in- the investment period than would have
creased productivity to the same extent been offered since direct loans to employ-
in many firms would be a general in- ees are prohibited by assumption. When
vestment and would have the same effect a firm gives a productive wage increase-
as general training, while an investment that is, an increase that raises productiv-
in health that increased productivity ity-"outside" investments are, as it
more in the firms making them would be were, converted into on-the-job invest-
a specific investment and would have ments. Indeed, such a conversion is a
the same effect as specific training. Of natural way to circumvent imperfections
course, most investments in health in the in the capital market and the resultant
United States are made outside firms, dependence of the amount invested in
in households, hospitals, and medical human capital on the level of wages.
offices. A full analysis of the effect on The discussion can be stated more
earnings of such "outside" investment in formally. Let W represent wages in the
health is beyond the scope of this paper, absence of any investment, and let a
but I would like to discuss a relation be- productive wage increase costing an
tween on-the-job and "outside" human amount C be the only on-the-job invest-
investments that has received much at- ment. Total costs to the firm would be
tention in recent years. 7r = W + C, and since the investment
When on-the-job investments are paid cost is received by employees as higher
by reducing earnings during the invest- wages, 7r would also measure total wages.
ment period, less is available for invest- The cost of on-the-job training is not
ments outside the job in health, better
32Imperfections in the capital market with re-
diet, schooling, and other factors. If these spect to investment in human capital are discussed
"outside" investments were more pro- in Sec. IV, D.
received as higher wages, so this formally store or truck system in nineteenth-cen-

distinguishes a productive wage increase tury Great Britain has been interpreted
from other on-the-job investments. The as partly designed to prevent an exces-
term MP can represent the marginal sive consumption of liquor and other
product of employees when wages equal debilitating commodities.35 The preva-
W, and G the gain to firms from the in- lence of employer paternalism in under-
vestment in higher wages. In full equi- developed countries has been frequently
librium, accepted as evidence of a difference in
temperament between East and West.
MP + G = W + C=r. (17)
An alternative interpretation suggested
Investment would not occur if the firm's by our study is that an increase in con-
gain was nil (G = o), for then total wages sumption has a greater effect on produc-
(r) would equal the marginal product tivity in underdeveloped countries, and
(MP) when there is no investment. that a productivity advance raises profits
We have shown that firms would bene- more there either because firms have
fit more from on-the-job investment the more monopsony power or because the
more specific the productivity effect, the advance is less delayed. In other words
greater their monopsony power, and the "paternalism" may simply be a way of
longer the labor contract; conversely, the investing in the health and welfare of
benefit would be less the more general employees in underdeveloped countries.
the productivity effect, the less their An investment in human capital
monopsony power, and the shorter the would usually steepen age-earnings pro-
labor contract. For example, a wage in- files, lowering reported earnings during
crease spent on a better diet with an im- the investment period and raising them
mediate impact on productivity might later on. But an investment in an in-
well be granted,33 but not one spent on crease in earnings may have precisely the
general education with a very delayed opposite effect, raising reported earnings
impact-." more during the investment period than
The effect of a wage increase on pro- later and thus flattening age-earning
ductivity depends on the way it is spent,
31 Marshall discusses delays of a generation or
which in turn depends on tastes, knowl-
more and notes that profit-maximizing firms in com-
edge, and opportunities. Firms might petitive industries have no incentive to grant such
exert an influence on spending by exhort- wage increases.
"Again, in paying his workpeople high wages and
ing employees to consume good food,
in caring for their happiness and culture, the liberal
housing, and medical care, or even by re- employer confers benefits which do not end with his
quiring purchases of specified items in own generation. For the children of his workpeople
share in them, and grow up stronger in body and in
company stores. Indeed, the company
character than otherwise they would have done. The
price which he has paid for labour will have borne
33 The more rapid the impact the more likely that
the expenses of production of an increased supply of
it comes within the (formal or de facto) contract pe-
high industrial facilities in the next generation: but
riod. Leibenstein apparently initially assumed a rap-
these facilities will be the property of others, who
id impact when discussing wage increases in under-
will have the right to hire them out for the best price
developed countries (see his "The Theory of Under-
they will fetch: neither he nor even his heirs can
employment in Backward Economies," Journal of
reckon on reaping much material reward for this
Political Economy, Vol. LXV [April, 1957]). In a later
part of the good that he has done" (op. cit., p. 566).
comment he argued that the impact might be de-
layed ("Underemployment in Backward Economies: 35 See G. W. Hilton, "The British Truck System
Some Additional Notes," Journal of Political Econ- in the Nineteenth Century," Journal of Political
omny, Vol. LXVI [June, 1958]). Economy, LXV (April, 1957), 246-47.
30 GARY S. BECKER
profiles. The cause of this difference is III. RELATION BETWEEN EARNINGS,
simply that reported earnings during the COSTS, AND RATES OF RETURN
investment period tend to be net of the

Thus far little attention has been paid
cost of general investments and gross of
to the factors determining the amount in-
the cost of a productive earnings in-
vested in human capital. The most im-
crease. 36
portant single determinant is the profit-
The productivity of employees de-
ability or rate of return, but the effect on
pends not only on their ability and the
earnings of a change in the rate of return
amount invested in them both on and off
has been difficult to distinguish em-
the job but also on their motivation, or
pirically from a change in the amount in-
the intensity of their work. Economists
vested. For investment in human capital
have long recognized that motivation in
usually extends over a long and variable
turn partly depends on earnings because
period, so the amount invested can-
of the effect of an increase in earnings on
not be determined from a known "invest-
morale and aspirations. Equation (17),
ment period." Moreover, the discussion
which was developed to show the effect
of on-the-job training clearly indicated
of investments outside the firm financed
that the amount invested is often merged
by an increase in earnings, can also show
with gross earnings into a single net earn-
the effect of an increase in the intensity
ings concept (which is gross earnings
of work "financed" by an increase in
minus the cost or plus the return on in-
earnings. Thus W and MP would show
vestment).
initial earnings and productivity, C the
In the following, some rather general
increase in earnings, and G the gain to
relations between earnings, investment
firms from the increase in productivity costs, and rates of return are derived.
caused by the "morale" effect of the in-
They permit one to distinguish, among
crease in earnings. The incentive to grant
other things, a change in the return from
a morale-boosting increase in earnings,
a change in the amount invested. The
therefore, would depend on the same fac-
discussion proceeds in stages from simple
tors as does the incentive to grant an in-
to complicated situations. First, invest-
crease used for outside investments.
ment is restricted to a single period and
Many recent discussions of wages in un-
returns to all remaining periods; then in-
derdeveloped countries have stressed the vestment is permitted to be distributed
latter,37 while earlier discussions often over a known group of periods called the
stressed the former.38 investment period. Finally, we show how
the rate of return, amount invested, and
36 If E represents reported earnings during the in-
the investment period can all be derived
vestment period and MP the marginal product when
there is no investment, E = MP - C with a general from information on net earnings
investment, E = MP with a specific investment alone.
paid by the firm, and E = MP + C with a pro-
ductive earnings increase. Let Y be an activity providing a per-
son entering at a particular age, called
37 See the papers by Leibenstein, op. cit., and
H. Oshima, "Underdevelopment in Backward Econ- age zero, with a real net earnings stream
omies: An Empirical Comment," Journal of Political
of Yo during the first period, Y1 the next
Economy, Vol. LXVI (June, 1958).
period, and so on until Yn is provided
38 For example, Marshall stressed the effect of an
increase in earnings on the character and habits of
during the last period. The general term
working people (op. cit., pp. 529-32, 566-69). "activity" rather than occupation or
another more concrete term is used to providing a net earning stream of X0,
indicate that any kind of investment in X1, . . . Xn, with a present value of V(X),
human capital is permitted, not just on- the present value of the gain from choos-
the-job training but also schooling, in- ing Y would be given by
formation, health, and morale. By "net"
d= V(Y) - V(X)
earnings I continue to mean that tuition
costs during any period have been sub- E _ _ X_ (19)
tracted and returns added to "gross"

i=o ( 1 +i)i0
earnings during the same period (see dis-
cussion in Sec. II). "Real" earnings are Equation (19) can be reformulated to
the sum of monetary earnings and the bring out explicitly the relation between
monetary equivalent of psychic earnings. costs and returns. The cost of investing
Since many persons appear to believe in human capital equals the net earnings
that the term "investment in human foregone by choosing to invest rather
capital" must be restricted to monetary than choosing an activity requiring no
costs and returns, let me emphasize that investment. If activity Y requires an in-
essentially all my analysis applies inde- vestment only in the initial period and if
pendently of the division of real earn- X does not require any, the cost of
ings into monetary and psychic com- choosing Y rather than X is simply the
ponents. Thus the analysis applies to difference between their net earnings in
health, an activity with a large psychic the initial period, and the total return
component, as well as to on-the-job would be the present value of the differ-
training, an activity with a large mone- ences between net earnings in later
tary component. When psychic compo- periods. If C= Xo- Yo, k= Yj-
nents dominate, the language associated Xj, j = 1,... n, and if R measures the
with consumer durable goods might be total return, the gain from Y could be
considered more appropriate than that written as
associated with investment goods, but to
simplify the presentation, I use invest- n k
ment language throughout. (I (1i)id-C=R-C. (20)
The present value of the net earnings

stream in Y would be The relation between costs and returns
can be derived in a different and, for our
purposes, preferable way by defining the
V( ) = (I Oi)+' (18) internal rate of return,40 which is simply
a rate of discount equating the present
where i is the market discount rate, as-
value of returns to the present value of
sumed for simplicity to be the same in
costs. In other words, the internal rate,
each period. If X were another activity
r, is defined implicitly by the equation
39 Our discussion assumes discrete income flows
and compounding, even though a mathematically 40 A substantial literature has developed on the
more elegant formulation would have continuous difference between the income gain and internal re-
variables, with sums replaced by integrals and dis- turn approaches. See, for example, Friedrich and
count rates by continuous compounding. The dis- Vera Lutz, The Theory of Investment of the Firm
crete approach is, however, easier to follow and yet (Princeton, N.J.: Princeton University Press, 1951),
yields the same kind of results as the continuous ap- chap. ii, and the articles in The Management of Cor-
proach. Extensions to the continuous case are porate Capital, ed. Ezra Solomon (Glencoe, Ill.: Free
straightforward. Press, 1959).
32 GARY S. BECKER
equate total costs and returns. In sym-

C +(1-+= r)i' (21) bols,
Cl= X. - V. O , . I
which clearly implies
rn-1
n Yj Xj C1 = O( I+ r)-,
(t
I
+-r)i+1
- Y
(1 + r)i+1 (22) 0
and
= d=o ,
C1C1
k I 1- I +=r ) 711,--ff
(?r In-_ 24 . (2 4)
since C = XO - YO and kj = Y- Xi.
So the internal rate is also a rate of dis-
count equating the present values of net If m = 1, this reduces to equation (23).
earnings. These equations would be con- Two serious drawbacks mar this ap-
siderably simplified if the return were the pealing straightforward approach. The
same in each period, or Yj = Xj + k, estimate of total costs requires a priori
j = 1, . . . n. Thus equation (21) would knowledge and specification of the in-
become vestment period. While the period
covered by formal schooling is easily de-
C = k l(+ r)-n] X (2 3)
r termined, the period covered by much
on-the-job training and other investment
where (1 + r)-Y is a correction for the
is not, and a serious error might result
finiteness of life that tends toward zero
from an incorrect specification: to take
as people live longer.
an extreme example, total costs would
If investment is restricted to a single
approach zero as the investment period
known period, cost and rate of return are
is assumed to be longer and longer.4"
easily determined from information on
A second difficulty is that the differ-
net earnings alone. Since, however, in-
ences between net earnings in X and Y
vestment in human capital is distributed
do not correctly measure the cost of in-
over many periods formal schooling is
vesting in Y since they do not correctly
usually more than ten years in the United
measure earnings foregone. A person
States, and long periods of on-the-job
who invested in the initial period could
training are also common the analysis
receive more than X1 in period 1 as long
must be generalized to cover distributed
as the initial investment yielded a posi-
investment. The definition of an internal
tive return.42 The true cost of an invest-
rate in terms of the present value of net
earnings in different activities obviously 41 Since
applies regardless of the amount and m-1
duration of investment, but the defini- C1= E(Xj- Y-)(I+ O-i,

0
tion in terms of costs and returns is not
generalized so readily. If investment n-1
were known to occur in Y during each liM C, (X X- IYj) (I + r ) i=0,

0
of the first m periods, a simple and super-

by definition of the internal rate.
ficially appealing approach would be to
42 If CO was the initial investment, ro its internal
define the investment cost in each of
rate, and if the return were the same in all years, the
these periods as the difference between amount
net earnings in X and Y, total invest-

Xc 1 = Xi d 1. ro o
ment costs as the present value of these
differences, and the internal rate would could be received in period 1.
ment in period 1 would be the total age of the rates of return on the indi-
earnings foregone, or the difference be- vidual investments. Any sequence of
tween what could have been received internal rates or investment costs is per-
and what is received. The difference mitted, no matter what the pattern of
between X1 and Y1 could greatly rises and declines, nor what form the
underestimate true costs; indeed, Yj investments take, be they a college edu-
might be greater than X1 even though a cation, an apprenticeship, ballet les-
large investment was made in period 1.43 sons, or a medical examination. Differ-
In general, therefore, the amount in- ent investment programs would have
vested in any period would be deter- the same ultimate effect on earnings
mined not only from net earnings in the whenever the average rate of return and
same period but also from net earnings in the sum of investment costs were the
earlier periods. same.45
If the cost of an investment is con- Equation (25) can be given an inter-
sistently defined as the earnings fore- esting interpretation if all rates of return
gone, quite different estimates of total were the same. The term k r would then
costs emerge. Although superficially a be the value at the beginning of the mth
less natural and straightforward ap- period of all succeeding net earning dif-
proach, the generalization from a single ferentials between Y and X discounted
period to distributed investment is
4 Y1 is greater than XI if
actually greatly simplified. So let Cj be
the foregone earnings in the jth period,
rj the rate of return on Cj, and let the O l-1+ ro) -n-1
return per period on Cj be a constant or if
kj, with k = Ykj being the total return

on the whole investment. If the number 1 ( + ro)
of periods was indefinitely large, and if where C1 is the investment in period 1.
investment occurred only in the first m
44 A proof is straightforward. An investment in
periods, the equation relating costs, re-
period j would yield a return of the amount ki =
turns, and internal rates has the striking- rjCi in each succeeding period if the number of peri-
ly simple form of44 ods was infinite and the return was the same in each.
Since the total return is the sum of individual re-
m-1
turns,
C= lCj=*, (25) m-1 m-1 m-1 r~

k= Elkj= E rjCj= CE C= rC
0
where 0 0 0 C
m-1 Cj I am indebted to Helen Raffel for important sugges-

r= Wj c tions which led to this simple proof.
0
45 Note that the rate of return equating the pres-
and ent values of net earnings in X and Y is not neces-
m-1
sarily equal to r, for it would weigh more heavily
than r does the rates of return on earlier investments.
Ew1=1. (26) For example, if rates were higher on investments in
0
earlier than later periods, the over-all rate would be
greater than i, and vice versa if rates were higher in
Total-cost, efined simply as the sum later periods. The difference between the over-all in-
of cost during each period, would equal ternal rate for X and Y and r would be small, how-
ever, as long as the investment period was not very
the capitalized value of returns, the rate
long and the systematic difference between internal
of capitalization being a weighted aver- rates not very great.
34 GARY S. BECKER
at the internal rate, r.46 Total costs would amount invested in each of the first m
equal the value also at the beginning of periods could be estimated from the net
the mth period-which is the end of the earnings streams in X and Y alone if the
investment period-of the first m differ- rate of return was the same on all invest-
entials between X and Y.47 The value of ments. For the internal rate r could be
the first m differentials between X and Y determined from the condition that the
must equal the value of all succeeding present value of net earnings must be the
differentials between Y and X, since r same in X and Y, and the amount in-
would be the rate of return equating the vested in each period seriatim from the
present values in X and Y. relations48
The internal rate of return and the
CO=XO-YO, Ci=X1-Yi+rCo
46 That is,
i-l (27)
co
Cj=Xj-Yj+rJ k =o
E ( Yj - Xj) ( + r)r-1-j
j=rn
So costs and the rate of return can be

= kE (1 + r)m-l=cestimated
k. from information on net earn-
r
ings. This is fortunate since the return
47 Since, by definition,
on human capital is never empirically
separated from other earnings and the
Xo-Yo= Co , X1-Y1 = Cl-rCo, cost of such capital is only sometimes and
and more generally incompletely separated.
j-1 The investment period of education
X- Yj=Cj- r ECh O j< m, can be measured by years of schooling,
k-o
but the period of on-the-job training, the
then
search for information, and other invest-
m-1 ments is not readily available. Happily,
(Xi- Yj)(I + r)n-1-i one need not know the investment
i=o
period to estimate costs and returns,
m-1 - since all three can be simultaneously
- (c- r Ci)(k + r)m11 estimated from information on net earn-
i=0 0
ings. If activity X were known to have
m-1 no investment (a zero investment period)
= ICj{ (I1+ r)m-1-i-rthe
[1 amount invested in Y during any
0
period would be defined by
+ (1+ r) + . + (1+ r )M-2-i]
48 If the rate of return was not the same on all in-
m-I vestments there would be 2m unknowns-CO,...
= Ecj=c. Cm - 1, and ro, . . . rmo-and only m + 1 equa-
0 tions-the m cost definitions and the equation
The analytical difference between the naive defi- m-1
nition of costs advanced earlier and one in terms of k = E risCi

foregone earnings is that the former measures total 0
costs by the value of earning differentials at the be-

ginning of the investment period and the latter by An additional m - 1 relation would be required to
the value at the end of the period. Therefore, C' = determine the 2rn unknowns. The condition ro =
C(1 + r)'-m, which follows from eq. (24) when n = r = . .. = rm-1 is one form these n - 1 relations
co can take.
j-1 definition is the same: investment occurs

Cj=Xj-Yj+r E Ck, allj, (28) in Y whenever earnings there are below
0
the sum of those in X and the income ac-
and total costs by cruing on prior investments. If costs were
co
found to be greater than zero before
some period m and equal to zero there-
C = 1Cj3. (29)0 after, the first m periods would be the
empirically derived investment period.
The internal rate could be determined in But costs and returns can be estimated
the usual way from the equality between from equation (28) even when there is no
present values in X and Y, costs in each simple investment period.
period from equation (28) and total costs A common objection to an earlier
from equation (29). draft of this paper is that the general and
The definition of costs presented here rather formal definition of costs ad-
simply extends to all periods the defini- vanced here is all right when applied to
tion advanced earlier for the investment on-the-job training, schooling, and other
period.49 The rationale for the general recognized investments, but goes too far
by also including as investment costs
49 Therefore, since the value of the first mn earning
differentials has been shown to equal many effects that should be treated
otherwise. For example, the protest runs,
m-1
ECj suppose that learning was essentially un-

() avoidable in an activity Z, so that earn-
ings "automatically" grows rapidly with
at period m (see n. 47), total costs could be esti-
mated from the value of all differentials at the end of
experience. Since earnings in Z would
the earning period. That is, tend to be lower than those in X at
00 00
younger ages and higher later on, my
approach would say that investment oc-
C= E Cj =E (xi- Yj) C-1-i.
0 0 curs in Z. Critics have argued that there
really is no investment in Z since the rise
Thus the value of all differentials would equal zero
at the beginning of the earning period-by definition in earnings results from unavoidable learn-
of the internal rate-and C at the end. The apparent ing rather than from an attempt to im-
paradox results from the infinite horizon, as can be prove skills, knowledge, or health. Al-
seen from the following equation relating the value
of the first differentials at the beginning of the gth though the argument is superficially
period to costs: plausible I am convinced it is as reason-
able to say that investment in human
f_1
capital occurs in Z as in activities requir-
V( f, g) = A (Xj- Yj)(1 + r)o-1-i
i=0 ing training or schooling. Indeed, an im-
portant virtue rather than defect in my
f-1
concept of human capital is that learn-
= YCj(l + r)-f- ing-both on and off the job-is included
i=0
along with training and schooling.

Whenf = co and g = 0, V = 0, but wheneverf = g,
If Z were preferred to X the higher
f-1 earnings at later ages presumably out-
V= Ec weigh the earnings foregone initially.
0
Similarly, a person entering an activity
In particular, if f = g =CO) V = C. requiring much education is said to value
36 GARY S. BECKER
the stream of future higher earnings that the total investment in any activity
more than the net earnings foregone can be determined?
initially. If the lower earnings due to The statement "nothing is invested in
education are called investment costs, an activity" means only nothing would
the higher earnings investment returns, be invested after the age when informa-
and if costs are related to returns by an tion on earnings first became available;
internal rate of return, logical consist- investment can have occurred before
ency and economic sense would require that age. If, for example, the data begin
that similar concepts apply to learning. at age eighteen, some investment in
Thus the lower initial earnings of high- schooling, health, or information surely
school graduates who enter occupations must have occurred at younger ages. The
"with a future" have as much right to be earning stream of persons who do not
considered investment, both from the invest after age eighteen would have to
social and private viewpoints, as do the be considered, at least in part, as a return
lower net earnings of those enrolled in on the investment before eighteen. In-
college. In general, since the private and deed, in the developmental approach to
social ranking of different economic ac- child-rearing (discussed in Selma Mush-
tivities depend only on their net earning kin's paper), most if not all of these earn-
streams, if one activity was said to re- ings would be so considered.
quire a given investment and to yield a The earning stream in an activity
given return, another activity with the with no investment beyond the initial
same net earning stream must be said to age (activity X) would be flat if the de-
require the same investment and yield velopmental approach was followed and
the same return, no matter how they earnings were said to result entirely from
differ in other respects. earlier investment." The minimum in-
So much in defense of our approach. vestment could then be determined if an
To estimate costs empirically still has assumption was made about its rate to
required a priori knowledge that nothing return. My discussion of the shape of the
is invested in activity X. Without such earning stream in X is, however, highly
knowledge, only the difference between conjectural,52 and further investigation
the amounts invested in any two activi- may well indicate that another approach
ties with known net earning streams is preferable.
could be estimated from the definitions Our assumption that lifetimes are
in equation (28). Were this done for all infinite, although descriptively unreal-
available streams the investment in any istic, is often a very close approximation.
activity beyond that in the activity with For example, I have shown elsewhere
the smallest investment could be deter- that the average rate of return on college
mined.50 The observed minimum invest- education in the United States could
ment would not be zero, however, if the
rate of return on some initial investment 51 If C measured the cost of investment before the
initial age and r its rate of return, k = rC would
was sufficiently high to attract everyone. measure the return per period. If earnings were at-
A relevant question is, therefore: can the tributed entirely to this investment, Xi = k = rC,
shape of the stream in an activity having where Xi represents earnings at the ith period past
the initial age.
zero investment be specified a priori so
52 But note that empirical evidence indicates that
50 The technique is applied and further developed age-earning profiles in unskilled occupations are very
by Mincer in his paper in this Supplement. flat.
only be slightly raised if people remained amount invested can be distinguished

in the labor force indefinitely. A finite empirically from the effect of a change in
earning period has, however, a greater rates of return. Our attention now turns
effect on the rate of return of invest- to the factors influencing the amount in-
ments occurring at later ages, say after vested in different activities and by dif-
age forty; indeed, it helps explain why ferent persons. Economists have long
schooling and other investments are pri- believed that the incentive to expand and
marily made at younger ages. improve physical resources depends on
An analysis of finite earning streams the rate of return expected. They have
can be approached in two ways. One been very reluctant, however, to inter-
simply applies the concepts developed pret improvements in the effectiveness
for infinite streams and says there is dis- and amount of human resources in the
investment in human capital when net same way, namely, as systematic re-
earnings are above the amount that could sponses or "investments" resulting in
be maintained indefinitely. Investment good part from the returns expected. In
at younger ages would give way to dis- this section I try to show that an invest-
investment at older ages until no human ment approach to human resources is a
capital remained at death (or retire- powerful and simple tool capable of ex-
ment). This approach has several im- plaining a wide range of phenomena, in-
portant applications and is used in parts cluding much that has either been ig-
of my study. An alternative that is more nored or given ad hoc interpretations.
useful for some purposes lets the earning An increase in the lifespan of an activ-
period itself influence the definitions of ity would, other things the same, increase
accrued income and cost. The income re- the rate of return on the investment
sulting from an investment during made in any period. The influence of
period j would be defined as lifespan on the rate of return and thus on
the incentive to invest is important and
r3C1
k - = -_(l+,j~j X (30) takes many forms. A few of these forms
will now be discussed.
The number of periods is obviously
where n + 1 is the earning period, and
affected by mortality and morbidity
the amount invested during j would be
rates, for the lower they are, the longer
defined by
the expected lifespan, and the larger the
Cj= Xj- Yj fraction of a lifetime that can be spent
k=j-l rkCk (31) at any activity. The major secular de-
? o 1 (1 + rk+)k-n cline of these rates in the United States
and elsewhere may have increased the
rates of return on investment in human
IV. THE INCENTIVE TO INVEST
capital,53 thereby encouraging such in-
A. NUMBER OF PERIODS
vestment. This conclusion is independent
The discussion summarized in equa-
53I say may because rates of return are adversely
tions (28) and (31) shows how total costs, affected by the increase in labor force that would re-
rates of return, and the investment sult from a decline in death and sickness. If the ad-
period can be estimated from informa- verse effect was sufficiently great, a decline in death
and sickness would reduce rates of return on human
tion on net earnings alone, and thus how
capital. I am indebted to my wife for emphasizingI-
the effect on earnings of a change in the this point.
38 GARY S. BECKER
of whether the secular improvement in for p periods only, where p varied be-
health itself resulted from investment; if tween o and n. The size of p would be
so, the secular increase in rates of return affected by many factors, including the
would be part of the return to invest- rate of obsolescence since the more
ment in health. rapidly an investment became obsolete
A relatively large fraction of younger the smaller p would be. The advantage in
persons are in school, enter upon on-the- being young would be less the smaller p
job training, change jobs and locations, was, since the effect of age on the rate of
and add to their knowledge of economic, return would be positively related to p.
political, and social opportunities. The For example, if p equaled two years, the
entire explanation of these differences rate would be the same at all ages except
between young and old persons may not the two nearest the "retirement" age. If
be that the young are more interested in the investment approach was correct, the
learning, more able to absorb new ideas, difference between the amount invested
less tied down by family responsibilities, at different ages would be positively cor-
more easily supported by parents, or related with p, which is not surprising
more flexible about changing their routine since an expenditure with a small p
and place of living. One need not rely would be less of an "investment" than
only on life-cycle effects on capabilities, one with a large p, and arguments based
responsibilities, or attitudes as soon as on an investment framework would be
one recognizes, as we have throughout, less applicable. None of the life-cycle
that schooling, training, mobility, and arguments seem to imply any correlation
the like are ways to invest in human with p, so this provides a powerful test of
capital and that younger people have a the importance of the investment ap-
greater incentive to invest because they proach.
can collect the return over more years. The time spent in any one activity is
Indeed, a greater incentive would be determined not only by age, mortality,
present even if age had no effect on and morbidity but also by the amount of
capabilities, responsibilities, and atti- switching between activities. Women
tudes. spend less time in the labor force than
Although the unification of these dif- men and, therefore, have less incentive
ferent kinds of behavior by the invest- to invest in market skills; tourists spend
ment approach is important evidence in little time in any one area and have less
its favor, other evidence is needed. A incentive than residents of the area to
powerful test can be developed along the invest in knowledge of specific consump-
following lines.55 Suppose that invest- tion opportunities;56 temporary migrants
ment in human capital raised earnings to urban areas have less incentive to in-
vest in urban skills than permanent resi-
54 Younger persons would also have a greater in-
centive to invest if the cost of any investment rose dents; and, as a final example, draftees
with age, say, because potential and thus foregone have less incentive than professional
earnings rose with age.
soldiers to invest in purely military
"5 This test was suggested by George Stigler's dis-
skills.
cussion of the effect of different auto-correlation pat-
terns on the incentive to invest in information (see Women, tourists, and the like have to
"The Economics of Information," Journal of Politi-
cal Economy, Vol. LXIX [June, 19611, and his paper 56 This example is from Stigler, "The Economics
in this Supplement). of Information," op. cit.
find investments that increase productiv- B. WAGE DIFFERENTIAlS AND
SECULAR CHANGES
ity in several activities. A womann wants
her investment to be useful both as a According to equation (30) the in-
housewife and as a participant in the la- ternal rate of return depends on the ratio
bor force, or a frequent traveler wants of the return per unit time to investment
to be knowledgeable in many environ- costs. A change in the return and costs by
ments. Such investments would be less the same percentage would not change
readily available than more specialized the internal rate, while a greater per-
ones-after all, an investment increasing centage change in the return would
productivity in two activities also in- change the internal rate in the same di-
creases it in either one alone, extreme rection. The return is measured by the
complementarity aside, while the con- absolute income gain, or by the absolute
verse does not hold; specialists, therefore, income difference between persons differ-
have greater incentive to invest in them- ing only in the amount of their invest-
selves than others do. ment. Note that absolute, not relative,
Specialization in an activity would be income differences determine the return
discouraged if the market were very and the internal rate.
limited; thus the incentive to specialize Occupational and educational wage
and to invest in oneself would increase as differentials are sometimes measured by
the extent of the market increased. relative, sometimes by absolute, wage
Workers would be more skilled the larger differences,59 although no one has ade-
the market, not only because "practice quately discussed their relative merits.
makes perfect," so often stressed in dis- Marginal productivity analysis relates
cussions of the division of labor,57 but the derived demand for any class of
also because a larger market would in- workers to the ratio of their wages to
duce a greater investment in skills.58 Put those of other inputs,60 so wage ratios are
differently, the usual analysis of the more appropriate in understanding forces
division of labor stresses that efficiency, determining demand. They are not, how-
and thus wage rates, would be greater the ever, the best measure of forces determin-
larger the market, and ignores the po- ing supply, for the return on investment
tential earnings period in any activity, in skills and other knowledge is deter-
while ours stresses that this period, and mined by absolute wage differences.
thus the incentive to become more ef-
ficient, would be directly related to 59 See A. M. Ross and W. Goldner, "Forces Af-
fecting the Inter-industry Wage Structure," Quar-
market size. Surprisingly little attention terly Journal of Economics, Vol. LXIV (May, 1950);
has been paid to the influence of market P. H. Bell, "Cyclical Variation and Trend in Occupa-
size on the incentive to invest in skills. tional Wage Differentials in American Industry
since 1914," Review of Economics and Statistics, Vol.
57 See, for example, Marshall, op. cit., Bk. IV, XXIII (November, 1951); F. Meyers and R. L.
chap. ix. Bowlby, "The Interindustry Wage Structure and
Productivity," Industrial and Labor Relations Re-
58 If "practice makes perfect" means that age-
view, Vol. VII (October, 1953); Stigler and Blank,
earnings profiles slope upward, then according to my
op. cit., Table 11; P. Keat, "Long-Term Trends in
approach it must be treated along with other kinds
Occupational Wage Differentials," Journal of Politi-
of learning as a way of investing in human capital.
cal Economy, Vol. LXVIII (December, 1960).
The distinction above between the effect of an in-
crease in the market on practice and on the incentive 60 Thus the elasticity of a substitution is usually
to invest would simply be that the incentive to invest defined as the percentage change in the ratio of quan-
in human capital is increased even aside from the tities employed per 1 per cent change in the ratio of
effect of practice on earnings. wages.
40 GARY S. BECKER
Therefore neither wage ratios nor wage interpretation suggested by our analysis
differences are uniformly the best meas- is that the spread of education and the
ure, ratios being more appropriate in de- increased investment in other kinds of
mand studies and differences in supply human capital were in large part induced
studies. by technological progress (and perhaps
The importance of distinguishing be- other changes) through the effect on the
tween wage ratios and differences, and rate of return, as measured by wage dif-
the confusion resulting from the practice frences and costs. Clearly a secular de-
of using ratios to measure supply as well cline in wage ratios is not inconsistent
as demand forces, can be illustrated by with a secular increase in real wage dif-
considering the effects of technological ferences if average wages were rising,
progress. If progress were uniform in all and, indeed, one important body of data
industries and neutral with respect to all on wages shows a decline in ratios and an
factors, and if there were constant costs, even stronger rise in differences.62
initially all wages would rise by the same The interpretation based on auton-
proportion and the prices of all goods, omous supply shifts has been favored
including the output of industries sup- partly because a decline in wage ratios
plying the investment in human capital,6" has erroneously been taken as evidence
would be unchanged. Since wage ratios of a decline in the return to skill. While a
would be unchanged, firms would have decision ultimately can be based only on
no incentive initially to alter their factor a detailed re-examination of the evi-
proportions. Wage differences, on the dence,63 the induced approach can be
other hand, would rise at the same rate made more plausible by considering
as wages, and since investment costs trends in physical capital. Economists
would be unchanged, there would be an have been aware that the rate of return
incentive to invest more in human capi- on capital could be rising or at least not
tal, and thus to increase the relative falling while the ratio of the "rental"
supply of skilled persons. The increased price of capital to wages was falling.
supply would in turn reduce the rate of Consequently, although the rental price
increase of wage differences and produce 62 Keat's data for 1906-53 in the United States
an absolute narrowing of wage ratios. show both an average annual decline of 0.8 per cent
In the United States during much of in the coefficient of variation of wages and an aver-
age annual rise of 1.2 per cent in the real standard
the last eighty years, a narrowing of wage deviation. The decline in the coefficient of variation
ratios has gone hand in hand with an in- was shown in his study (op. cit); I computed the
creasing relative supply of skill, an as- change in the real standard deviation from data
made available to me by Keat.
sociation that is usually said to result
63 For those believing that the evidence over-
from the effect of an autonomous increase whelmingly indicates a secular decline in rates of re-
in the supply of skills-brought about turn on human capital, I reproduce Adam Smith's
statement on earnings in some professions. "The lot-
by the spread of free education or the
tery of the law, therefore, is very far from being a
rise in incomes-on the return to skill, as perfectly fair lottery; and that, as well as many
measured by wage ratios. An alternative other liberal and honourable professions, is, in point
of pecuniary gain, evidently under-recompensed"
61 Some persons have argued that only direct in- (The Wealth of Nations [New York: Modern Library,
vestment costs would be unchanged, indirect costs or 1937], p. 106). Since economists tend to believe that
foregone earnings rising along with wages. Neutral law and most other liberal professions are now over-
progress implies, however, the same increase in the compensated relative to non-professional work "in
productivity of a student's time as in his teacher's point of pecuniary gain," the return to professional
time or in the use of raw materials, so even foregone work could not have declined continuously if Smith's
earnings would not change. observations were accurate.
of capital declined relative to wages over culations I have made suggest that there
time, the large secular increase in the is much uncertainty about the return to
amount of physical capital per man-hour human capital.64 The response to un-
is not usually considered autonomous, certainty is determined by its amount
but rather induced by technological and and nature and by tastes or attitudes.
other developments that, at least tempo- Many have argued that attitudes of in-
rarily, raised the return. A common ex- vestors in human capital are very differ-
planation based on the effects of eco- ent from those of investors in physical
nomic progress may, then, account for capital because the former tend to be
the increase in both human and physical younger, 65 and young persons are sup-
capital. posed to be especially prone to overesti-
mate their ability and chance of good
C. RISK AND LIQUIDITY
fortune.66 Were this view correct, a hu-
An informed, rational person would in- man investment which promised a large
vest only if the expected rate of return return to exceptionally able or lucky
was greater than the sum of the interest persons would be more attractive than a
rate on riskless assets and the liquidity similar physical investment. However, a
and risk premiums associated with the in- "life-cycle" explanation of attitudes to-
vestment. Not much need be said about ward risk may be no more valid or nec-
the "pure" interest rate, but a few words essary than life-cycle explanations of
are in order on risk and liquidity. Since why investors in human capital are rela-
human capital is a very illiquid asset-it tively young (discussed on pp. 37-38).
cannot be sold and is rather poor col- Indeed, an alternative explanation of re-
lateral on loans a positive liquidity actions to large gains has already ap-
premium, perhaps a sizable one, would peared.67
be associated with such capital. 64 For example, Marshall said: "Not much less
The actual return on human capital than a generation elapses between the choice by par-
varies around the expected return be- ents of a skilled trade for one of their children, and
his reaping the full results of their choice. And mean-
cause of uncertainty about several fac- while the character of the trade may have been al-
tors. There always has been considerable most revolutionized by changes, on which some
uncertainty about the length of life, one probably threw long shadows before them, but
others were such as could not have been foreseen
important determinant of the return. even by the shrewdest persons and those best ac-
People are also uncertain about their quainted with the circumstances of the trade" (op.
ability, especially younger persons who cit., p. 571), and "the circumstances by which
the earnings are determined are less capable of being
do most of the investing. In addition, foreseen [than those for machinery]" (ibid.).
there is uncertainty about the return to 65 Note that our argument on p. 38 implied that
a person of given age and ability because investors in human capital would be younger.
of numerous events that are not predict- 66 Smith said: "The contempt of risk and the pre-
sumptuous hope of success, are in no period of life
able. The long time required to collect more active than at the age at which young people
the return on an investment in human choose their professions" (op. cit., p. 109). Marshall
capital reduces the knowledge available, said that "young men of an adventurous disposition
are more attracted by the prospects of a great suc-
for required is knowledge about the en-
cess than they are deterred by the fear of failure"
vironment when the return is to be re- (op. cit., p. 554).
ceived, and the longer the average period 67 See M. Friedman and L. J. Savage, "The Util-
between investment and return the less ity Analysis of Choices Involving Risk," reprinted in
Readings in Price Theory, ed. G. J. Stigler and
such knowledge is available. K. Boulding (Chicago: Richard D. Irwin, Inc.,
Informed observation as well as cal- 1952).
42 GARY S. BECKER
D. CAPITAL MARKETS AND KNOWLEDGE 1. Since large expenditures would be

If investment decisions respond only more difficult to finance, investment in
to earning prospects, adjusted for risk (say) a college education would be more
and liquidity, the adjusted marginal rate affected than in (say) short-term migra-
of return would be the same on all in- tion.
vestments. The rate of return on educa- 2. Internal financing would be com-
tion, training, migration, health, and mon, and consequently wealthier families
other human capital is supposed to be would tend to invest more than poorer
higher than elsewhere, however, because ones.
of financing difficulties and inadequate 3. Since employees' specific skills are
knowledge of opportunities. These will part of the intangible assets or good will
now be discussed briefly. of firms and can be offered as collateral
Economists have long emphasized that along with tangible assets, capital would
it is difficult to borrow funds to invest in be more readily available for specific
human capital because such capital can- than for general investments.
not be offered as collateral and courts 4. Some persons have argued that op-
have frowned on contracts which even portunity costs (foregone earnings) are
indirectly suggest involuntary servitude. more readily financed than direct costs
This argument has been explicitly used because they require only to do "with-
to explain the "apparent" underinvest- out," while the latter require outlays.
ment in education and training and also, Although superficially plausible, this
although somewhat less explicitly, under- view can easily be shown to be wrong:
investment in health, migration, and opportunity and direct costs can be
other human capital. The importance at- financed equally readily, given the state
tached to capital market difficulties can of the capital market. If total invest-
be determined not only from the discus- ment costs were $800, potential earnings
sions of investment but also from the $1,000, and if all costs were foregone
discussions of consumption. Young per- earnings, investors would have $200 of
sons would consume relatively little, earnings to spend; if all were direct costs,
productivity and wages might be related, they would initially have $1,000 to
and some other consumption patterns spend, but just $200 would remain after
would follow only if it were difficult to paying "tuition," so their net position
capitalize future earning power. Indeed, would be exactly the same as before. The
unless capital limitations applied to con- example can be readily generalized and
sumption as well as investment, the latter the obvious inference is that indirect and
could be indirectly financed with "con- direct investment costs are equivalent in
sumption" loans."8 imperfect as well as perfect capital
Some other implications of capital markets.
market difficulties can also be men- While it is undeniably difficult to use
tionrdl the capital market to finance invest-
ments in human capital, there is some
68 A person with an income of X and investment
costs of Y (1 < X) could either use X for consuinp-
reason to doubt whether otherwise
tion and receive an iinvestmient loan of Y. or use equivalent investments in physical capi-
X - Y for consumption, Y for investment, and re- tal can be financed much more easily.
ceive a consulmption loan of Y. He ends up with the
same consumption and investment in both cases, the
Consider an eighteen-year-old who wants
only difference being in the names attached to loans. to invest a given amount in equipment
for a firm he is starting rather than in a The eighteen-year-old in our example

college education. What is his chance of who could not finance a purchase of
borrowing the whole amount at a "mod- machinery might, without too much cost,
erate" interest rate? Very slight, I be- postpone the investment for a number of
lieve, since he would be untried and have years until his reputation and equity
a high debt equity ratio; moreover, the were sufficient to provide the "personal"
collateral provided by his equipment collateral required to borrow funds.
would probably be very imperfect. He, Financing may prove a more formidable
too, would either have to borrow at high obstacle to investors in human capital
interest rates or self-finance. Although because they cannot postpone their in-
the difficulties of financing investments vestment so readily. Perhaps this ac-
in human capital have usually been re- counts for the tendency of economists to
stress capital market imperfections when
lated to special properties of human
discussing investments in human capital.
capital, in large measure they seem also
to beset comparable investments in
V. SOME EFFECTS OF HUMAN CAPITAL
physical capital.
A. EXAMPLES
A recurring theme is that young per-
sons are especially prone to be ignorant Differences in earnings among persons,
of their abilities and of the investment areas, or time periods are usually said to
opportunities available. If so, investors result from differences in physical capi-
in human capital, being younger, would tal, technological knowledge, ability, or
be less aware of opportunities and thus institutions (such as unionization or
more likely to err than investors in socialized production). Our analysis indi-
tangible capital. I suggested earlier (pp. cates, however, that investment in hu-
man capital also has an important effect
37-38) that investors in human capital
on observed earnings because earnings
are younger partly because of the cost in
tend to be net of investment costs and
postponing their investment to older
gross of investment returns. Indeed, an
ages. The desire to acquire additional
appreciation of the direct and indirect
knowledge about the return and about
importance of human capital appears to
alternatives provides an incentive to
resolve many otherwise puzzling em-
postpone any risky investment, but since
pirical findings about earnings. Consider
an investment in human capital is more the following examples:
costly to postpone, it would be made l. Almost all studies show that age-
earlier and presumably with less knowl- earnings profiles tend to be steeper among
edge than comparable non-human in- more skilled and educated persons. I
vestments. Therefore, investors in hu- argued earlier (pp. 14-15) that on-the-
man capital may not have less knowledge job training would steepen age-earning
because of their age; rather both might be profiles and the analysis of Section III
a joint product of the incentive not to generalizes the argument to all human
delay investing.69 capital. Since observed earnings are
gross of returns and net of costs, invest-
69 Marshall (op. cit., pp. 571-73) appears to argue ment in human capital at younger ages
that it is also intrinsically more difficult to acquire
would reduce observed earnings then and
knowledge about the return from an investment in
human capital. raise them at older ages, thus steepening
44 GARY S. BECKER
the age-earnings profile]0 to estimate empirically the elasticity of

2. In recent years students of inter- substitution between capital and labor.
national trade theory have been some- Usually a ratio of the input of physical
what shaken by findings that the United capital to the input of labor is regressed
States, said to have relative scarcity of on the wage rate in different areas or time
labor and abundance of capital, ap- periods, the regression coefficient being
parently exports relatively labor-inten- an estimate of the elasticity of substitu-
sive commodities and imports relatively tion.73 Countries, states, or time periods
capital-intensive commodities. For ex- that have relatively high wages and in-
ample, one study found that export puts of physical capital also tend to have
industries pay higher wages than import much human capital. Just as a correla-
competing ones.7" tion between wages, physical capital and
An interpretation consistent with the human capital seems to obscure the rela-
Ohlin-Heckscher emphasis on the relationship between relative factor supplies
tive abundance of different factors argues and commodity prices, so it obscures the
that the United States has an even more relationship between relative factor sup-
(relatively) abundant supply of human plies and factor prices. For if wages were
than of physical capital. An increase in high primarily because of human capital,
human capital would, however, show up a regression of the relative amount of
as an apparent increase in labor intensity physical capital on wages could give a
since earnings are gross of the return on seriously biased picture of the effect of
such capital. Thus export industries factor proportions on wages.74
might pay higher wages than import
72 This kind of interpretation has been put for-
competing ones primarily because they ward by many writers; see, for example, the discus-
employ more skilled or healthier work- sion in W. Leontief, "Factor Proportions and the
Structure of American Trade: Further Theoretical
ers.72
and Empirical Analysis," Review of Economics and
3. Several recent studies have tried Statistics, Vol. XXXIII (November, 1956).
73 Interstate estimates for several industries can

70 According to eq. (28) earnings at age j can be
be found in J. Minasian, "Elasticities of Substitution
approximated by
and Constant-Output Demand Curves for Labor,"
k=j-l
Journal of Political Economy, LXIX (June, 1961),
Yj = Xj + E rkCk-C1, 261-70; intercountry estimates in Kenneth Arrow,
k =o Hollis B. Chenery, Bagicha Minhas, and Robert M.
Solow, "Capital-Labor Substitution and Economic
where Xi are earnings at j of persons who haveEfficiency,"
not Review of Economics and Statistics (Au-
invested in themselves, Ck is the investment at age gust, 1961); unpublished papers by Philip Nelson
k, and rk is its rate of return. The rate of increase in and Robert Solow contain both interstate and time-
earnings would be at least as steep in V as in X at series estimates.
each age and not only from "younger" to "older"
7 Minasian's argument (op. cit., p. 264) that in-
ages if and only if
terstate variations in skill level necessarily bias his
Yj > A Xi estimates toward unity is actually correct only if
skill is a perfect substitute for "labor." (In corre-
4j - 4'
or spondence Minasian states that he intended to make
this condition explicit.) If, on the other hand, human
> AC- and physical capital were perfect substitutes the es-
timates would always have a downward bias, regard-
less of the true substitution between labor and capi-
This condition is usually satisfied since rjCj > 0 and tal. Perhaps the most reasonable assumption would
the amount invested tends to decline with age.
be that physical capital is more complementary with
71 See I. Kravis, "Wages and Foreign Trade," human capital than with labor; I have not, however,
Review of Economics and Statistics, Vol. XXXIII been able to determine the direction of bias in this
(February, 1956). case.
INVESTiMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 45
4. A secular increase in average earn- sonality, persistence, and intelligence.

ings has usually been said to result from Accordingly, some writers have gone to
increases in technological knowledge and the opposite extreme and argued that the
physical capital per earner. The average only relevant way to measure economic
earner, in effect, is supposed to benefit talent is by results, or by earnings them-
indirectly from activities by entrepre- selves.77 Persons with higher earnings
neurs, investors, and others. Another ex- would simply have more ability than
planation put forward in recent years others, and a skewed distribution of earn-
argues that earnings can rise because of ings would imply a skewed distribution
direct investment in earners.75 Instead of of abilities. This approach goes too far,
only benefiting from activities by others, however, in the opposite direction. The
the average earner is made a prime main reason for an interest in relating
mover of development through the in- ability to earning is to distinguish its ef-
vestment in himself.76 fects from differences in education,
training, health, and other such factors,
B. ABILITY AND THE DISTRIBUTION
and a definition equating ability and
OF EARNINGS
earnings ipso facto precludes such a dis-
An emphasis on human capital not
tinction. Nevertheless, results are very
only helps explain differences in earnings
relevant and should not be ignored.
over time and among areas but also
A compromise might be reached
among persons or families within an area.
through defining ability by earnings only
This application will be discussed in
when several variables had been held
greater detail than the others because a
constant. Since the public is very con-
link is provided among earnings, ability,
cerned about separating ability from
and the incentive to invest in human
education, on-the-job training, health,
capital.
and other human capital, the amount in-
Economists have long been aware that
vested in such capital would have to be
conventional measures of ability-in-
held constant. Although a full analysis
telligence tests or aptitude scores, school
would also hold discrimination, nepo-
grades, and personality tests while un-
tism, and several other factors constant,
doubtedly relevant at times, do not re-
a reasonable first approximation would
liably measure the talents required to
say that if two persons have the same
succeed in the economic sphere. The
investment in human capital, the one
latter requires a particular kind of per-
who earns more is demonstrating greater
75 The major figure here undoubtedly is T. W. economic talent.
Schultz. Of his many articles see esp. "Education
Since observed earnings are gross of
and Economic Growth" in Social Forces Influencing
American Education (Sixtieth Yearbook of the Na- the return on human capital they are af-
tional Society for the Study of Education, Part II fected by changes in the amount and
[Chicago: University of Chicago Press, 1961]). rate of return. Indeed, after the invest-
" One caveat is called for, however. Since ob-
ment period earnings (Y) can be simply
served earnings are not only gross of the return from
investments in human capital but also are net of approximated by
some costs, an increased investment in human capi-
tal would both raise and reduce earnings. Although Y = X + rC, (32)
average earnings would tend to increase as long as
the rate of return was positive, the increase is less 77 Let me state again that whenever the word
than it would be if the cost of human capital, like "earnings" appears I mean real earnings, or the sum
that of physical capital, was not deducted from of monetary earnings and the monetary equivalent
national income. of psychic earnings.
46 GARY S. BECKER
where C measures total investment costs, One is that the tendency for abler per-
r the average rate of return, and X earn- sons to migrate, continue their educa-
ings when there is no investment in hu- tion,82 and generally invest more in them-
man capital. If the distribution of X is selves can be explained without recourse
ignored for now, Y would depend only to an assumption that non-economic
on r when C was held constant, so "abil- forces or demand conditions favor them
ity" would be measured by the average at higher investment levels. A second
rate of return on human capital.78 implication is that the separation of
The amount invested is not the same "nature from nurture" or ability from
for everyone, nor even in a very imper- education and other environmental fac-
fect capital market rigidly fixed for any tors is apt to be difficult, for high earn-
given person, but depends in part on the ings would tend to signify both more
rate of return. Persons receiving a high ability and a better environment. Thus
marginal rate of return would have an the earnings differential between college
incentive to invest more than others.79 and high-school graduates does not
Since marginal and average rates are pre- measure the effect of college alone since
sumably positively correlated80 and since college graduates are abler and would
ability is measured by the average rate, earn more even without the additional
one can say that abler persons would in- education. Or reliable estimates of the
vest more than others. The end result income elasticity of demand for children
would be a positive correlation between have been difficult to obtain because
ability and the investment in human higher income families also invest more
capital,8' a correlation with several im- in contraceptive knowledge.83
portant implications. The main implication, however, is in
the field of personal income distribution.
78 Since r is a function of C, Y would indirectly as
well as directly depend on C, and therefore the dis-
At least ever since the time of Pigou
tribution of ability would depend on the amount of economists have tried to reconcile the
human capital. Some persons might rank high in strong skewness in the distribution of
earnings and thus high in ability if everyone were
unskilled, and quite low if education and other train-
earnings and other income with a pre-
ing were widespread. sumed symmetrical distribution of abili-
79 In addition, they would find it easier to invest ties.84 Pigou's own solution, that prop-
if the marginal return and the resources of parents erty income is not symmetrically dis-
and other relatives were positively correlated.
tributed, does not directly help explain
80 According to a well-known formula the skewness in earnings. Subsequent at-
( 1 tempts have largely concentrated on
r = ral+ +1
developing ad hoc random and other
where rm is the marginal rate of return, ra the average
probabilistic mechanisms that have little
rate, and ea the elasticity of the average rate with
82 The first is frequently alleged (see, for example,
respect to the amount invested. The rates rm and ra
Marshall, op. cit., pp. 199, 684). Evidence on the sec-
would be positively correlated unless ra and l/ea
ond is discussed in my forthcoming study for the
were sufficiently negatively correlated.
National Bureau of Economic Research.
81 This kind of argument is not new; Marshall
83 See my "An Economic Analysis of Fertility" in
argued that business ability and the ownership of
Demographic and Economic Change in Developed
physical capital would be positively correlated: Countries (Princeton, N.J.: Princeton University
"[economic] forces . . . bring about the result that
Press, 1960).
there is a far more close correspondence between the
ability of business men and the size of the businesses 84 See A. C. Pigou, The Economics of Welfare (4th
which they own than at first sight would appear ed.; London: Macmillan & Co., 1950), Part IV,
probable" (op. cit., p. 312). chap. ii.
relation to the mainstream of economic capable, therefore, of reconciling a strong

thought.85 The approach presented here, positive skewness in earnings with a pre-
however, offers an explanation that is sumed symmetrical distribution of abili-
not only consistent with economic analy- ties.
sis but actually relies on one of its Variations in X help explain an im-
fundamental tenets; namely, that the portant difference among skill categories
amount invested is a function of the rate in the degree of skewness. The smaller
of return expected. In conjunction with the fraction of total earnings resulting
the effect of human capital on earnings
this tenet can explain several well-known 86 Jacob Mincer ("Investment in Human Capital
and Personal Income Distribution," Journal of Po-
properties of earnings distributions.
litical Economy, Vol. LXVI [August, 1958]) con-
By definition, the distribution of earn- cluded that a symmetrical distribution of investment
ings would be exactly the same as the dis- in education implies a skewed distribution of earn-
ings because he defines educational investment by
tribution of ability if everyone invested school years rather than costs. If we follow Mincer
the same amount in human capital; in in assuming that everyone was equally able, that
particular, if ability were symmetrically schooling was the only investment, and that the cost
of the nth year of schooling equaled the earnings of
distributed, earnings would also be. persons with n - 1 years of schooling, then, say, a
Equation (32) shows that the distribu- normal distribution of schooling can be shown to im-
tion of earnings would be exactly the ply a log-normal distribution of school costs, and
thus a log-normal distribution of earnings.
same as the distribution of investment if The difference between the earnings of persons
all persons were equally able; again, if with n - 1 and n years of schooling would be kA =
investment were symmetrically dis- Y- Yn,- = rnCn. Since r, is assumed to equal r for
all n, and Cn = Y.-l, this equation becomes Y, =
tributed, earnings would also be.86 If
(1 + r) Yn.4 and therefore
ability and investment both varied,
Cl = Yo
earnings would tend to be skewed even
when ability and investment were not, C2= Y,= Yo(1+r)
but the skewness would be small as
long as the amount invested was sta- C3 = Y2 = Yj(1 + r) = YO(1 + r)2
tistically independent of ability.87 Q = Yn-1 = Yo(1 + r)n 1 ,
Our analysis has shown, however, that
or the cost of each additional year of schooling in-
abler persons would tend to invest more creases at a constant rate. Since total costs have the
than others, so ability and investment same distribution as (1 + r)n, a symmetrical, say a
would be positively correlated, perhaps normal, distribution of school years, n, implies a log-
normal distribution of costs and hence by eq. (32) a
quite strongly. Now the product of two log-normal distribution of earnings. I am indebted to
symmetrical distributions is more posi- Mincer for a helpful discussion of the comparison
tively skewed the higher the positive and especially for the stimulation provided by his
pioneering work. Incidentally, his article and the
correlation between them, and might be
dissertation on which it is based cover a much
quite skewed.88 The economic incentive broader area than has been indicated here.
given abler persons to invest relatively 87 For example, C. C. Craig has shown that the
large amounts in themselves does seem product of two independent normal distributions is
only slightly skewed (see his "On the Frequency
85 A sophisticated example can be found in
Function of XY," A nnals of AMathematical Statistics,
B. Mandelbrot, "The Pareto-Levy Law and the
VII [March, 1936], 3).
Distribution of Income," International Economic
Review, Vol. I (May, 1960). In a recent paper, how- 88 Craig (op. cit., pp. 9-10) showed that the prod-
ever, Mandelbrot has brought in maximizing be- uct of two normal distributions would be more posi-
havior (see "Paretian Distributions and Income tively skewed the higher the positive correlation be-
Maximization," Quarterly Journal of Economics, tween them, and that the skewness would be con-
Vol. LXXVI [February, 1962]). siderable with high correlations.
48 GARY S. BECKER
from investment in human capital-the older ages because the presumed negative
smaller rC relative to X-the more i-1
would the distribution of earnings be correlation between -C3 and E riCi

0
dominated by the distribution of X.
would counteract the positive correlation
Higher skill categories have a greater
between ability and investment.
average investment in human capital
A simple analysis of the incentive to
and thus presumably a large rC relative
invest in human capital seems capable of
to X. The distribution of "unskilled
explaining, therefore, not only why the
ability," X, would, therefore, tend to
over-all distribution of earnings is more
dominate the distribution of earnings in
skewed than the distribution of abilities,
relatively unskilled categories while the
but also why earnings are more skewed
distribution of a product of ability and
among older and skilled persons than
the amount invested, rC, would dominate
among younger and less skilled ones. The
in skilled categories. Hence if abilities
renewed interest in investment in hu-
were symmetrically distributed, earn-
man capital may provide the means of
ings would tend to be more symmetrical-
bringing the theory of personal income
ly distributed among the unskilled than
distribution back into economics.
among the skilled.89
Equation (32) holds only when in-
VI. SUMMARY AND CONCLUSIONS
vestment costs are small, which tends to
be true at later ages, say after age thirty- Most investments in human capital
five. Net earnings ar earlier ages would both raise observed earnings at older
be given by ages, because returns are added to earn-
ings then, and lower them at younger
i-1
ages, because costs are deducted from
Yj= X-j+ E rCI+ (-Cj),
0
earnings then. Since these common ef-
fects are produced by very different kinds
where j refers to the current year and i of human capital, a basis is provided for
a unified and powerful theory. The
to previous years, Ci measures the invest-
ment cost of age i, Cj current costs, andanalysis proceeded from a discussion of
specific kinds of human capital, with
ri the rate of return on Ci. The distribu-
tion of - Cj would be an important de- greatest attention paid to on-the-job
terminant of the distribution of Yj since training because it clearly illustrates and
investment is large at these ages. Hence emphasizes the common effects, to a
our analysis would predict a smaller general theory applying to any kind.
(positive) skewness at younger than at The general theory has a wide variety
of important implications, ranging from
89 As noted earlier, X does not really represent interpersonal and interarea differences in
earnings when there is no investment in human capi-
tal, but only earnings when there is no investment
earnings, to the shape of age-earning
after the initial age (be it fourteen, twenty-five, or profiles, to the effect of specialization on
six). Indeed, the developmental approach to child- skill. For example, since earnings are
rearing argues that earnings would be close to zero if
there was no investment at all in human capital. The
gross of the return on human capital,
distribution of X, therefore, would be at least partly some persons may earn more than others
determined by the distribution of investment before simply because they invest more in them-
the initial age, and if it and ability were positively
correlated, X might be positively skewed, even
selves. And since "abler" persons tend to
though ability was not. invest more than others, the distribution
of earnings could be very unequal and employment among unskilled than skilled
even skewed, even though "ability" workers may result from the latter hav-
were symmetrically and not too unequal- ing more specific capital; or incompletely
ly distributed. To take another example, vested pension plans may be a means of
learning, both on and off the job, and insuring firms against a loss on their
other activities appear to have exactly specific investments.
the same effects on observed earnings as This paper has concentrated on de-
do education, training, and other tradi- veloping a theory of investment in hu-
tional investments in human capital. We man capital, with an emphasis on empiri-
argue that a relevant concept should cal implications rather than on formal
cover all activities with identical effects generalization. Of course, empirical use-
and show that the total amount in- fulness is the only justification for any
vested in a generalized concept of human theory, and although I did not try to
capital and its rate of return can be esti- bring in even the quite limited evidence
mated from information on earnings on the role of human capital, the em-
alone. pirical work reported in this volume, my
Some investments in human capital do own work, and that of many others sup-
not affect earnings because costs are paid port the view that investment in human
and returns are collected by the firms, capital is a pervasive phenomenon and a
industries, or countries using the capital. valuable concept. The next few years
These "specific" investments range from should provide much stronger evidence
hiring costs to executive training and are on whether the recent emphasis placed
more important than is commonly be- on this concept is just another fad or a
lieved. To take a couple of examples, we development of great and lasting im-
showed that the well-known greater un- portance.

Investment in Human Capital - Becker

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Investment in Human Capital - Becker

Uploaded by

Copyright:

Available Formats

Investment in Human Capital: A Theoretical Analysis

Author(s): Gary S. Becker

Columbia University and National Bureau of Economic Research

I. INTRODUCTION prove the physical and mental abilities of

affected: on-the-job training primarily lating these differences primarily to dif-

schooling, on-the-job training, medical may be useful in attempts to understand

on earnings and consumption, in the college and high-school education in the

that, placing special emphasis on the where W equals wages or expenditures

Rn -I t n -I E t as the sum of opportunity costs and out-

or a doctor trained (interned) at one Equation (7) is reduced to

to be subtracted or a positive apprecia- period and equal afterwards, they would

the cost of, and return from, the invest- FIG. 1

further demonstrated by contrasting training in a wide variety of skills and

MP'+ aC = W + C, (13) of such training is borne entirely by em-

received as higher wages, so this formally store or truck system in nineteenth-cen-

profiles. The cause of this difference is III. RELATION BETWEEN EARNINGS,

investment period tend to be net of the

tracted and returns added to "gross"

ment language throughout. (I (1i)id-C=R-C. (20)

The present value of the net earnings

equate total costs and returns. In sym-

applies regardless of the amount and m-1

duration of investment, but the defini- C1= E(Xj- Y-)(I+ O-i,

were known to occur in Y during each liM C, (X X- IYj) (I + r ) i=0,

of the first m periods, a simple and super-

net earnings in X and Y, total invest-

kj, with k = Ykj being the total return

C= lCj=*, (25) m-1 m-1 m-1 r~

m-1 Cj I am indebted to Helen Raffel for important sugges-

So costs and the rate of return can be

The analytical difference between the naive defi- m-1

nition of costs advanced earlier and one in terms of k = E risCi

costs by the value of earning differentials at the be-

j-1 definition is the same: investment occurs

ECj suppose that learning was essentially un-

along with training and schooling.

only be slightly raised if people remained amount invested can be distinguished

find investments that increase productiv- B. WAGE DIFFERENTIAlS AND

D. CAPITAL MARKETS AND KNOWLEDGE 1. Since large expenditures would be

for a firm he is starting rather than in a The eighteen-year-old in our example

the age-earnings profile]0 to estimate empirically the elasticity of

73 Interstate estimates for several industries can

4. A secular increase in average earn- sonality, persistence, and intelligence.

relation to the mainstream of economic capable, therefore, of reconciling a strong

would the distribution of earnings be correlation between -C3 and E riCi

You might also like