P. 1
Invt Human Capital

Invt Human Capital

|Views: 6|Likes:
Published by Shweta Kulkarni

More info:

Published by: Shweta Kulkarni on Mar 16, 2012
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less





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

Becker Reviewed work(s): 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: 08/03/2012 12:34
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp 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.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Political Economy.


GARY S. BECKER Columbia University and National Bureau of Economic Research I. INTRODUCTION

ture well-being,while others have their main impact in the present. Dining is an example of the latter,while purchaseof a car exemplifies former. the Both earningsand consumptioncan be affected: on-the-job training primarily affects a earnings, new sail boat primarily affects consumption, and a college education is said to affectboth. The effectsmay operateeither through physical resources,such as a sail boat, or through human resources,such as a college education. This paper is concerned with activities that influencefuture real inthe come through imbeddingof resources in people. This is called investing in human capital. The many ways to invest include schooling, on-the-job training, medical care, vitamin consumption,and acquiring information about the economic in system.They differ the relativeeffects on earnings and consumption, in the amount of resourcestypically invested, in the size of returns, and in the extentto which the connection between investis mentand return perceived.But all im1 I am greatly indebtedto the CarnegieCorporation of New York forthe supportgiven to the NationalBureau ofEconomicResearchto studyinvestmentin educationand otherkindsofhumancapital. greatlyfrommany discussionswith my I benefited colleagueJacob Mincer,and also withotherparticipants in the Labor Workshopof Columbia Univervaluable comsity. Althoughmany personsoffered I mentson thedraftpreparedfortheconference, am especially indebted to the detailed comments of Johnand Shirley TheodoreSchultz,GeorgeStigler, son.

SOME activities primarily affect fu-

prove the physicaland mentalabilitiesof people and thereby raise real income prospects. People differ substantially in their economic well-being,both among countries and among familieswithin a given country.For a while economistswere relating these differences primarilyto difin ferences the amountof physicalcapital since richer people had more physical capital than others. It has become increasingly evident,however,from studies of income growth2that factors other than physical resources play a larger role than formerly believed, thus focusing attention on less tangible resources, like the knowledgepossessed. A concern with investment in human capital, ties in closely with the new therefore, emphasis on intangible resources and may be usefulin attemptsto understand the inequality in income among people. The original aim of my study was to estimate the money rate of return to college and high-schooleducation in the United States. In orderto set these estimates in proper context I undertook a of briefformulation the theoryof investment in human capital. It soon became clear to me, however, that more than a restatementwas called for: while important and pioneeringwork had been done on the economic returnto various
2 The evidence for the United States appears to show that the growthin capital per capita explains in onlya small part ofthe growth per capita income and that the growthin "technology"explainsmost and Progress ofit. On thissee S. Fabricant,Economic of EconomicChange:34thAnnual Report tIe National Bureau of EconomicResearch(New York: National Bureau of Economic Research, 1954).




occupationsand educationclasses,3there have been few,if any, attemptsto treat the process of investingin people from a general viewpoint or to work out a broad set of empirical implications. I began then to prepare a generalanalysis in of investment human capital. As the work progressed, it became clearerand clearerthat much more than a gap in formaleconomicanalysis would be filled,for the analysis of human inexplanationof a vestmentoffered unified a wide range of empirical phenomena whichhad eitherbeen given ad hocinterpretations or had baffledinvestigators. (1) Amongthese are the following: Earnings typicallyincrease with age at a decreasing rate. Both the rate of increase and the rate of retardationtend to be positivelyrelatedto the level of skill. (2) Unemploymentrates tend to be negatively related to the level of skill. (3) Firms in underdeveloped countries appear to be more "paternalistic" toward employeesthan those in developed countries. (4) Younger persons change jobs and receivemoreschoolmorefrequently ing and on-the-job training than older of personsdo. (5) The distribution earnings is positively skewed, especially among professional and other skilled workers.(6) Abler persons receive more education and other kinds of training than others. (7) The division of labor is limitedby the extentof the market. (8)
I In additionto the earlierworksof Smith,Mill, and Marshall,see H. Clark,LifeEarningsin Selected in Occupations theU.S. (New York: Harper & Bros., 1937); J. R. Walsh, "Capital Concept Applied to Man," Quarterly Journal of Economics,February, 1935; M. Friedmanand S. Kuznets,Incomefrom Independent Professional Practice(New York: National Bureau ofEconomicResearch, 1945); G. Stiglerand D. Blank, The Demand and Supply ofScientific Personnel (New York: National Bureau of Economic Research,1957); and T. W. Schultz,"Investmentin Man: An Economist'sView," Social ServiceReview, June,1959.

The typical investorin human capital is more impetuousand thus more likelyto in errthan is the typicalinvestor tangible capital. What a diverseand possiblyeven array! Yet all these as well as confusing many other importantempirical implications can be derived fromvery simple theoretical arguments. The purpose of this paper is to set out these arguments in some generality,with the emphasis placed on empirical implications, although little empirical material is presented. My own empiricalwork will appear in a later study. First, a lengthydiscussion of on-thejob trainingis presentedand then,much in discussionsofinvestment morebriefly, and health. Onschooling,information, the-job trainingis dealt with so elaborately not because it is more important in than otherkindsof investment human capital-although its importanceis often underrated-but because it clearlyillustrates the effectof human capital on earnings, employment,and other economic variables. For example, the close connectionbetween foregoneand direct of costs or the effect human capital on earnings at differentages is vividly broughtout. The extendeddiscussionof on-the-job training paves the way for muchbriefer discussionsofotherkindsof in investment human beings.

Theories of firmbehavior, no matter in how they differ otherrespects,almost of invariablyignorethe effect the productive process itselfon workerproductivity. This is not to say that no one recognizes that productivityis affected has by the job itself;but the recognition not been fonnalized, incorporated into economic analysis, and its implications worked out. We now intend to do just





that, placing special emphasis on the broader economicimplications. Many workersincrease their productivityby learningnew skills and perfecting old ones while on the job.. For example, the apprentice usually learns a new skill while the interndecompletely velops skills acquired in medical school, and both are moreproductiveafterward. On-the-job training, therefore, is a process that raises future productivity fromschool trainingin that and differs is an investment made on the job rather that specializes in than in an institution teaching.Presumably,futureproductivity can be improvedonly at a cost, for otherwisethere would be an unlimited Included in cost are demand fortraining. of a value placed on the time and effort trainees, the "teaching" provided by others,and the equipmentand materials used. These are costs in the sense that they could have been used in producing currentoutput if they were not used in raisingfutureoutput. The amount spent and the duration of the trainingperiod depend partly on the type of trainingmore is spent for a longer time on an internthan on an operative-partly on production possibilities,and partly on skills. the demand fordifferent Each employeeis assumed to be hired timeperiod (in the limiting fora specified case thisperiodapproacheszero), and for the moment both labor and product markets are assumed to be perfectly competitive.If therewere no on-the-job training,wage rates would be given to the firm and would be independentof its A profit-maximizing would firm actions. products whenmarginal be in equilibrium equaled wages, that is, when marginal receiptsequaled marginal expenditures. In symbols MP=W, (1)

where W equals wages or expenditures and MP equals the marginalproduct or receipts. Firms would not worry too much about the relation between labor conditions in the present and future partly because workerswere only hired forone period,and partlybecause wages and marginalproductsin futureperiods would be independentof a firm's current behavior.It can therefore be legitimately assumed that workershave unique marginal products (for given amounts of other inputs) and wages in each period, which are, respectively,the maximum in productivity all possible uses and the market wage rate. A more complete set ofequilibrium conditions would be the set




where t refersto the tth period. The equilibrium positionfor each periodwould depend only on the flows during that period. These conditionsare altered when account is taken of on-the-jobtraining and the connectiontherebycreated between presentand futurereceiptsand expenditures. Training might lower current receipts and raise currentexpenditures, yet firmscould profitablyprovide this if training futurereceiptswere sufficiently raised or future expenditures sufficientlylowered. Expenditures during each period need not equal wages, receipts need not equal the maximum possible productivity,and expenditures and receiptsduringall periods would be interrelated. The set of equilibrium conditions summarized in equation (2) would be replaced by an equality between the presentvalues of receiptsand expenditures.If Et and Rt representexpendituresand receiptsduringperiod t, and i the marketdiscount rate,then the equilibriumconditioncan be writtenas

to produce currentoutput. job trainingproduced some general reSince the termk onlymeasuresthe actual sults summarizedin equations (3) and outlay on trainingit does not entirely (7) ofwide applicability. If this MPO+ +1 ( 1 O t +0Z implied that G = C.Equation (7) shows that marginal product would equal wages in the initial period only when the return equals costs. call this MPo and what is pro. but such wherek measuresthe outlay on training. and.in addition to the firm providingit. from. made on the job and costs. training.the difference betweenG and C measuresthe difference betweenthe return and the cost of. and receipts during all periods would equal marginal products.12 Rn -I GARY S. as a duced.(6) becomes MP'+G=Wo+C.the present value of the marginal product streamwould have to equal the present value of the wage stream. for excluded is crete results require more specific asthe time that a person spends on this sumptions. that marginal product equals wages. General.discussed in turn: general and specific. The equilibrium condition of equation (2) has been generalized. equation (4) can be writtenas 1. marginal product (4) would equal wages in the initial period. . MPo. or G = C.General trainingis usefulin many firms duced. Those familiarwith capital theorymightargue that this generalization of the simple equality between marginal product and wages is spurious because a full equilibriumwould require equality between the returnfroman investment-in this n-i MRt _ case. There is much to be said for the rele+ =WO k+E t)tX vance of a conditionequating the return froman investment with costs. The differ.In this and the following sectime that could have been used tion two types of on-the-jobtrainingare training. is the opportunitycost of machinisttrained in the army findshis the time spentin training. a conditiondoes not implythat G = C or If a new termis defined. the excess of future receiptsover future outlays.is a measure of the returnto the firm fromproviding training. Our treatmentof on-the(6) APo P+G=Wo+k. depend on all other receipts and expenditures. however. (7) where n represents the number of periods. ence betweenwhat could have been pro. and E.If C is defined skills of value in steel and aircraft firms. Equation (3) becomes The term G.the converseneed not hold. BECKER v 0 t +i) n -I Et + (3) (I t=o as the sum of opportunity costs and outlays on training. it would be greateror less than wages as the return was smalleror greaterthan costs. and R. expenditures during the initialperiodwould equal wages plus the outlay on training. for if marginal productequals wages in each period.but moreconmeasure training costs.expendituresduring other periods would equal wages alone. Obviously. therefore. If trainingwere given only duringthe initial period. The followingdiscussion demonstrates G= E M1-Wt (5) that great care is requiredin the applicat=__ (1 +iW tion of this condition to on-the-job investment.

Equation (10) has many other and the rest of this section implications. is devoted to developing the more importantones.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 13 or a doctor trained (interned) at one hospital finds his skills useful at other hospitals. . In otherwords. MPo=Wo+ or Wo=MPo-k.and a capital or stock term. In other words. but general training would also increase their marginal product in many otherfirms well. wage rates would rise by exactly the same amount as the marginal product and the firmsproviding such training could not capture any of the return.but later I show (pp.and therefore __ =__0MP- Equation (7) is reduced to MP. t ( 8) 8 t=1 ( The wage of trainees would not equal their opportunitymarginal product but would be less by the total cost of training.not by firms.n-1 .=WWO+C. or VO=MP. Most on-the-jobtrainingpresumably increases the future marginal providing productof workersin the firm it. in Why. Since wages and marginalproducts are raised by the same amount. trainingcosts would have to be included in any study of the relationbetweenwages and productivity. Hence the cost as well as the return from general training would be borne by trainees. Persons receiving general trainingwould be willing to pay these costs since training raises their future wages. These firmscould capture some of the return from training only if their marginal product rose by more than theirwages. Such an interpretationof net productivitycould formallysave the equalitybetweenmarginal productand wages here. Consequently. so that the capital and income accounts would be closely intermixed. earningsof persons receiving on-the-jobtrainingwould be net of . Since in a competitive as labor marketthe wage rates paid by any by firm determined marginalproducare futurewage rates tivitiesin other firms. then.employeeswould pay for general trainingby receivingwages below their current (opportunity) productivity. 18-25) that it cannot always be saved. with changes in either affectingwages. for all t = 1. (9) (10) (9') (10') In termsof actual marginalproduct (1+it - t=?. as well as marginal products would increase to firmsprovidinggeneral training. Some mightargue that a really "net" of definition marginal product obtained by subtracting training costs from "gross" marginal product must equal wages even for trainees. "Perfectlygeneral" training would be equally useful in many firms and marginalproductswould riseby the same extentin all of them. must equal W.regardlessof which inis terpretation used."Earnings" duringthe trainingperiod would be the between an income or flow difference term. MP.training costs. -C. . Employees pay forgeneral on-the-job trainingby receivingwages below what could be receivedelsewhere.potential marginalproduct. Moreover.for why provide trainingthat brings no return? The answer is that firms would provide general training only if they did not have to pay any of the costs. k. These and other implicationsof general training can be more formally demonstratedwith equation (7).. do rational firms competitive labor markets provide general training.

since traineeswould tend to have lower 6 A difference signsis impossible Friedman's in in currentand highersubsequent earnings analysis of consumerbehavior because he assumes and than other youth.14 GARY S.appear to differ luctance to treat people as capital and of depreciation than in its existence. 1955. even though their long. Education." August. .6 Doubt has been cast on the frequent throughoutthis paper.7 A depreciation-type of ing conventionsused fortransactionsin item is deducted. I I say "young people" ratherthan "young families" because as J. and Welfare (forthcoming)]) that educated personsshould be permittedto subtract fromincomea depreciationallowance on tuition payments. currentand a husband's long-runincome greater and long-run than that betweena husband'scurrent income. Marshall.human capital.1957. which subtracts all investmentcosts from"gross" earn." Econometrica. 1949). Januto Wealthand the Wage Rate. JournalofPolitical Economy.PrinciplesofEcotion with long-run earnings."The underlying morein the timepattern ence undoubtedlyis the widespread re. Interest. we are suggesting correlatedforyoungpersons. Goode has argued (see and theIncome Tax.J. Participation of wives. R. betweena family's makes the correlation therefore.. In particular.Economics HigherEducation [Washington: United States Department of Health.that transitory long-run(that is. one that too much or too mightargue. C. but the 4 Of course. "The Relation of Consumption and W. at least fromthe earnmaterial goods-which separate income ings due to on-the-jobtraining.life.for the from capital accounts to prevent a cost would be deducted duringthe traintransaction in capital from ipso factor4 ing period. Mushkin[ed. Mincer has shown(in a paper to be publishedin a National Bureau of Economic Revolume on labor economics). only be much weaker than the correla. definition earningsfromthe account. Hence human and tangible capital off. a and negative depreciation term would have youngpersonsreceivesome training. (8thed. p." in "Educational Expenditures of Selma J.assertion that no allowance is made in our departurewith this the income accounts fordepreciationon ings. on is rapid depreciation permitted such investment. "write-off"of human capital is what Intermixingthe capital and income should oftenbe emphasized and studied.1959]). the correlation be. forexample. Hamburger.the search conference participationof wives is positivelycorlabor-force betweenhusbands' longrelatedwiththe difference run and current income. 352.may be negatively I See. New York: MacmillanCo. and Our point can be put differently accounts could make the reported "inThe ideal depreciation comes" oftraineesunusuallylow and per. permanent) tween currentconsumptionand current incomes are uncorrelated(see his A Theoryof the Function[Princeton. Since a considerablefractionof the period.if value rose.on the contrary. Eshift betweenassets having differthe income account would affect ent productivities on material goods even with currentaccounting practices. Therefore. Depreciation on tangible the income side is not capri.N."Patinkin on Money.on a capital asset during any period run or lifetimeincomes were well above would equal its change in value during average.capital does not bulk so large in any one affecting cious but is groundedin a fundamental period because it is usually "writtenoff" in or depreciated during a period of time difference betweentheway investment materialand human capital are "written designed to approximate its economic cause of this differ.8 on the accompanyingtendencyto treat all and the effect wage income of a rapid wage receiptsas earnings.A.: Princeton Consumption earnings of young people' would not University thatthey Press. nomics Christ. 8 In a recent paper. haps negative. ary.and Prices.Such an allowanceis apparentlynot required for on-the-jobtrainingcosts.more rigorously.]. indeed.. BECKER investmentcosts and would correspond signs of these correlationsmight even to the definitionof net earnings used differ.

9 Training has an important effecton the relation between earnings and age. that training raised the level of marginal producon tivitybut had no effect the slope. Indeed.the economic"value" of a trainee would at firstincrease rather than decrease with age.hIthe total . and higherearningsat later ages because the returnis collected then. C. however.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANAILYSIS 15 to be subtractedor a positive appreciation termadded to the income fromthe asset. may really be an important part of costs. X being consumption goods. so that themarginalproductivity trained of persons was also independentof age. Since trainingcosts would be deducted from earningsduringthe training period. Not only does training make the curve steeper but.E lX1)?+Y wherepi is the marketprice of the ith good. 1). * X) X . all costs appear as foregoneearnings to workers T zc/ I. as indicated by Figure 1. subject to the constraint P-ix=wV (h . all costs appear as lowerearningsthan could be received elsewhere. impartinga concave appearance to the curve as a whole. earnings of trained persons would be below marginal productivityduring the training 9 In my study forthe National Bureau of Economic Research I try to measure the relationbetween depreciationand age for several education classes.i . the difference being greater the greater the cost of. In this extremecase an extreme concavityappears.A householdcan be assumedto maximize utilityfunction U(K1. Trained personswould receive lowerearningsduringthe training period because training is paid for then. and returnfrom. and only later would it begin to decrease. 1 receivingon-the-jobtraining. showingneitherslope nor concavity._ U AGE FIG. X 2. rise sharply at the end of the training periodand thenlevel off(as shownby the dashed line T'T' in Fig. If earningsequaled marginal product.that is. to take an extreme case. although direct outlays. althoughneglected. in less extremecases the principlewould be the same and the concavitymore continuous. Suppose that untrainedpersonsreceived the same earnings regardlessof age. they would period and equal afterwards. Suppose. the rate of increase in earnings is affected moreat youngerthan at older ages.cost of much human capital and shouldbe treatedon the same as footing directoutlays. steeper than that of untrainedpersons. that is. as shown by the horizontal line UU in Figure 1. The combined effect payingforand collectingthe reof turn fromtrainingin this way would be to make the age earningscurveoftrained persons. shown by TT in Figure 1. Since. The arbitrarinessof the division between foregone and direct costs and the resulting advantage of treatingtotal costs as a whole" can be 10 The equivalence between foregone and direct costsapplies to consumption wellas to investment as a decisions. TT would merelybe parallel to and higher than UU. I14the average wage rate. Foregone earnings are an important.the investment. y non-wageincome. also more concave.

and thecause ofdifferences amongincomeclasses in priceelasticities demand.Firms that both paid fortraining and less than the marketwage fortrained persons would have the worst of both worlds. and hopefully This hope. Firms that did not pay trained persons the satismarketwage would have difficulty and would fyingtheirskill requirements also tend to be less profitable than other firms.It has had. well over 90 per cent of United States commercialairlinepilots received much of their training in the armed forces. of trainingin a wide variety of skills and many-such as pilotingand machine repair-are very usefulin the civilian sector.and hj the numberof hoursrequiredto consume a unit of thejth good. A shiftof trainingfromschools to on the job would. its graduates make part of the supply in up the predominate several civilian occupations.of course.Usually only the direct cost of school trainingis emphasized. Vol. bor and "leisure.11 Persons withthese skillsleave the militarymore readily because they can receive much higher wages in the civilian sector.The military.and labor costs would be relatively high. not The military onlywantsto eliminatethe inverse would like to createa strong relationbut apparently positiverelationbecause theyhave such a large inin vestment heavily trainedpersonnel(see ibid. however. and the accompanyingdiscussion. the military. however. I expectto showin another paper that this formulation household decisions gives of extremely usefulinsights intoa number important of such as the choice betweenlaeconomicproblems. BECKER further demonstrated by contrasting school and on-the-jobtraining. For example.16 GARY S. many persons would seek training." the effectof price control on prices. in consequence.however. I. By transposing terms the constraint as can be written 2(p?+Whti))X =Wh+y . relativelyeasy access to "students" and heavy losses of "graduates.few would quit duringthe training period." Indeed.forthey would attract too many traineesand too fewtrainedpersons. Training is provided duringpart or enlistment periodand used all ofthe first duringthe remainderof the firstperiod duringsubsequentperiods. The militaryis a conspicuousexample of an organization that both pays at least part of trainingcosts and does not pay marketwages to skilledpersonnel. Income maximizingfirmsin competitive labor markets would not pay the cost of general trainingand would pay trained persons the market wage.theroleofqueues.Not surprisingly. What about the old argument that and Compensation " See ManpowerManagement (Washington: GovernmentPrintingOffice.).trainingcosts were paid.is thwartedby the rates tend to be fact that re-enlistment inversely related to the amount of civilian-type skills provided by the military. even though the foregone cost is sometimes(as with college education) an importantpart of the total.even when direct outlays were important.reversethe emphasis and make all costs appear as foregone earnings.1957). . Chart 3. is not a commercial organization judged by profitsand losses and has had no difficulty survivingand even thriving. and the foregone earningsperunit. If. Net military wages for those receivingtrainingare higherrelathe than tive to civilianwages during first periodsbeduringsubsequentenlistment cause trainingcosts are largely paid by therefore. The military offers numberofhoursavailable foreitherconsumption or work. The total cost or priceofconsuming unitoftheith a the good is thesumoftwocomponents: market price or directoutlayper unit.Whi.pi. for enlistments skilledjobs are first-term obtained much more easily than are reenlistments. These principles have been clearly demonstratedduringthe last few years in discussions of problems in recruiting military personnel.

the firmsacquiring the knowledgethan 2.usefulin in otherfirms. althoughhe compares trainingto land-tenure systems. pp.. for a skill cannot be used Resources are usually spent by firms without permissionof the person pos.-Completelygeneraltrain. Much on-the-job search and developmentis often drawn training is neither completely specific since a firmdeveloping a process that not completely general but increases cannot be patented or kept secretwould productivitymore in firmsprovidingit impart external economies to competi.Propertyrightsin because productivity is raised in the skills. Specific. 12 lyzed. Classrooms theFactories[New York: New York University Press. An analogy with re. 19551. cally vested. however.The rest increasesproductivity were to pay trainingcosts.fighter pilots. in Sloan. and the knowledge so is the source of the incentiveto invest in acquired is a form of specific training training and explains why an analogy because productivityis raised more in withunowneq innovationsis misleading.in familiarizing new employeeswiththeir sessing it.op.the scope of specific training are pretitionfortheirservices. are automati.use to civilians:astronauts.formalorientation coursesare quite common.and these innova.paper in this Supplement the "costs of crease productivity by a different 13 To judge by a sample of firmsrecentlyanaThese arguments be foundin Marshall. Trainindustries cannot establish property ing is also offered that is only of minor rightsin innovations. iv). Without trainingthat are extremely usefulin the patents or secrecy.in other firms. can cit. . and missile men all illustrate this to a Patent systems try to establish these greater or lesser extent. Clark and H.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 17 firms competitivelabor marketshave amount in firmsprovidingthe training in no incentiveto provide on-the-jobtrain.firmsin competitive civilian sector. sented before a formal analysis is deThe difference between investmentin veloped.Training that ining because trainedworkers would be bid creases productivitymore in firmsproaway by other firms?Firms that train viding it will be called specifictraining. S.and falls withinthe definition specific of tors. 565-66. such as employmentagency fees. Clearly some kinds of trainingin.militarybut not (much) elsewhere. The previous section discussed genby otherfirms."2 This argumentand analogy would training. of traineesby exactly the same amount the expensesincurredby new employees in firms providing training in other in finding the as jobs (what Stigler calls in his firms.shift eral training and this one will cover costs to traineesand have an in. are supposed to impart external Completely specifictrainingcan be deworkers economies to other firms because the finedas training that has no effect the on latter can use these workersfreeof any productivityof trainees that would be training charge. Other kinds of hiring ing increases the marginal productivity costs. F. by at least as much in other firmsand apply if firms for they would suffera "capital loss" falls withina definition general trainof whenevertrainedworkers were bid away ing. tions become fair game for all comers. training and in research and developThe military offers some forms of ment can be put very simply.than in other firms. Such training rightsso that incentivescan be provided falls withinthe scope of specifictraining to invest in research.as already noted. chap. Firms can. A few illustrationsof training centiveto do so when faced with compe. at least in large firms(see H.on the other hand. This propertyright in skills organization.specifictraining.

MPo is the opportunitymarginalproduct of trainees. werecompletely If all training the wage that an employee could get elsewherewould be independent of the amount of traininghe had received.etc.Expenditureson acquiringknowledge of employee talents would be a specific investment if the knowledge could be kept fromotherfirms.Very strong mightbe completelyinsumonopsonists lated from competitionby other firms. specific.trial and error. In the only to restof this sectionI usually refer on-the-jobspecifictrainingeven though the analysis applies to all on-the-job specificinvestment. If the analysis of completelyspecifictraininggiven in the preceding paragraph was correct. and Wt and MPt are the wage and marginal product in period t. or the time employedin interand viewing. Long-run requiresthat the equilibrium competitive present value of the return exactly equals costs. theyare in human capital because theylose their value wheneveremployeesleave. Accordingto equations (5) and (7) the equilibriumof a firmproviding trainingin competitivemarkets can be writtenas 0~ G = + [E +0-t = ( 11) lvo+C given only whereC is the cost of training in the initial period. then. firms usuallyknowonlya limitedamount about the ability and potential of new employees. checkingreferences. firmswould have to pay training costs. They try to increase their knowledge in various ways-testing. rotation among departments. in The effect investment employees of on theirproductivity elsewheredepends on market conditionsas well as on the nature of the investment. firms extremely competitive labor marketswould face a constant threat of raiding and would have fewer available. in bookkeepingdo not so obviouslyraise the knowledge of new employees. be provided whenever the return-discounted at an appropriate rate-was at least as large as the cost. W would always equal the wage that could . testing. Firms would collect the returnfromsuch trainingin resultingfrom the formof largerprofits and trainingwould higherproductivity. One might plausibly argue. for no rational employee would pay for training that did not benefithim. Even after hiringcosts are incurred. This set is now treated abstractlyin order that a general formal analysis can be developed. On the in other hand. because outlays They are an investment over a short period create distributed effects productivity.-for greater knowledge permits a more efficient utilizationof manpower. and practically all investmentsin their labor force would be specific. Wo is the wage paid to trainees.18 GARY S. althoughnot training. Empirical situations are brought in again after several major implicationsof the formal analysis have been developed. specificinvestments These examples convey some of the surprisingly large variety of situations inthat come underthe rubricof specific vestment. that the would also be indewage paid by firms pendent of training.they are specific on because productivity raised primarily is in the firms makingthe outlays. but investment theytoo are a form specific of in human capital. can be stated more These propositions formallywith the equations developed earlier. BECKER search"). If so. then for productivity would be raised morein the firms than elsemakingthe expenditures where.

Wages would be greater than actual marginal product if some productivity was foregone part of the trainingproas gram. closely depend on the likelihood of labor turnover. Is it more plausible that firmsrather than workerspay for and collectand returnfromtraining? An answer can be foundby reasoning lines. a worker fired after he had paid for specific trainingwould be unable to collect any furtherreturn and would also suffera capital loss.they would be less in the futurebecause the differences between future marginal productsand wages constitute return the to training and are collectedby the firm.MP' = Wo.in fullequilibrium. MPt . If a firmhad along the following paid forthe specific trainingof a worker who quit to take anotherjob.wages equal marginal product. The second is that. Likewise. for expenditure no furtherreturn could be collected.no one suffers from turnover. turnover is ignoredin traditionaltheorybecause it plays no importantrole within the framework the theory. portunity.therefore.But could not one equally well argue that workerspay all specific training costs by receiving appropriately lower wages initiallyand collect all returns by receivingwages equal to marginal productlater? In termsof equation (11). The firstis that the equality between wages and marginal productin the initialperiod involves opnot actual marginal product. just as with general training. even if wages equaled marginalproduct initially.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 19 be receivedelsewhere. Training could raise the new employee's productivitybut would require additional expendituresby the In firm. As a consequence G = C. quit aftercompleting Accordingto our earlier analysis he would have been receivingthe marketwage and a new employee could be hired at the same wage. Suppose a firm paid all the specifictrainingcosts of a worker who it. return the fromtrainingequals costs.In other words. Before claimingthat the usual equality betweenmarginalproductand wages holds when completelyspecifictraining is considered.Any personleaving one firm could do equally well in other firms. It would not matter whether a firm'slabor force always contained the same persons or a rapidly changing group. The of willingness workersor firms pay for to specific training should. or. All ofthisfollows from assumption the that firmspay all costs and collect all returns. its capital would be partly wasted. and since wages and marginal product are assumed to be the same in many firms. Wtwould equal MPt. Since MPo measuresthe marginal product elsewhere and WO would measure the wage elsewhere of trainees. otherwords. of Turnover becomes important when costs are imposed on workersor firms.In the usual analysis of competitive firms.Wt would be the fullreturn t fromtraining in given in 0. and G would be the present value of thesereturns. whichare preciselythe effects specific of training.the reader should bear in mind two points.a firm hurtby the is departureof a trainedemployeebecause an equally profitable new employee could not be obtained. If the new employee were not given training. G would equal zero. To bringin turnover thispointmay at seem like a deus ex machinesince it is almost always ignored in traditional theory. and Wo=MP'-C.and his employercould replace him withoutany change in profits. In the same way an employeewho pays forspecifictrain- .his marginal product would be less than that of the one who quit since presumably training raised the latter's productivity.

Since firmsdo not pay any of completelygeneral costs and costs. To bring turnover into the analysis of specifictraining is a not. employeessome of the they would offer returnfromtraining. of course. with the formerbeing relatively larger the greater the effecton wages in other firmsrelative to the firmsproviding the training.20 GARY S.however.overestimatethe average returnon all trainingexpenditures. while probablyhe could not get half (op.or positivelyreof lated to the specificity the training. positivesince that is presumably liquiditypremium capital investedin specifictrainingis very illiquid (see the discussionin Sec.. The final step would be to shift some trainingcosts as well as returnsto supplymore bringing thereby employees. IV. that salaryelsewhere" cit. by a bargainbetweenthemand theiremployers. When the finalstep is completed firms no longer pay all trainingcosts nor do they collect all the return but they share both with emThe shares of each depend on ployees.such as the cost of funds. and thenhis departimesthevalue turewouldperhapsinjureit byseveral of his salary.) However. he overstressedthe element of indeterminacy in these wages ("their earningsare determined.) Firms could do even better. (12) If G' measures the return collected by employees."4 the relation between quit rates and wages. 15 The rate used to discountcosts and returns is the sum of a (positive) rate measuringthe cost of and a funds. would be the sum of G and G'.and on otherfactorsnot discussed here. Let a representthe fractionof the total returncollected by firms.the othercompletely cific..they mightreduce the likelihood of failure itself by higherwages aftertrainingthan offering could be received elsewhere.'5 If trainingwere not completelyspewould increasein other cific. therefore. In full equiwould equal total libriumthetotal return costs. the terms of which are theoreticallyarof bitrary"[ibid.If G is the presentvalue of the rethe collectedby firms.attitudes toward risk. and desiresforliquidity.p.the specific onlypart of completely would be fractionof costs paid by firms negativelyrelated to the importanceof the general component. (My italics. layoffrates and profits.the use of whichhe could in But in some cases sell at a highprice to rival firms. othercases it is ofa kindto be ofno value save to the businessin whichhe already is. BECKER a ing would suffer loss frombeinglaid off because he could not find an equally good job elsewhere. Since G = aG" and G" = C. deus ex machine but is made necessary by the important link betweenthem. and the wage that could be received elsewhere would also in- crease. in line withdemand.G". and rationingwould be required. spepletelygeneral. turnfrom training fundamentalequation is MP' + G = W + C. productivity firmsas well.. (The reance the loss from turn on "successes"-those remainingwould. Our conclusionscan be stated formally in terms of the equations developed earlier. In effect. C).the total return. or G" = C.. Firms paying for specific training might take account of turnovermerely large return by obtaining a sufficiently from those remaining to counterbalthoseleaving.fn. for the higher wage would make the supply of traineesgreaterthan the demand.Matters would be improvedin some respectsbut worsened in others.]) because he ignoredthe effect wages on turnover. Such trainingcan be looked upon one comas the sum of two components.by that the likelihoodof a quit recognizing is not fixedbut depends on wages.a (positive or negative) riskpremium. . equation (12) can be writtenas 14 Marshall was clearly aware of specifictalents "Thus on and theireffect wages and productivity: the head clerk in a business has an acquaintance with men and things. Instead of merely recouping on successes what is lost on failures. 626).

for employees with extremely specific training and most for those receiving such general trainingthat productivity was raised less in firmsproviding the trainingthan elsewhere.A readermight namely. however. since.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 21 MP'+ or aC = W + C. tends to make age-earningsprofilessteeper and more concave. and or MIP'+ aC+ an = W+C. than employees with no or general training.which implies that quit and layoff rates would be inversely related to the amount of specific Turnoverwould be least training. only the more regular movements are discussed. so a comparisonwith alternativewages gives a misleading impression of the on absolute effect wages of different types of training. not external economies accruing to other firms. G" = G + G' = C +-n n > 0 so that the present would be greaterthan total value ofthe total return costs. none of the earlierequations are satisfactory. Note.afterall.for from quits would preventfirms capturing the fullreturn costspaid by them. unlikegeneral. theseequationswouldbe slightlymore complicated. W = AP'- [ (1-a )C-anl]. etc.since current or futurewages would not be affected. 1 . and equation (14) recollected duces to equation (10).These propositions are as applicable to the large amount of irregular quits and layoffs that continually occur as to the more regular cyclical and secular movements in turnover.Moreover. The part of specifictrainingpaid by employees has effectssimilar to those discussed earlier for general training:it is also paid by a reductionin wages during the trainingperiod. Then G = aG" = aC + an. This view.training would Specific. Rational firmspay generally trained employeesthe same wage and specifically trained employees a higher wage than theycould get elsewhere. if firms all the returnfromtraining. A few major implications of this analysis of specific training are now developed. competition for persons with the latter is apt to be weaker than for those with the former. generaltraining would commanda higher than specific wage relativeto alternatives training does. and (14) reduces to MP' = Wo. that these are external diseconomies imposed on the employees or employers firms of providing trainthe ing. for example. overlooks the fact that general training raises the wages that could be received elsewhere while (completely) specifictrainingdoes not. The part paid by firms has none of these implications. in this section. however. which generalizes the results obtained earlier. Employees with specifictraininghave less incentiveto quit. For if trainingwere completely general.and on layoffs would do the same to employees.a = 1. firms are not too concerned about the turnover of employees with general trainingand have no incentive to offerthem a premium above wages elsewherebecause the cost 16 If G" did not equal C. a = o. if 0 < a < 1. Firms are concernedabout the turnover of employees with specific and a premiumis offered retraining. and firms have less incentive to firethem. of such trainingis borne entirely emby ployees. Suppose. produce certain "external" effects. as they collect in returns. to duce their turnover because firmspay part of theirtrainingcosts. however. Consider a firmthat experiences an unexpected decline in demand for its . P'- (13) (14) I == (1 - a)C.a.that easily believe the contrary.16 Employees pay the same fraction of costs.

for the firm'sinvestmentin his 17 trainingmight be lost forever. The likelihoodof theirtaking Factor of Producter Oi in "Labor as a Quasi-fixed of other jobs would be inversely related. beinglaid off rate of laid-offworkerstends to be significantly to the extentof theirown investment in if greaterthan that of employedworkers. not to There is an incentive.Any workerlaid off would tryto finda new job.D. 18 Actually one need only assume that the quit would be positivelyrelated. the rest of the economy being unaffected. The marginal product of specificallytrained employees initially would have been greaterthan wages.otherjobs. lay off workers with specific training when their marginal product is only below wages. laid-off . trainingseem with specific Thus workers less likelyto be laid offas a consequence ofa declinein demandthan are untrained or even generallytrainedworkers. A declinein demand would reduce these marginal products too. less thanthe initialdifference firmshave no incentive to lay offsuch employees. would the firm just proceed to lay them off until the marginal product was brought into equalitywithwages? To show the danger here. The conclusion here can be briefly summarized.if only because their marginal product were thantheirwage.For sunk costs are sunk. When one firmalone experiences an unexpected decline in demand. If specificallytrained workers were not laid off. A workercollectingsome of the return trainingwould have less infromspecific centive to find a new job when tempothan otherswould: he does rarilylaid off not want to lose his investment.to invest in their training. dissertation. and the larger temporarily a firm'sinvestmentthe greater the incentivenot to lay offsuch workers. tion" (unpublished and thereforethe likelihood of their Chicago).therefore. University Ph.for if it were known that he would not readily take another job.'8 The firmmight be hurt if a new job was found.His behavior while laid offin turn affectshis chances of being laid off. and their employmentwould be reduced to prevent their marginal productivityfromfallingbelow wages.22 GARY S. If thedegreater initially all cline were permanent. only betraining.'7 If the decline in demand were sufficiently great so that even the marginal product of specificallytrained workers was pushed below wages. workerswould be laid offwhen theirmarginal product became less than theirwage and all those laid off would have to find jobs elsewhere. the firmcould lay him off without much fear of losing its investment. in retrospect. sincenothing would bind him to the old one. If the decline were temporary. The marginal product of employees without specific trainingsuch as untrained or generally trained initiallyequaled employees presumably wages.assume that all the cost and return fromspecifictrainingwas paid and collected by the firm.and thereis no incentiveto lay offemployees whose marginal product is greaterthan wages. relativelyfew workerswith specific trainingwould be laid off. but as long as theywere reducedby withwages. no matter how unwise it was. trainedworkersmightnot be specifically laid off even though their marginal product were less than their wage beif cause the firmwould suffer they took is A verysimilarargument developedby Wal. BECKER output. the firm would lose now because marginal product would be less than if wages but would gain in the future the decline in demand proved temporary. for cause the cost ofsearching anotherjob is less for The analysis can easily be extendedto workers.

44-46). The discussion has concentrated on rates. would have a greaterincentiveto A firm trained workersthan lay offspecifically when it alone experiencesa decline beworkers would be less likely cause laid-off to find other jobs when unemployment was widespread.but they have less need to discourage them and would be more willingto pay fortrainingcosts if insurance was provided. 20 In recentyearspensionshave also been an imhas been portanttax-savingdevice. Although quits and layoffsare influenced by considerationsother than investment costs. A pension plan appear with incomplete vesting privileges'9 penalizes employees quitting before reand thus provides an incentive tirement -often an extremely powerfulone not to quit. casual reading of historysuggests that long-termcontracts have.These are important implicationsthan can be tested with the data available.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 23 cover general declines in demand. In other respects the implications of a general decline with are wage rigidity the same as those of a declinein one firm alone.mostplans stillhave incompletevesting(see D. For specifically trained workersinitially receive higher wages than are available elsewhere and the wage rise elsewhere would have to be greaterthan the initial difference before they would consider quitting. suppose. If the decline reduced marginal productbelow wages.Let me assume that wages are stickyand remainat the initiallevel. they were probablyalways the exceptionhere largely because courts have considered them a form of involuntaryservitude.pp. workerswith specifictraining would not be laid off even thoughotherswould be.some ofthese. are more than may relatedto investments strongly at firstblush. Insurance is needed forspecifically trainedemployeesbecause theirturnover would impose capital losses on firms. forexample. a crucialfactorin theirmushrooming 21 The military industryare and entertainment the major exceptions. just as before. 19According the National Bureau ofEconomic to Researchstudyofpensions. .2'and while they have declined in importance over time. only one modification in the previous analysis is required.such as the presence of pension plans. If the decline in business activity were to not sufficient reduce the marginal product below the wage. Firms would be more willingto pay forall kinds of trainingassumingfuture wages were set at an appropriate level-since a contract. a general cyclical declineoccurred. These contractsare seldom used today in the United States.converts all training into comA pletelyspecific training.Thus both the quit and layoff would trainedworkers rate of specifically be relatively low and fluctuaterelatively less duringbusinesscycles. just as pensions do.20 An effective long-term contractwould insure firmsagainst quits. 1960]. Firms can discouragesuch quits by sharing training costs and the returnwith employees. in effect. At the same time pension plans "insure" firms against quits forthey are given a lump sum-the non-vestedportion of payments-whenever a worker quits.whichcertainly growth. indeed. and also insure employees against layoffs. Holland's reportin A Respectfor Facts: National Bureau of Economic ResearchAnnual Report[New York: National Bureau of Economic Research.The effects the inon centiveto investin one's employeesmay have been a major stimulus to the developmentof pension plans.but the same kind of reasonlayoff ing shows that a rise in wages elsewhere would cause fewer quits amongspecifically trained workersthan among others. primarily been a means of inducing firmsto undertake large investmentsin employees.

or 24 Specific. 3).as can be seen examples.In one respect. For experfectly ample. fessional Look at IndustrialTrainingin MercerCounty. be a misleading estimate of what "migrants" could receive. The usual explanation would be that workerswere eitherirrationalor obstacles in movfaced with formidable trainingwere ing. ing. occupation.Such training would tend to be paid by trainees. Rottenberg("The Baseball Players' Labor June. a real possibility.occupation.it is similar to specifictraining. theentry 22 Sometimes firms co-operatein payingtraining thatmonopsony costs.An analysis is of specifictraining. some and perhaps most of the difference must be considereda returnon investment even were there no uncertainty players. occupational. 19591.it would occur even in a perfectlycompetitiveenvironment where there is investment in specific training. so has the ratio of marginalproduct to wages been suggested as a measure of monopsony power. or country"turnover" would be less than average. Althougha discrepancybetweenmarginal product and wages is frequently in taken as evidenceof imperfections the competitivesystem. Does the differencebetween the marginal product and the earnings of majorleague baseball players." JournalofPolitical Economy. .1956. BECKER contractcould Moreover. there may be relatively little worker mobility. if specific24 in differences earningswould important.and countriesand yet firms. and TrainBureau of Apprenticeship [Washington in countries question. measuremonopsonypower or the return on a team's investment?Since teams do spend a greatdeal on developingplayers. try (via migration)than other workers.The investmentapproach proof interpretation vides a very different some commonphenomena.since a singlefirm and could not readilycollectthe return. can vary widely. or geographicallocation. or countryare less likely to occupation. in would be ineffective the United States. p. and it might be rational not to move. except that a firm ratherthan an industry.or counleave that industry. especially when trainingapprentices(see A threatof new leagues. unSince performance happy workerscould usually "sabotage" to operationsto induce employers release them fromcontracts.22 in this respectwould be the same as general training.24 GARY S. not the quality of performance. to the firms. Market.23 about the abilities of different greatly among Earnings might differ industries. for example. of on restrictions entry p.such as have occurredin prois basketballand football. For example. Some trainingmay be useful neither but in a in most nor only in a singlefirm definedby product. However. carpentry training would raise productivityprimarilyin the construcand French legal training tion industry. language and legal with its different institutions.therefore.that is.J. The same resultis obtained for specific training.however. the average Frenchlawyercould not earnthe average Americanlegal incomesimplyby migrat- 23 S. althoughFrenchlawyersearn less than American lawyers.type of set of firms work. But specifictrainingwould also make this ratio greater than one. so their industrial. industries. 254) arguesthatthestrong teams intothe major leagues is primafacieevidence or but poweris important. helpful of the effects ceralso in understanding tain types of "general" training. Workers with training"specific" to an industry. N.and just as the ratio of product price to marginal cost has been suggested as a measure of monopolypower.any enforcible at best specifythe hours required on a job. fromthe following A positive differencebetween marginal product and wages is usually said to be evidenceof monopsonypower. or countryis used as the unit of observation in measuringturnover.

there between learning are complementarities and workand betweenlearningand time. say. of OccupationalLicensing" [paper givenat the Naproductand wages in monopsonies on therefore.specialize in one skill. Schools and firms are oftensubstitutesourcesof particular skills. indusMost trainingin the construction try is apparentlystill best given on the job. betweenmarginal school training large difference mightmeasure. would appear to increase the importance of specifictrainingand the incentive for firms to invest in human capital. like those forbarbers. Sjaastad. therefore.These kinds of monopsony increase the importanceof specific training and thus the inventiveto invest The effect trainingof on in employees. Or to take an example closerto home. 38. But monopsonypower as a whole. themselves this partly explains why young persons migrate whilea formal of statement the principles 25 27 State occupational often requirements licensing permit on-the-job training to be substituted for 26 A relatively "The Economics (see S.persons whohave notyetinvestedin ized and can be formallypresented in a course. The tion specializing in the production of complementarity and schools with firms training. no reason why specific training should be more (or less) important to him. . offerstraining in conjunction with the productionof goods. in largeinvestment employees. combined effectof economic power tional Bureau of Economic Research Conference the Labor Economics. Some schools. and all training. others require prolonged specialization. The shiftthat has occurred over time in both law and engineeringis a measure of this substitution. would be specificto the firm. while others.In acquiring legal skills the shifthas been from apprenticeships in law firms to law schools. SCHOOLING the "art" of in apprenticeship mastering A school can be definedas an institu. by for job alternatives both trainedand untrainedworkersare nil.25 In extremetypes of monopsony. That is. and in engineeringskills from on-the-job experience to engineering schools. Physicians receive apprenticeshiptraining as interns and residentsafter several years of concenin trated instruction medical schools.1960]). and a relatively explanationsee morethan older ones. a research economistnot only spends many yearsin schoolbut also a ratherextensive B. wouldhave an incentiveto migrate. for he would have to invest in learning English and Americanlaw and procedures. Rottenberg.I see no reason why training should have a systematically differenteffect on the foregone earningsof his employeesthan of those in competitivefirms and. Monopsony combined with control of a product or an occupation (due. like universities.empirical and theoreticalresearch.27 Some types of knowledge can be related masteredbetterifsimultaneously to a practical problem. exemplified an isolated company town. including the more extrememanifestations. to anti-pirating agreements)convertstrainingspecificto that product or occupation into firmspecific training.as distinct from a firm that dependsin part on the amount offormalknowledge available. Consider the monopsonistwho pays his workersthe best wage available elsewhere. while the trainingof physicistsrequires a long period of specialized effort.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 25 ing to the United States. also see the paper in this by Supplement L.26 less extreme monopsony positions is more difficultto assess. The development of certain skills requires both specializationand experience and can be had partly fromfirmsand partly from schools.price theory Ofcourse. no matter what the nature.April. For a further my discussionon p.offera large and diverse set.

forotherwiseearningswould as have to be defined the sum ofobserved and foregoine earnings. and foregone earningsare a major cost of high school. BECKER and handlingempirical used in gathering materialsis lacking. In symbols. Our definition studentnet earnings of may seem strangesince tuitionand other direct costs are not usually subtracted from "gross" earnings. (15) where MP is actual marginal product (assumed equal to earnings)and k is directcosts. however. haps this is most apparent when a stucontrolled by dent worksin an enterprise his school. and allow fordepreciationon human capital.mix together accounts. OTHER KNOWLEDGE and total costs. Consistent accounting. unusual transportation and lodging expenses are other. fees. including direct "tuition' costs. in this respectdiffer be treated as a special kind of firmand although at one time many apprenticesalso had studentsas a special kind of trainee. WT= MP-k. ing would steepen the age-earningsprothe incomeand capital file.28 Regardless of whether all costs or merely indirect costs are subtracted schoolingwould from potentialearnings. engineeringskills were initially acquired on the job. college.If MPo is themarginalproduct that could have been received. engineering A studentdoes not work forpay while in school but may do so "after" or "before" school. have the same kind of implications as Thus schoolgeneralon-the-jobtraining. would requirethat earnings of students be defined in the same way. which is particularly important when comparingearningsof personstrainedin schooland on the job.more direct.26 GARY S. primarilyby increasing the knowledge which suggests that a sharp distinction at a person's command. Information is between schools and firms not always 28 Studentsoften have negativenet earningsand necessary:forsome purposesschools can frommost on-the-jobtrainees. books and supplies. Tuition. and over time schoolshave been developed. For example. Moreover." His earningsare usually less than if he were not in school since he cannot work as beThe difference much or as regularly. but as demand develops. that indirectschool costs are implicitly subtracted. . some of the training shifts to schools. and adult schooling.Per. Training in a new industrial skill is given on the job. Note. costs. earningsof on-the-jobtraineeswould be net of all their costs.negativeearnings. or during"vacations. tween what could have been and is earned is an importantand indirectcost of schooling. since firms usually first tend to be the firstto be aware of its value. These relationsshouldbe On-the-joband school trainingare not familiarsince they are the same as those the only activitiesthat raise real income derived for general on-the-jobtraining. introduce a negative relative between the permanent and current earnings of young persons. This supports our earlier assertion that an analysis of on-the-jobtrainingleads to generalresultsthat apply to otherkinds of investmentin human capital as well. which frequentlyoccurs at many colleges. Net earningscan be definedas the between actual earnings and difference directschool costs.equation (15) can be writtenas W = MPo - (MPo - MP + k) = MPO-C (16) whereC is the sum ofdirectand foregone costs and where net earnings are the differencebetween potential earnings C.

An EconomicTheory of Democracy (New York: Harper & Bros.talents.the larger. 1957). I (Fall.pp.30These expenditures lated from competition in the labor in constitutean investment information markethave an incentiveto pay thecosts about job opportunities that would yield of workerscomingfromelsewheresince a returnin the formof higherearnings they have little to worry about in the firms. Presumablymoney systemsee C.: 1860-1885(Cambridge.AmericanIndustryand the wagesare considered be moredispersed to geographi. way of competingneighboring If workerspaid costs and collected the addition. Vol.fact. United States and has been infrequently additional time and resourceswould be used. . Even the system of using one's time to examine want ads. was singularlyunsuccessfulin the newjob requires geographicalmovement. paper in this Supplement.In thanwould otherwise have been received.because some costs-such as an employearningsprofiles. in means of protectingfirmsagainst turnStigler'slanguage by "search. In Westerncountries botha changein employment consumpand quiring 31 For a carefuldiscussionof the contract-labor tion-have tendedto emphasizethejob changemore than the consumption change. although it must be noted be repeated at each job change.added to consumptionrather than demand over resources. pay theirown way. immigrantsalmost always had to ment agencies and situation-wanted ads.firms would be willingpartlyto return. 1958).depreciation. skill. which we have seen is a or talkingto friends and visitingfirms. 1957). Informationabout the political or the fraction costs that have to be paid of social system-the effect of different by workers.not the smaller.an investmentin search would pay forsearchwithina geographicalarea have the same implicationsabout age. today earn30 Studiesof largegeographical moves-those re." is to improve emotional and physical Journalof Law and Economics.If firms paid costs about the wages offeredby different and collected the return. cally than prices.is increased as number of alternatives made available distinctfromknowledge of a particular by a change. Whether workers or firms pay for In both examples information about the search depends on the effectof a job economic system.health. like the sellerswould enable a personto buy from direct costs of schooling. Erickson.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 27 about the prices charged by different that the direct costs of search.29 also significantly found many firmsthat could use their should have been Let us considerin more detail invest.and the ment agency's fee-would be specificto like as general on-the-job trainingand the firm doingthe hiringsince theymust schooling. 94-105). PRODUCTIVE cally discussedin A. are usually the cheapest. Harvard University Press. and One way to invest in human capital in morebriefly my "Competitionand Democracy. of consumption and change on alternatives: the larger the production possibilities. A better job might be portingworkersto the United States. thereby raising his com.and thesefirms ment in information about employment reluctantto pay the large cost of transopportunities.3" Firms that are relatively insuspent in moving." When the over. contractlabor. to parties or social arrangements-could Immigrants the United States usually raise real incomes. or information ducted fromearnings. 29 The role of political knowledgeis systematiWAGE INCREASES D.EuropeanImmigrant. Consider a few examples. In found by spending money on employ. Mass. Downs.search would firms would enable him to work forthe have the same implicationsas on-the-job firm paying the highest (see Stigler's specifictraining..

Of in course. strength influenceon earnings. coffee working breaks. but I would like to discuss a relationbetween on-the-joband "outside" human investments that has receivedmuch attentionin recentyears.28 GARY S.and let a productive wage increase costing an amount C be the only on-the-jobinvestment. would be to offerhigher wages during the investmentperiod than would have since directloans to employbeen offered ees are prohibitedby assumption. absence of any investment. converted into on-the-job investments.emotionalhealth increasingly is consideredan importantdeterminant of earnings in all parts of the world. A declinein the death rate at working ages may improve earning the prospectsby extending periodduring whichearningsare received. The analysis assumes. amount of "outside" investmentcould be financed with borrowed funds.most investments healthin the United States are made outside firms.or steeringthem away from activities with high accident and death rates.If these diet. and medical offices. Total costs to the firmwould be 7r= W + C. The only way to pay. an increasethat raises productivity-"outside" investments are. better and otherfactors. and elsewherestill.32 A firm would be willingto pay forinvestmentin human capital made by emif ployees outside the firm it could beneincreasein producfitfromthe resulting tivity. luncheons. in health that increased productivity more in the firms makingthemwould be a specific investment and would have the same effectas specifictraining. or an improvement conditions-higherwages. can be improved Health.that the capital market is extremelyimperfect. earnings and other income being a major source of funds. hospitals. Firms can invest in the health of employees through medical examinations. and so on-might affectmorale and productivity.A full analysis of the effecton in earningsof such "outside" investment health is beyond the scope of this paper. BECKER ings are much more closely geared to knowledge than to strength. IV.When a firm gives a productivewage increasethat is. in households. The amount invested outside the job would be related to current earningsonly if the capital market for otherwise any was very imperfect.but in an earlierday. however. An investmentin health that into creased productivity the same extent in many firmswould be a general investment and would have the same effect while an investment as general training. Indeed. The cost of on-the-job training is not 32Imperfections in the capital market with rein spect to investment human capital are discussed in Sec. as it were. Before I proceed further.schooling. less is available forinvestments outside the job in health. had a significant Moreover. in manyways. The discussion can be stated more Let W representwages in the formally. therefore. some on-the-job investments would not be undertaken even though theywereveryproductiveby "absolute" standards. like knowledge. and thus in earning capacity. 7rwould also measuretotal wages. and since the investment cost is received by employees as higher wages.a betterdiet adds strength and stamina. "outside" investmentswere more pro- ductive. . D. When on-the-jobinvestments paid are by reducingearningsduringthe investment period. such a conversion is a imperfections natural way to circumvent in the capital market and the resultant dependence of the amount invested in human capital on the level of wages.one point needs to be made.

cit. whichin turn depends on tastes. Hilton. We have shownthat firms would benefitmore fromon-the-jobinvestmentthe more specific the productivity effect. and G the gain to firms vestmentin higherwages. 566).the liberal whichdo not end withhis employer confers benefits For the childrenof his workpeople own generation." Journal of in PoliticalEconomy." Journal of Political LXV (April. knowledge.loweringreportedearnings during the investmentperiod and raisingthem later on.and that a productivity advance raisesprofits more there either because firms have more monopsonypower or because the advance is less delayed.1957]). and the shorterthe labor contract. forthentotal wages (r) would equal the marginal product (MP) when thereis no investment." store or truck systemin nineteenth-century Great Britain has been interpreted as partly designed to prevent an excessive consumption of liquor and other debilitatingcommodities. housing. The price whichhe has paid forlabour will have borne the expensesofproduction an increasedsupplyof of but high industrial facilitiesin the next generation: these facilitieswill be the propertyof others. in share in them. Indeed.conversely.The term MP can represent the marginal product of employeeswhen wages equal fromthe inW. LXV [April. "The BritishTruck System in the Nineteenth Century.35The prevalence of employerpaternalismin underdeveloped countrieshas been frequently in accepted as evidence of a difference temperamentbetween East and West.Leibenstein apparently initially assumeda rapid impactwhen discussingwage increasesin underdevelopedcountries(see his "The Theoryof Underemployment Backward Economies. and opportunities. 35 See G. .. impact-. MP + G = W + C=r. 1957). the benefitwould be less the more general the productivityeffect. a wage increase spent on a betterdiet with an immediate impact on productivitymight well be granted. For example." JournalofPolitical Economny.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 29 receivedas higher wages. An alternative interpretation suggested by our study is that an increase in conon sumptionhas a greatereffect productivity in underdevelopedcountries. so thisformally distinguishes productivewage increase a fromother on-the-jobinvestments. W.the less their monopsonypower. highwages and "Again.who willhave the right hirethemout forthe best price to they will fetch: neitherhe nor even his heirs can reckon on reaping much material reward for this part of the good that he has done" (op. In a later commenthe argued that the impact mightbe delayed ("Underemployment BackwardEconomies: in Some AdditionalNotes.and medical care. Firms might exertan influence spendingby exhorton ing employees to consume good food. (17) Investmentwould not occur if the firm's gain was nil (G = o). Economy. and the longerthe labor contract. Vol.1958]).and growup stronger body and in characterthan otherwise theywouldhave done. Vol. But an investmentin an inthe crease in earnings may have precisely opposite effect. in payinghis workpeople in caringfortheirhappinessand culture.p. An investment in human capital would usually steepen age-earningsprofiles.33 but not one spent on general education with a very delayed The effect a wage increase on proof ductivity depends on the way it is spent. 246-47. or even by requiring purchases of specifieditems in company stores. LXVI [June. raisingreportedearnings more duringthe investmentperiod than later and thus flattening age-earning 31 Marshall discusses delays of a generationor firms moreand notesthat profit-maximizing in competitiveindustries have no incentiveto grant such wage increases. In full equilibrium. the greatertheirmonopsonypower. the company 33 The more rapidtheimpactthe morelikelythat it comeswithinthe (formal de facto) contractpeor riod. In other words "paternalism" may simply be a way of investing in the health and welfare of employeesin underdevelopedcountries.

The cause of this difference is simplythat reportedearningsduringthe investmentperiod tend to be net of the cost of generalinvestments and gross of the cost of a productive earnings in36 crease. the discussion of on-the-job training clearly indicated that the amountinvestedis oftenmerged withgrossearningsinto a singlenet earnings concept (which is gross earnings minus the cost or plus the returnon investment).and the investment period can all be derived from information on net earnings alone. 37 See the papers by Leibenstein.pp. LXVI (June. For investment human capital in usually extendsover a long and variable period."Underdevelopment Backward Econin omies:An EmpiricalComment.and G the gain to firmsfromthe increase in productivity caused by the "morale" effect the inof creasein earnings.37while earlier discussions often stressedthe former.Thus W and MP would show initial earningsand productivity. with a real net earningsstream of Yo duringthe first period. and so on until Yn is provided duringthe last period. on earningsof a change in the rate of return has been difficultto distinguish empiricallyfrom changein the amount ina vested. called age zero.E = MP with a specific investment paid by the firm.a change in the returnfrom a change in the amount invested. The most imis portant single determinant the profitbut the effect abilityor rate of return..op. Economists have long recognizedthat motivationin turnpartlydepends on earningsbecause of of the effect an increasein earningson morale and aspirations.30 GARY S. 529-32. cit. can also show of the effect an increase in the intensity of work "financed" by an increase in earnings. Equation (17). COSTS. AND RATES OF RETURN profiles.. BECKER III. = MP . The general term "activity" rather than occupation or . tors as does the incentiveto grantan increase used for outside investments. In the following. The productivity of employees depends not only on their ability and the amountinvestedin themboth on and off the job but also on theirmotivation. The discussionproceedsin stages fromsimple to complicated situations. then investmentis permittedto be distributed over a knowngroup of periods called the investment period. Finally." Moreover.Marshall stressedthe effect an of increasein earningson the characterand habits of working people (op.we show how the rate of return. 566-69). some rathergeneral relations between earnings.C witha general E investment. First. Thus farlittleattentionhas been paid to the factors determining amountinthe vested in human capital. cit. so the amount invested cannot be determined a from known"investment period. amount invested. RELATION BETWEEN EARNINGS. Many recentdiscussionsof wages in underdevelopedcountrieshave stressedthe latter. and H.38 36 If E represents reported the earnings during investment periodand MP themarginal productwhen thereis no investment. Y1 the next period.or the intensityof their work."JournalofPolitical Economy. which was developed to show the effect outside the firm of investments financed by an increasein earnings. investment costs. Let Y be an activity providinga person enteringat a particular age. They permit one to distinguish. the C increase in earnings. and E = MP + C with a productiveearningsincrease. Oshima.among otherthings. investment is restricted a single period and to returns all remaining to periods. Vol. 38 For example. and rates of return are derived.1958).The incentiveto grant a morale-boostingincrease in earnings. would depend on the same factherefore.

. an activity with a large psychic component. not just onthe-job trainingbut also schooling. When psychic components dominate.The discreteapproach is. X1.the gain from Y could be writtenas d-C=R-C.Yo.40 a rate of discount equating the present where i is the market discount rate.ed. and the total return would be the presentvalue of the differences between net earnings in later periods. me emphasize that essentiallyall my analysis applies independentlyof the division of real earnings into monetary and psychic components. but to simplify the presentation.and morale. would have continuous more elegant formulation variables. (1i)i (I n k (20) The relationbetweencosts and returns can be derivedin a different and.I use investmentlanguage throughout. each period. asvalue of returnsto the present value of sumed for simplicityto be the same in costs.J. . forour the purposes.Ill. ii.The cost of investing in human capital equals the net earnings foregoneby choosing to invest rather than choosing an activity requiringno investment. the cost of choosing Y ratherthan X is simply the difference between their net earningsin the initial period.. the presentvalue of the gain fromchoosing Y would be given by d= V(Y) - V(X) E _ X_ _ (19) i=o ( 1 +i)i0 to Equation (19) can be reformulated bringout explicitlythe relationbetween costs and returns.: Free Press. Friedrichand of Vera Lutz. 1959).withsums replaced by integralsand discount rates by continuouscompounding. By "net" earningsI continueto mean that tuition costs duringany period have been subtracted and returns added to "gross" earningsduringthe same period (see discussion in Sec. The presentvalue of the net earnings streamin Y would be providing a net earning stream of X0.easier to followand yet apyieldsthe same kindofresultsas the continuous proach. is definedimplicitly the equation by 40 A substantialliterature has developed on the betweenthe incomegain and internalredifference turn approaches. . activity with a large monean tary component. and the articlesin The Management CorporateCapital.the language associated with consumerdurable goods might be consideredmore appropriate than that associated withinvestment goods.preferable way by defining = ) V( (18) Oi)+' (I whichis simply internalrate of return. and if R measures the total return. k= YjXj. j = 1. . Xn. If C= Xo. Since many persons appear to believe that the term "investment in human to capital" must be restricted monetary let costs and returns. the internalrate. See. n. "Real" earningsare the sum of monetary earnings and the monetary equivalent ofpsychicearnings.If activity Y requiresan investmentonly in the initialperiod and if X does not require any. for example. Extensions to the continuous case are straightforward. N. however. In other words. If X were another activity r. 39Our discussionassumes discreteincome flows even though a mathematically and compounding. health.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 31 another more concrete term is used to indicate that any kind of investmentin human capital is permitted.witha presentvalue of V(X). Ezra Solomon (Glencoe. .: PrincetonUniversity (Princeton. II). 1951). of chap.information. The Theoryof Investment the Firm Press.. as well as to on-the-job training. Thus the analysis applies to health.

In symbols. and long periods of on-the-job trainingare also common the analysis must be generalizedto cover distributed of The definition an internal investment.V. rate in termsof the presentvalue of net activitiesobviously earningsin different applies regardless of the amount and duration of investment.These equations would be conif werethe siderablysimplified the return same in each period. If investment were known to occur in Y during each of the first periods. 0 and In C1 k I (?rI + r ) 711. j = 1. So the internalrate is also a rate of discount equating the presentvalues of net earnings. ro o could be receivedin period 1. X- of by definition the internalrate. liM C..a simpleand superm ficiallyappealing approach would be to define the investment cost in each of between these periods as the difference net earnings in X and Y. BECKER C +(1-+= r)i' (21) which clearlyimplies n I equate total costs and returns. n.and ifthe return amount Xc 1 = Xi d 1. investmentin human capital is distributed over many periods formalschoolingis usuallymorethan tenyearsin the United States. total costs would approach zero as the investmentperiod is assumed to be longerand longer. or Yj = Xj + k. this reduces to equation (23).YO and kj = Y. O Cl= X. total investment costs as the presentvalue of these and the internalrate would differences. its internal ro CO werethe same in all years. Since. .Xi. . If investmentis restricted a single to knownperiod. C1 since C = XO . Thus equation (21) would become C = k l(+ r r)-n] X (2 3) where (1 + r)-Y is a correctionfor the finiteness life that tends toward zero of as people live longer. If m = 1.4" A second difficulty that the differis ences between net earningsin X and Y do not correctly measure the cost of investingin Y since they do not correctly measure earnings foregone.but the definition in termsof costs and returnsis not generalized so readily.cost and rate of return are easily determinedfrom informationon net earnings alone.42 true cost of an invest41 Since m-1 C1= E(Xj0 n-1 Y-)(I+ O-i. The estimate of total costs requires a priori knowledge and specificationof the investment period.the rate.the period covered by much and otherinvestment on-the-job training is not. I C1 (t +r)i+1 Yj Y (1 +r)i+1 = d=o . however. . A person who invested in the initial period could receivemore than X1 in period 1 as long as the initial investmentyielded a posiThe tive return. . .--ff (2 4) 124 = -_ . Two serious drawbacks mar this appealing straightforward approach. 42 If was the initialinvestment. While the period covered by formalschoolingis easily determined.32 GARY S. . Xj (22) = rn-1 O( I+ r)-. and a serious errormight result froman incorrectspecification:to take an extreme example. 0 (X IYj)(I + r) i=0.

rj the rate of returnon Cj. DifferIn general. 0 (26) Total-cost.43 sons. quite different estimates of total were the same.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 33 ment in period 1 would be the total age of the rates of returnon the indiearnings foregone. Any sequence of tween what could have been received internalrates or investmentcosts is perand what is received.vidual investments.rjCi in each succeedingperiodifthe numberofperiturns.45 If the cost of an investmentis conEquation (25) can be given an intersistently defined as the earnings fore. for it would weigh more heavily thanr does the ratesofreturn earlierinvestments. nor what form the underestimate true costs. no matter what the pattern of between X1 and Y1 could greatly rises and declines.If the number of periods was indefinitely large.as long as the investment periodwas not very long and the systematic difference betweeninternal rates not verygreat. ods was infinite the return and was the same in each. the generalizationfroma single ferentialsbetween Y and X discounted period to distributed investment is Y1is greaterthan XI if actually greatly simplified. or a medical examination. and let the l-1+ O ro) -n-1 returnper period on Cj be a constant orif kj. in investmentoccurredonly in the firstm A proofis straightforward.the same ultimate effect on earnings mined not only fromnet earningsin the wheneverthe average rate of returnand same periodbut also from earningsin the sum of investment costs were the net earlierperiods. 0 m-1 Since the total returnis the sum of individual returns.estinginterpretation all rates of return if gone.The difference betweenthe over-allinternalrate forX and Y and r would be small.however. with k = Ykj being the total return 1 ( + ro) on the whole investment.ent investment programs would have vested in any period would be deter.would equal the capitalized value of returns. and vice versa ifrateswerehigherin laterperiods. The termk r would then costs emerge. 45 Note that the rate of return equatingthe present values of net earningsin X and Y is not necessarily equal to r. indeed. and m-1 Ew1=1. (25) where r= m-1 Wj 0 k= Elkj= 0 m-1 m-1 0 E rjCj= m-1 CE r~ 0 C= C rC Cj c I am indebtedto Helen Raffel important for suggestionswhichled to this simpleproof. therefore.periodj would yield a returnof the amount ki = and internalrates has thestriking. ballet leslarge investment was made in period 1. same. ly simpleformof44 4 44 C= lCj=*. the equation relating costs. investment An in periods. rate the of capitalization being a weightedaver- . re. efinedsimplyas the sum of cost duringeach period.the amount in. be they a college edumightbe greaterthan X1 even thougha cation. Although superficiallya be the value at the beginning the mth of less natural and straightforward ap. let Cj be So the foregoneearningsin the jth period. on For example. The difference mitted.period of all succeedingnet earningdifproach. the difference or be.if rates were higheron investments in earlierthan laterperiods.the over-allrate would be greaterthan i. and if whereC1 is the investment period 1. Yj investments take. an apprenticeship.

. (27) j=rn E ( Yj - Xj) ( = + r)r-1-j (1 + kEr)m-l=c k.47 The value of betweenX and Y m the first differentials must equal the value of all succeeding between Y and X. 0 m-I So costs and the rate of returncan be estimatedfrominformation net earnon ings.. . . rmo-and only m + 1 equations-the m cost definitions the equation and m-1 betweenthe naive defiThe analyticaldifference nitionof costs advanced earlierand one in termsof k= risCi measurestotal 0 earningsis that the former foregone at costs by the value of earningdifferentials the beperiod and the latterby An additionalm . = determinethe 2rn unknowns.. the search forinformation. one need not know the investment period to estimate costs and returns. since r differentials would be the rate of returnequating the presentvalues in X and Y. The investmentperiod of education can be measured by years of schooling. This is fortunatesince the return on human capital is never empirically separated from other earnings and the cost ofsuch capital is onlysometimes and incompletely separated.34 GARY S. otherinvestand ments is not readily available. Cj=Xj-Yj+rJCkO k=o CO=XO-YO. (24) when n = C(1 + r)'-m. (Xim-1 i=0 Yj)(I + r)n-1-i - (c- r Ci)(k 0 + [1 r)m11 = ICj{ 0 m-1 (I1+ r)m-1-i-r + (1+ r) + . co amount invested in each of the firstm periods could be estimated fromthe net earningsstreamsin X and Y alone if the rate of return was the same on all investments. For the internalrate r could be determinedfromthe conditionthat the presentvalue ofnet earningsmust be the same in X and Y. and the amount invested in each period seriatimfromthe relations48 Ci=X1-Yi+rCo i-l <i< m-i.r. O j< and moregenerally Xthen m-1 i=o Yj=Cj- r ECh k-o m. E . since all three can be simultaneously estimatedfrominformation net earnon ings.. r 47 Since.. Xo-Yo= Co .The conditionro = eq.1.46 equal the value also at the beginningof the mthperiod-which is the end of the period-of the first differm investment entialsbetweenX and Y. If activity X were known to have no investment zero investment (a period) the amount invested in Y during any period would be definedby 48 If the rate ofreturn was not the same on all investmentsthere would be 2m unknowns-CO.1 relations co can take. whichfollowsfrom r = . + (1+ r)M-2-i] = Ecj=c. BECKER Total costs would at the internalrate. = rm-1is one form these n. Therefore. The internal rate of return and the 46 Thatis.1 relationwould be requiredto ginningof the investment C' the value at the end of the period. Happily. . by definition. j-1 X1-Y1 = Cl-rCo. but the periodof on-the-jobtraining. and ro. Cm.

The definition costs presentedhere of simplyextends to all periods the definition advanced earlierforthe investment period.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 35 Cj=Xj-Yj+r and total costs by j-1 E 0 Ck. costswere If found to be greater than zero before some period m and equal to zero thereafter. The apparent paradox resultsfromthe infinite horizon. 0 (29) The internalrate could be determined in the usual way fromthe equality between presentvalues in X and Y. C= E Cj =E (xi.the protestruns.49The rationale for the general 49 Therefore. Since earnings in Z would tend to be lower than those in X at younger ages and higher later on. A common objection to an earlier draftofthispaper is that the generaland rather formal definition of costs advanced here is all rightwhen applied to on-the-jobtraining. sincethe value ofthe first earning mn differentials been shownto equal has m-1 ECj () at period m (see n. (28) co C = 1Cj3.That is.and other recognizedinvestments.Indeed. V= Ec 0 In particular.Yj) C-1-i. Although the argument is superficially plausible I am convincedit is as reasonable to say that investmentin human capital occursin Z as in activitiesrequiring trainingor schooling. but goes too far by also including as investment costs many effects that should be treated otherwise. an important virtue ratherthan defectin my concept of human capital is that learning-both on and off job-is included the along with trainingand schooling.the firstm periods would be the empirically derived investment period.For example. total costs could be estimatedfrom value ofall differentials the end of the at the earningperiod. or health.but wheneverf= g. V = 0. f if = f-1 g =CO) V = C. Criticshave argued that there reallyis no investment Z since the rise in in earnings resultsfrom unavoidable learning ratherthan froman attempt to improve skills. so that earnings "automatically" growsrapidlywith experience. schooling. g) = f_1 A (Xji=0 Yj)(1 + r)o-1-i f-1 YCj(l i=0 = + r)-f- Whenf = co and g = 0. Similarly. my approach would say that investmentoccurs in Z. allj.as can be seen from following the equation relatingthe value of the first differentials the beginning the gth at of periodto costs: V( f. 0 0 00 00 Thus the value of all differentials would equal zero at thebeginning the earning of period-by definition oftheinternal rate-and C at the end. suppose that learningwas essentiallyunavoidable in an activityZ.a person enteringan activity requiring much educationis said to value . But costs and returnscan be estimated fromequation (28) even whenthereis no simpleinvestment period. costs in each periodfrom equation (28) and total costs fromequation (29). If Z were preferred X the higher to earnings at later ages presumably outweigh the earnings foregone initially. definition the same: investment is occurs in Y wheneverearningsthere are below the sum of those in X and the incomeaccruingon priorinvestments. knowledge. 47).

to Xi whereXi represents earningsat the ith period past the initialage. The earning stream in an activity with no investmentbeyond the initial age (activityX) would be flat if the developmentalapproach was followedand from earningswere said to resultentirely earlier investment. Without such knowledge. forexample. least in part. BECKER the stream of future higher earnings more than the net earnings foregone initially." The minimum inif vestmentcould then be determined an assumption was made about its rate to return.My discussionof the shape of the earningstream in X is.if one activity was said to require a given investmentand to yield a given return. Thus the lower initial earningsof highschool graduates who enter occupations "with a future"have as much rightto be considered investment. however. 52 But note that empirical evidenceindicatesthat in age-earning profiles unskilled occupationsare very flat.since the private and economic acsocial rankingof different tivitiesdepend only on theirnet earning streams. or information surely must have occurredat youngerages.as do the lower net earnings of those enrolled in college.both from the social and private viewpoints. For example. in the developmentalapproach to child-rearing (discussed in Selma Mushkin's paper). mostifnot all ofthese earnings would be so considered. .only the difference between the amounts invested in any two activities with known net earning streams could be estimated fromthe definitions in equation (28). So much in defenseof our approach.another activity with the same net earningstreammust be said to require the same investmentand yield the same return.is oftena veryclose approximation.although descriptivelyunrealistic.highly and furtherinvestigation conjectural.no matter how they in differ otherrespects. investment can have occurred before that age. If earningswere attributedentirely this investment. = k = rC. the higherearningsinvestmentreturns. in that the total investment any activity can be determined? The statement"nothingis investedin an activity" means only nothingwould be invested afterthe age when information on earningsfirstbecame available.36 GARY S.50 ment would not be zero.k = rC would measurethe return per period. Indeed.52 may well indicate that anotherapproach is preferable. Our assumption that lifetimes are infinite.health.as a return at on the investmentbefore eighteen. the data begin at age eighteen. some investment in schooling.if the rate of returnon some initialinvestment was sufficiently to attracteveryone.logical consistency and economic sense would require that similar concepts apply to learning. The earning stream of persons who do not invest afterage eighteenwould have to be considered. however. To estimate costs empiricallystill has requireda prioriknowledgethat nothing is invested in activity X. In general. If. Were this done for all in available streamsthe investment any activitybeyond that in the activitywith the smallest investmentcould be deterThe observed minimuminvestmined. If the lower earnings due to education are called investment costs. and if costs are related to returnsby an internal rate of return. high A relevantquestion is. I have shown elsewhere on that the average rate of return college education in the United States could 51If C measuredthe cost ofinvestment before the initial age and r its rate of return. therefore: the can shape of the streamin an activityhaving zero investmentbe specifieda priori so 50The techniqueis applied and further developed by Mincerin his paper in this Supplement.

An increasein the lifespanof an activitywould. say after age forty.and the largerthe fractionof a lifetimethat can be spent at any activity. This approach has several importantapplicationsand is used in parts of my study. The income resulting from an investment during periodj would be definedas k-= r3C1 -_(l+.53thereby encouraging such investment. I am indebtedto my wifeforemphasizingIthis point. The influenceof lifespanon the rate of return and thus on the incentiveto invest is importantand takes many forms. a greater effecton the rate of return of investments occurringat later ages. k=j-l rkCk (1 rk+)k-n + TO INVEST (31) o 1 THE INCENTIVE A. forthe lower they are. increase the rate of return on the investment made in any period. One simply applies the concepts developed forinfinite streamsand says thereis disinvestmentin human capital when net earnings above the amountthat could are be maintained indefinitely.j~j X (30) where n + 1 is the earningperiod. to interpret improvementsin the effectiveness and amount of human resourcesin the same way.A few of these forms will now be discussed.however. the longer the expected lifespan.Yj ? IV.including much that has either been ignored or given ad hocinterpretations.a declinein death verseeffect on and sicknesswould reduceratesof return human capital. Economists have long believed that the incentiveto expand and improve physical resources depends on the rate of returnexpected. finite A earning period has.If the adwas sufficiently great.otherthingsthe same. rates of return. They have been very reluctant. Investment at youngerages would give way to disinvestment older ages until no human at capital remained at death (or retirement). . NUMBER OF PERIODS The discussion summarized in equations (28) and (31) showshow total costs. and the investment period can be estimated frominformation on net earningsalone. and the amount invested duringj would be definedby Cj= Xj.An alternativethat is more usefulforsome purposeslets the earning of period itselfinfluencethe definitions accrued incomeand cost. namely. The major secular decline of these rates in the United States and elsewheremay have increased the rates of returnon investmentin human capital. however.This conclusionis independent 53I say maybecause ratesof return are adversely affected the increasein labor forcethat would reby sult froma declinein death and sickness.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 37 only be slightly raised ifpeople remained in the labor force indefinitely. as systematic responses or "investments" resulting in good part fromthe returnsexpected. The number of periods is obviously affected by mortality and morbidity rates.Our attentionnow turns to the factorsinfluencing amount inthe vested in different activities and by different persons. it helps explain why schoolingand otherinvestments priare marilymade at youngerages. and thus how on the effect earningsof a change in the amount invested can be distinguished empirically from the effect a change in of rates of return. An analysis of finiteearning streams can be approached in two ways.indeed. In this sectionI tryto show that an investment approach to human resourcesis a powerfuland simple tool capable of explaininga wide range of phenomena.

The advantage in personsare in school.being young would be less the smallerp of job training. responsibilities.The For example. "The Economics of Information. soldiers to invest in purely military "5 This testwas suggested by GeorgeStigler'sdis.and the like have to ternson the incentiveto investin information (see "The Economicsof Information. difference less tied down by familyresponsibilities. because potentialand thus foregone have less incentive than professional earningsrose with age. The time spent in any one activity is can collect the returnover more years. one witha large p. or related with p. If in the investment be that the young are more interested approach was correct.tion opportunities. but also by the amount of present even if age had no effecton and morbidity capabilities.includingthe would be part of the returnto invest. was. One need not rely would be less of an "investment" than on effects capabilities. and arguments the like are ways to invest in human withp." JournalofPolitical Economy. Indeed.56 temporary migrants ment in human capital raised earnings to urban areas have less incentiveto invest in urban skills than permanentresi54 Younger personswould also have a greater incentiveto investif the cost of any investment rose dents. say. and argumentsbased only on life-cycle or responsibilities. touristsspend ment approach is importantevidence in littletime in any one area and have less its favor. other evidence is needed. BECKER of whetherthe secular improvementin for p periods only. so thisprovidesa powerful test of capital and that youngerpeople have a the importance of the investment apgreaterincentiveto invest because they proach. where p varied behealth itselfresultedfrominvestment. the between the amount invested learning. and.if tween o and n.change jobs and locations. and his paper in this Supplement).if p equaled two years. The size of p would be so. a greater incentive would be determinednot only by age.skills. since the effect age on the rate of and add to theirknowledgeof economic.to invest in marketskills. the secular increasein rates of return affectedby many factors. LXIX [June.more able to absorb new ideas. have less incentive ferentkinds of behavior by the invest. op. A relativelylarge fractionof younger the smallerp would be.38 GARY S. of Althoughthe unification these dif.men and.and atti. cussionofthe effect different of auto-correlation patWomen. tourists. the entire explanation of these differences rate would be the same at all ages except betweenyoung and old personsmay not the two nearest the "retirement"age. as a final example. therefore. returnwould be positivelyrelated to p. and social opportunities. None of the life-cycle seem to implyany correlation that schooling. Women spend less time in the labor forcethan tudes.as we have throughout. training. enterupon on-the. which is not surprising about changing theirroutine since an expenditure with a small p moreflexible and place of living. Vol.mobility. less applicable. . 56 This example is fromStigler. A incentive than residentsof the area to test can be developed along the invest in knowledgeof specificconsumppowerful followinglines.switching between activities. attitudes as soon as on an investmentframework would be one recognizes. political. mortality." cit.rate of obsolescence since the more rapidly an investmentbecame obsolete ment in health.19611. draftees with age.55Suppose that invest. at different ages would be positivelycormore easily supported by parents.

demand. The returnis measured by the complementarity aside.forthe returnon investment tential earningsperiod in any activity.. Note that absolute. op. limited. Keat." JournalofPolitical Economy." ReviewofEconomicsand Statistics."Forces Afficient. IV. Such investmentswould be less the same percentage would not change readily available than more specialized the internal rate. specialists.change the internalrate in the same dicreases it in either one alone. WAGE DIFFERENTIAlS AND findinvestments that increaseproductivSECULAR CHANGES ity in several activities. relative. howdivisionof labor stressesthat efficiency. not only because "practice quately discussed their relative merits. an investment increasing centage change in the return would productivityin two activities also in.Vol.ternalrate of return per unit timeto investment bor force. sometimes by absolute.. extreme rection. . Stiglerand Blank. determinethe return Specialization in an activitywould be income differences discouraged if the market were very and the internalrate. Bk. makes perfect. They are not. "Long-Term Trends in OccupationalWage Differentials. 1950). "Cyclical Variationand Trendin Occupaof tional Wage Differentialsin American Industry size on the incentiveto invest in skills.costs. M. L. cit. XXIII (November.59 the market.the best measureofforces determinlarger the market. Meyers and R.thenaccording my approach it must be treatedalong with otherkinds of learning a way of investingin human capital. incomedifference have greaterincentiveto investin them. H. Vol. thus the incentive to becomemore ef59See A.57but the derived demand for any class of also because a larger market would in. ix. thus the incentiveto specialize Occupational and educational wage are and to investin oneselfwould increaseas differentials sometimesmeasured by the extent of the market increased. while a greater perones-after all. for example. of wages. 57 58 If "practice makes perfect"means that ageearnings profiles to slope upward. Vol.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 39 B. Ross and W. LXIV (May.ing supply.Marginal productivity analysis relates cussions of the division of labor. wage although no one has adeWorkerswould be moreskilledthe larger differences. 1951). as The distinction above between the effect an inof 60 Thus the elasticity creasein themarket practiceand on theincentive on of a substitution usually is to investwouldsimply thattheincentive invest defined thepercentage be to as changein theratioofquanin human capital is increased even aside fromthe titiesemployedper 1 per cent change in the ratio of effect practiceon earnings. 1953). would be directly related to fectingthe Inter-industry Wage Structure. while the con. has been paid to the influence market P. since 1914.therefore.ing only in the amount of their investment. chap. Vol.60 wage ratiosare forces the usual analysis of the moreappropriatein understanding differently. F."Industrial and Labor Relations Review. and ignores the po. VII (October.or a frequenttraveler wants of the return and costsby to be knowledgeablein many environ.absolute income gain.A changein the return ments. LXVIII (December.A womann wants her investmentto be useful both as a According to equation (30) the independs on the ratio housewife and as a participantin the la."Quarmarketsize. Table 11. cit.58 in so Put those of otherinputs. Bell. selves than othersdo. determining and thuswage rates.workersto the ratio of their wages to duce a greaterinvestment skills. Marshall. Goldner.would be greaterthe ever. See. not relative."so oftenstressedin dis. Surprisingly little attention terly JournalofEconomics. op. and mined by absolute wage differences. 1960). or by the absolute betweenpersonsdifferversedoes not hold. in skills and other knowledge is deterwhile ours stressesthat this period. "The Interindustry Wage Structureand Productivity. Bowlby. P.

very far frombeing a fair lottery.therewould be an incentiveto invest more in human capital.If progresswere uniform all and neutralwith respectto all industries and if therewere constantcosts.a narrowing wage ratios has gone hand in hand with an increasing relative supply of skill.The declinein the coefficient variation was shown in his study (op.can be illustratedby consideringthe effectsof technological in progress.8 per cent of in the coefficient variationof wages and an average annual rise of 1.2 per cent in the real standard of deviation. I computed the change in the real standard deviation from data made available to me by Keat. The increased supply would in turn reduce the rate of and increaseof wage differences produce of an absolute narrowing wage ratios. is teryof the law. in point of pecuniary gain. In the United States duringmuch of of the last eightyyears. An alternative interpretation suggestedby our analysis is that the spread of education and the increased investmentin other kinds of human capital were in large part induced by technologicalprogress (and perhaps on other changes) throughthe effect the rate of return. otherliberal and honourable professions. evidently under-recompensed" 61 Some personshave argued that only directin(The Wealth Nations[New York: ModernLibrary. of tend to believe that indirect costs or 1937].6" would be unchanged. firmswould have no incentiveinitiallyto alter theirfactor on proportions. and thus to increase the relative supply of skilled persons. Wage differences. an association that is usually said to result increase the of from effect an autonomous in the supply of skills-brought about by the spread of free education or the to risein incomes-on the return skill. Neutral law and mostotherliberalprofessions now overforegone work "in progressimplies. Since economists costswouldbe unchanged. would rise at the same rate as wages.40 GARY S. Clearly a secular decline in wage ratios is not inconsistent with a secular increase in real wage differences if average wages were rising. measured by wage difas frencesand costs. and.p. therefore. 106). observations earningswould not change.however.as measuredby wage ratios. even foregone workcould not have declinedcontinuously Smith's wereaccurate. BECKER Thereforeneitherwage ratios nor wage the are differences uniformly best measure. includingthe output of industriessupin plyingthe investment human capital. indeed. ratiosbeing more appropriatein dein mand studies and differences supply studies. initiallyall wages would riseby the same proportionand the prices of all goods. Since wage ratios would be unchanged. vestment are earningsrisingalong with wages. and since investment costs would be unchanged.62 even stronger The interpretationbased on autonomous supply shifts has been favored partly because a decline in wage ratios has erroneously been taken as evidence of a declinein the return skill.63the induced approach can be made more plausible by considering trends in physical capital. factors. beThe importanceof distinguishing and tween wage ratios and differences. Economists have been aware that the rate of return on capital could be risingor at least not falling while the ratio of the "rental" price of capital to wages was falling. fromthe practice resulting the confusion of using ratios to measure supply as well as demand forces. the other hand. as well as many perfectly is. Consequently. one importantbody of data on wages showsa declinein ratiosand an rise in differences. and that.While a to decisionultimatelycan be based only on a detailed re-examinationof the evidence.although the rental price 62 Keat's data for 1906-53 in the United States show both an average annual declineof 0. ." the returnto professional if so timeor in the use ofraw materials.the same increasein the compensatedrelative to non-professional of productivity a student's time as in his teacher's point of pecuniarygain. I reproduceAdam Smith's in on "The lotstatement earnings some professions. cit). 63 For those believing that the evidence overindicatesa seculardeclinein rates of rewhelmingly turn on human capital.

RISK AND LIQUIDITY rationalpersonwould inAn informed. Many have argued that attitudes of investorsin human capital are very different fromthose of investorsin physical capital because the formertend to be 65 younger. J. Marshall said: "Not much less elapses betweenthe choiceby parthan a generation and ents of a skilledtrade forone of theirchildren. cit. Since human capital is a veryilliquid asset-it cannot be sold and is rather poor collateral on loans a positive liquidity premium.one uncertainty important determinant of the return. Informedobservation as well as cal- culations I have made suggestthat there about the returnto is much uncertainty human capital. but otherswere such as could not have been foreseen even by the shrewdestpersons and those best acof quainted with the circumstances the trade" (op. cit. Boulding (Chicago: Richard D. a "life-cycle"explanation of attitudes toward risk may be no more valid or necessary than life-cycle explanations of in whyinvestors human capital are relatively young (discussed on pp. "The Utilin ityAnalysisofChoicesInvolvingRisk. The actual returnon human capital varies around the expected return because of uncertaintyabout several factors.Not much need be said about rate.. 67 See M.perhaps a sizable one.66 Were this view correct.However.but a fewwords the "pure" interest are in orderon risk and liquidity. 554).. . Stigler and K. 571). foreseen 65 Note that our argument on p.a human investmentwhich promiseda large return to exceptionally able or lucky personswould be more attractivethan a similarphysical investment. 109). on which some probably threw long shadows before them. are in no period of life moreactive than at the age at whichyoungpeople (op.p. C.. 66 Smithsaid: "The contempt risk of and thepresumptuoushope of success. Indeed. and "the circumstancesby which are are the earnings determined less capable ofbeing [thanthose formachinery]"(ibid.ed. p. at least tempootherdevelopments rarily. Savage. account for the increasein both human and physical capital.and the longerthe average period between investmentand returnthe less such knowledgeis available. G.67 64 For example.).INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 41 of capital declinedrelativeto wages over time. the large secular increase in the amount of physicalcapital per man-hour is not usually considered autonomous. In addition. then. about the returnto thereis uncertainty of given age and abilitybecause a person of numerousevents that are not predictable. 1952). vest only if the expected rate of return was greaterthan the sum of the interest rate on risklessassets and the liquidity associated withthe inand riskpremiums vestment. but ratherinduced by technologicaland that. 38 implied that investorsin human capital would be younger.raised the return. Friedmanand L. People are also uncertain about their ability. Marshall choose theirprofessions" disposition said that "youngmen ofan adventurous are moreattractedby the prospectsof a great success than they are deterredby the fear of failure" (op. and young persons are supposed to be especially prone to overestimate their ability and chance of good fortune.A common explanation based on the effectsof economic progressmay.64 The response to uncertaintyis determinedby its amount and nature and by tastes or attitudes. Irwin.. especially youngerpersons who do most of the investing. cit. The long time required to collect the returnon an investmentin human capital reduces the knowledgeavailable.There always has been considerable about the lengthof life.And meanwhile the characterof the trade may have been almost revolutionizedby changes.p. J." reprinted Readings in Price Theory. 37-38). for requiredis knowledgeabout the enwhen the returnis to be revironment ceived. would be associated with such capital. an alternativeexplanationof reactions to large gains has already appeared. his reapingthefullresultsoftheirchoice. Inc.

and otherhuman capital. migration. latter could be indirectlyfinancedwith "consumption"loans." while the latter require outlays.given the state of the capital market.These will now be discussed briefly. BECKER AND KNOWLEDGE If investmentdecisions respond only to earning prospects. This argumenthas been explicitlyused to explain the "apparent" underinvestment in education and trainingand also.and consequently would tend to invest more than poorer ones.however. or use loan X . health. The importanceattached to capital market difficulties can be determined not only fromthe discussions of investmentbut also from the discussionsof consumption. but just $200 would remainafter paying "tuition. ifall weredirectcosts. the onlydifference beingin the names attachedto loans.the adjusted marginalrate of returnwould be the same on all investments. He ends up withthe and in same consumption investment bothcases. this view can easily be shown to be wrong: opportunity and direct costs can be financedequally readily.000 to spend. capital would be more readily available for specific than forgeneralinvestments.000. underalthoughsomewhatless explicitly. If total investment costs were $800.investmentin (say) a college education would be more migraaffected than in (say) short-term tion. investment in health." so their net position would be exactlythe same as before. forinvestment. there is some reason to doubt whether otherwise in equivalent investments physical capital can be financed much more easily."8 Some other implications of capital market difficultiescan also be mentionrdl 68 A personwithan incomeof X and investment costsof Y (1 < X) could eitheruse X forconsuinption and receive an iinvestmient of Y. to While it is undeniablydifficult use the capital market to finance investments in human capital. and reY ceive a consulmptionloan of Y.The example can be readily generalized and is the obvious inference that indirectand costs are equivalent in directinvestment imperfect as well as perfect capital markets. 4. Economistshave long emphasizedthat it is difficult borrowfundsto investin to human capital because such capital cannot be offered collateral and courts as have frownedon contracts which even indirectly suggestinvoluntary servitude. 3. productivity and wages mightbe related. and if all costs were foregone earnings. 1.The rate of returnon education.Y forconsumption. potentialearnings $1. Since employees' specificskills are part of the intangibleassets or good will as of firms and can be offered collateral along with tangibleassets. Indeed. migration. unless capital limitations applied to conas the sumption wellas investment. they would initially have $1. Since large expenditureswould be to more difficult finance. Internal financingwould be comfamilies wealthier mon. and other human capital is supposed to be higherthan elsewhere. training. CAPITAL MARKETS GARY S.Young persons would consume relatively little. Although superficially plausible.42 D. Some personshave argued that opportunitycosts (foregoneearnings) are more readily financedthan direct costs because they require only to do "without. and some other consumption patterns to would follow only if it were difficult capitalize futureearningpower. who Consideran eighteen-year-old wants to invest a given amount in equipment .because of financingdifficulties and inadequate knowledge of opportunities. 2.investorswould have $200 of earningsto spend. adjusted for risk and liquidity.

thus steepening . sincehe would be untriedand have a high debt equity ratio. Since observed earnings are gross of returnsand net of costs.69 69 Marshall (op.ability.would be less aware of opportunitiesand thus more likely to err than investors in tangible capital. What is his chance of the borrowing whole amount at a "moderate" interestrate? Very slight.He. would eitherhave to borrowat high Although interest rates or self-finance. that investmentin human capital also has an importanteffect on observed earnings because earnings tend to be net of investmentcosts and gross of investmentreturns. it would be made earlierand presumablywith less knowledge than comparable non-human investments. 14-15) that on-thejob trainingwould steepen age-earning profilesand the analysis of Section III generalizes the argument to all human capital. The desire to acquire additional knowledge about the returnand about alternatives provides an incentive to but since postponeany riskyinvestment. of the difficulties financinginvestments in human capital have usually been related to special properties of human capital. machinery for postpone the investment a numberof years until his reputation and equity to were sufficient provide the "personal" collateral required to borrow funds. or time periods are usually said to in result fromdifferences physical capital. SOME EFFECTS OF HUMAN CAPITAL A. an appreciation of the direct and indirect importanceof human capital appears to resolve many otherwise puzzling emabout earnings. A recurring theme is that young persons are especially prone to be ignorant of their abilities and of the investment opportunitiesavailable. I believe. in The eighteen-year-old our example who could not finance a purchase of might. I suggestedearlier (pp. technologicalknowledge.Indeed. I argued earlier (pp. If so. Therefore. moreover.Our analysis indicates. or institutions (such as unionization or socialized production)..pp.the collateral provided by his equipment would probably be very imperfect. an investmentin human capital is more costly to postpone. investors in human capital.Consider pirical findings the following examples: l. cit. Perhaps this accounts forthe tendencyof economiststo when stresscapital marketimperfections in discussinginvestments human capital. too. investment in human capital at youngerages would reduceobservedearningsthenand raise themat older ages. ratherboth mightbe a joint product of the incentive not to delay investing. V.investors in human capital may not have less knowledge becauseoftheirage. 571-73) appears to argue that it is also intrinsically more difficult acquire to knowledge about the returnfroman investment in humancapital. EXAMPLES in Differences earningsamong persons. being younger. Financing may prove a more formidable obstacle to investors in human capital because they cannot postpone their investment so readily. areas. 37-38) that investorsin human capital are youngerpartlybecause of the cost in postponing their investment to older ages. Almost all studies show that agetendto be steeperamong earnings profiles more skilled and educated persons.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 43 he fora firm is startingratherthan in a college education. however. in large measure they seem also to beset comparable investments in physical capital. withouttoo much cost.

intercountry Bagicha Minhas. XXXIII (February. (28) earnings at age j can be to k=j-l Yj = Xj + k=o E rkCk-C1.72 to estimate empiricallythe elasticity of substitutionbetween capital and labor. Kravis. sion in W. I have not.LXIX (June. the directionof bias in this been able to determine case. Vol. and rk is its rate of return.)If. earningswould be at least as steep in V as in X at each age and not only from"younger" to "older" ages if and only if or 4j Yj > AXi - 4' > ACsincerjCj > 0 and This condition usuallysatisfied is the amountinvestedtends to declinewithage. 71 See I.so it obscuresthe and commodity between relative factorsuprelationship plies and factorprices. the regressioncoefficient an estimate of the elasticityof substitution. apparently exports relatively labor-intensive commoditiesand importsrelatively capital-intensive commodities.7" An interpretation consistentwith the Ohlin-Heckscheremphasis on the relafactorsargues tive abundance ofdifferent that the United States has an even more (relatively) abundant supply of human than of physical capital. Usually a ratio of the input of physical capital to the input of labor is regressed areas or time on thewage rate in different being periods.1956).1956). Minasian. "Wages and Foreign Trade. Several recent studies have tried approximatedby 70According eq. cit. highprimarily a regressionof the relative amount of physical capital on wages could give a seriouslybiased picture of the effectof on factorproportions wages. In recent years students of international trade theory have been somethat the United what shakenby findings States. 3.states. Hollis B." (In correspondenceMinasian states thathe intendedto make explicit.44 GARY S. and Robert M. For example.74 72 This kind of interpretation has been put forsee."Factor Proportionsand the Structureof AmericanTrade: FurtherTheoretical and Empirical Analysis. Chenery. regardtimateswouldalwayshave a downward betweenlabor and capiless of the truesubstitution tal. or time periods that have relativelyhigh wages and inputs ofphysicalcapital also tend to have much human capital. . Just as a correlation betweenwages. estimatesin Kenneth Arrow. 7 Minasian's argument (op. For ifwages were because of human capital. on the otherhand. BECKER the age-earnings profile]0 2." Reviewof Economicsand Statistics. Leontief. 73 Interstateestimatesforseveral industries can be foundin J." Reviewof Economicsand Statistics. whereXi are earningsat j of persons who have not at investedin themselves.. Perhaps the most reasonableassumptionwould with be thatphysicalcapital is morecomplementary humancapital thanwithlabor. however. physical capital and human capital seems to obscure the relationshipbetween relativefactorsupplies prices.show up as an apparent increasein labor intensity since earningsare gross of the returnon such capital. forexample.the discusward by manywriters. is the investment age Ck The rate of increasein k. unpublishedpapers by Philip Nelson and timeand Robert Solow containboth interstate seriesestimates. said to have relative scarcity of labor and abundance of capital. Thus export industries might pay higher wages than import competing ones primarilybecause they employ more skilled or healthier workers." and Constant-Output Journalof Political Economy. one study found that export industriespay higherwages than import competingones. 261-70. 1961). 264) that interstatevariationsin skill level necessarilybias his estimates toward unity is actually correctonly if skill is a perfectsubstitutefor "labor."ElasticitiesofSubstitution Demand Curves forLabor." gust. 1961).Vol. "Capital-Labor Substitutionand Economic ReviewofEconomicsand Statistics(AuEfficiency.human thiscondition the substitutes esand physicalcapital wereperfect bias.p.73Countries. XXXIII (November. An increase in human capital would.however. Solow.

or by earningsthemselves. a reasonable firstapproximation would say that if two persons have the same investment in human capital.nepotism. This approach goes too far. results are very relevantand should not be ignored. and a definitionequating ability and earningsipso facto precludes such a distinction.the increase is less than it would be if the cost of human capital. the sum or that of physical capital. Part II of [Chicago: University Chicago Press. " One caveat is called for.and personalitytests while undoubtedly relevant at times. Nevertheless.75 in Instead of onlybenefiting fromactivitiesby others. the one greater who earns more is demonstrating economic talent. persistence.and several otherfactorsconstant. health. Since observedearnings not onlygrossof the return are from investments human capital but also are net of in some costs. (32) tal would both raise and reduceearnings.Indeed. "Education and Economic Growth"in Social ForcesInfluencing AmericanEducation (SixtiethYearbook of the National Societyforthe Study of Education. A secular increase in average earnings has usually been said to resultfrom increasesin technological knowledgeand physicalcapital per earner.INVESTiMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 45 4. Since the public is very concerned about separating ability from education. Economistshave long been aware that conventional measures of ability-intelligencetestsor aptitude scores.77 Persons with higher earnings would simply have more ability than of others. on-the-job training.The main reason for an interestin relating its ability to earningis to distinguish effects from differences in education. Of his many articles see esp. and a skeweddistribution earnings would imply a skewed distribution of abilities. ABILITY AND THE DISTRIBUTION OF EARNINGS An emphasis on human capital not in onlyhelps explain differences earnings over time and among areas but also withinan area. the average earner is made a prime mover of development through the investmentin himself.health. do not reliably measure the talents required to succeed in the economic sphere. amongpersonsor families This application will be discussed in greaterdetail than the othersbecause a link is providedamong earnings.an increasedinvestment human capiin Y = X + rC.ability.in effect. and the incentive to invest in human capital. A compromise might be reached abilityby earningsonly defining through when several variables had been held constant. and other such factors. however. some writershave gone to Accordingly. of psychicearnings. sonality. and otherhuman capital. supposed to benefit is indirectlyfrom activities by entrepreneurs.in the opposite direction.school grades.investors.The average earner.however. Schultz.Anotherexplanation put forward in recent years argues that earningscan rise because of directinvestment earners. W. after the investment period earnings (Y) can be simply approximatedby . and intelligence. 1961]). Although a full analysis would also hold discrimination.Although average earningswould tend to increaseas long as 77Let me state again that wheneverthe word the rate of returnwas positive. and others. Since observed earnings are gross of the returnon human capital theyare affected by changes in the amount and rate of return. like "earnings"appears I mean real earnings. and argued that the the oppositeextreme only relevant way to measure economic talent is by results. The latter requires a particular kind of per- 75 The major figure here undoubtedlyis T. the amount invested in such capital would have to be held constant. training.76 B. was not deducted from of monetary earningsand the monetary equivalent nationalincome.

A second implication is that the separation of "nature from nurture" or ability from faceducation and other environmental for tors is apt to be difficult. 83 See my "An EconomicAnalysis of Fertility"in Demographicand Economic Change in Developed Countries (Princeton.Subsequent attempts have largely concentrated on developing ad hoc random and other mechanismsthat have little probabilistic 82 The first frequently alleged (see.8' a correlationwith several importantimplications. forexample. The Economics Welfare (4th of ed.82 generally selves can be explained withoutrecourse to an assumption that non-economic forcesor demand conditionsfavor them at higher investment levels.so "ability" would be measured by the average rate of returnon human capital. If the distributionof X is ignored for now. that property income is not symmetricallydistributed.46 GARY S. 199. Pigou.J. chap. London: Macmillan & Co. continue their educainvestmorein themand tion.but depends in part on the rate of return. . theaverage ra whererm is the marginal rate. and ea the elasticityof the average rate with respectto the amountinvested.684). C.Persons receivinga high marginal rate of returnwould have an incentive to invest more than others. cit. . 78 Since r is a function C.The ratesrmand ra would be positivelycorrelatedunless ra and l/ea correlated. it easier to invest 79 In addition. is Marshall. ing werewidespread.is in the fieldof personalincome distribution. Marshall argued that business ability and the ownershipof physical capital would be positively correlated: "[economic]forces. N. bring about the result that betweenthe thereis a farmoreclose correspondence abilityofbusinessmenand the size ofthe businesses which they own than at firstsight would appear probable" (op. 1960). Part IV.nor even in a very imperfect capital market rigidlyfixedforany givenperson.79 Since marginaland average ratesare preand sumablypositivelycorrelated80 since ability is measured by the average rate.however. Evidence on the secstudy for the ond is discussed in my forthcoming National Bureau of Economic Research.op. whereC measurestotal investment and X earnr the average rate of return.pp.78 The amount invested is not the same foreveryone.does not directlyhelp explain the skewnessin earnings. At least ever since the time of Pigou economists have tried to reconcile the strong skewness in the distributionof earnings and other income with a predistribution abiliof sumed symmetrical ties..84Pigou's own solution. one can say that abler personswould invest more than others. Y would depend only on r when C was held constant..p. and therefore diswell as directly of tribution abilitywould depend on the amount of human capital. ings when thereis no investmentin human capital.83 The main implication. ii. high earnings would tend to signifyboth more Thus ability and a better environment.theywould find if the marginalreturnand the resourcesof parents and otherrelativeswerepositivelycorrelated. Or reliable estimates of the income elasticityof demand forchildren have been difficultto obtain because higherincome familiesalso invest more in contraceptiveknowledge. 1950). BECKER costs. The end result would be a positive correlationbetween ability and the investment in human capital. negatively were sufficiently 81 This kind of argumentis not new. One is that the tendencyforabler persons to migrate.: Princeton University Press. Some persons might rank high in earningsand thus high in ability if everyonewere and quite low ifeducationand othertrainunskilled.. . Y would indirectly as of the depend on C. between college the earningsdifferential and high-school graduates does not of measure the effect college alone since college graduates are abler and would earn more even without the additional education. cit. 312).. 80 formula to According a well-known r = ral+ ( +1 1 rateofreturn. 84 See A.

1962]).Mincer for a helpfuldiscussionof the comparison symmetrical tively skewed the higher the positive and especially for the stimulationprovided by his his article and the between them.ties. Vol. that the portant difference among skill categories amountinvestedis a function the rate in the degree of skewness. 1958]) conBy definition. In a recentpaper. Cl = Yo earningswould tend to be skewed even when ability and investmentwere not.= rnCn.however." InternationalEconomic Review. personswith n . positive skewnessin earningswith a prethought. that the cost of the nth year of schoolingequaled the earningsof distributed." QuarterlyJournal of Economics. earnings would also be.1 and n years of schoolingwould be kA= Since investment were symmetrically dis. distribution earn.pp."JournalofPoproperties earningsdistributions. cit.and mightbe pioneeringwork.then. and Cn = Y.-l.in education impliesa skewed distribution earnings because he defineseducational investmentby tributionof ability if everyoneinvested school years ratherthan costs." A nnals ofAMathematical VII [March. 1936]. only slightlyskewed (see his "On the Frequency Statistics. therefore.say. 9-10) showedthat the prodwould be moreposiuct of two normaldistributions betivelyskewedthe higherthe positivecorrelation tween them. that and particular.86 If (1 + r) Yn. however..n. again. same as the distribution investment of if The difference between the earningsof persons all persons were equally able. impliesa logof normaldistribution costs and henceby eq. I (May. am indebtedto distributionsis more posi. so ability and investment same distribution (1 + r)n. 88 Craig (op. 1960). in in assuming that everyonewas equally able.normaldistribution schoolingcan be shownto imof tion of earnings would be exactly the ply a log-normaldistributionof school costs. Incidentally. offersan explanation that is sumed symmetrical distribution abiliof not onlyconsistent with economicanaly.3).INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 47 relation to the mainstreamof economic capable. this equation becomes Y. LXVI [August. = tributed. correlation dissertationon which it is based cover a much The economic incentive broaderarea than has been indicatedhere. . a as say of would be positively correlated. C.85 however. quite skewed. C2= Y. a symmetrical. if withn . "The Pareto-Levy Law and the Distribution of Income. Craig has shown that the is large amounts in themselvesdoes seem productof two independentnormaldistributions 85 A sophisticated example can be found in B.if ability were symmetrically schoolingwas the onlyinvestment. and of thus a log-normal distribution earnings. (32) a Now the product of two log-normal quite strongly. a Equation (32) shows that the distribu.87 Q = Yn-1 = Yo(1 + r)n 1 . In conjunctionwith the fractionof total earnings resulting the effect human capital on earnings of 86Jacob Mincer ("Investmentin Human Capital thistenetcan explain severalwell-known and Personal Income Distribution.Vol. If we followMincer the same amount in human capital.4 and therefore ability and investment both varied. Mandelbrot has brought in maximizingbehavior (see "Paretian Distributions and Income Maximization.. r. is assumedto equal r for all n.Y.perhaps normal. earnings would also be. of litical Economy. The smaller of of returnexpected. Since total costs have the than others.and that the skewnesswould be considerablewithhighcorrelations.= Yo(1+r) but the skewness would be small as C3 = Y2 = Yj(1 + r) = YO(1 + r)2 long as the amount invested was statisticallyindependentof ability.cludedthata symmetrical the of of distribution investment of ingswould be exactlythe same as the dis. Mandelbrot.1 years of schooling. reconciling strong of a The approach presentedhere. LXXVI [February. namely. Functionof XY.Yn. C. Our analysishas shown.88 given abler persons to invest relatively 87 For example.that or the cost of each additional year of schoolinginabler personswould tend to invest more creasesat a constantrate. sis but actually relies on one of its Variations in X help explain an imfundamental tenets.distribution schoolyears. Vol. I of distribution earnings.

48 GARY S. The distribution of . SUMMARY AND CONCLUSIONS (-Cj). which tends to be trueat later ages. even thoughabilitywas not. six).89 Equation (32) holds only when investmentcosts are small.X does not really represent in whenthereis no investment humancapiearnings tal. The analysis proceeded froma discussion of specific kinds of human capital.Cj would be an importantdeof of terminant the distribution Yj since investment large at these ages. and if it and abilitywere positively correlated. and lower them at younger ages. with greatest attention paid to on-the-job because it clearlyillustratesand training emphasizes the common effects. the effect specializationon skill. say afterage thirtyfive. are more skewed but also why earnings among older and skilled persons than among youngerand less skilledones. The of would be at least partly distribution X. And since "abler" personstend to investmore than others. a basis is provided for a unified and powerful theory. therefore. Since these common efkinds fectsare producedby verydifferent of human capital. some personsmay earn more than others simplybecause theyinvestmorein themselves. but only earningswhen thereis no investment or twenty-five. Hence is our analysis would predict a smaller (positive) skewness at younger than at 89 As noted earlier. because returnsare added to earnings then.rC. Indeed. the developmentalapproach to childwould be close to zero if arguesthatearnings rearing at therewas no investment all in humancapital. A simple analysis of the incentive to investin human capital seems capable of not only why the explaining. would.Ci measuresthe investment cost of age i. Most investmentsin human capital both raise observed earnings at older ages. afterthe initial age (be it fourteen.rangingfrom in and interareadifferences interpersonal earnings. of over-all distribution earningsis more of skewed than the distribution abilities. The distribution of "unskilled ability. BECKER frominvestmentin human capital-the smaller rC relative to X-the more would the distribution of earnings be dominated by the distribution of X. earningswould tend to be moresymmetrically distributedamong the unskilledthan among the skilled. wherej refersto the currentyear and i to previousyears.the distribution . Hence if abilities were symmetricallydistributed. because costs are deducted from earnings then. to the shape of age-earning of to profiles. therefore.therefore." X. The generaltheoryhas a wide variety of importantimplications. Cj currentcosts.tend to dominate the distribution earningsin of relativelyunskilled categorieswhile the of distribution a product of ability and theamountinvested. For example. since earnings are gross of the return on human capital. before of by determined the distribution investment the initialage.X might be positively skewed.to a general theoryapplying to any kind. The renewed interest in investment in human capital may provide the means of bringingthe theory of personal income back into economics. distribution VI. Net earningsar earlier ages would be given by Yj= X-j+ E rCI+ 0 i-1 olderages because thepresumednegative correlationbetween -C3 and i-1 E 0 riCi would counteractthe positivecorrelation between ability and investment.would dominate in skilled categories. Higher skill categories have a greater average investment in human capital and thus presumablya large rC relative to X. and ri the rate of returnon Ci.

my own work. To take a couple of examples.both on and offthe job. withan emphasison empirical implications rather than on formal generalization. The next few years should provide much strongerevidence on whetherthe recent emphasis placed on this concept is just another fad or a development of great and lasting importance. . training.and that of many otherssupport the view that investment human in capital is a pervasive phenomenonand a valuable concept.Of course. and other activities appear to have exactly the same effects observed earningsas on do education. even though "ability" weresymmetrically not too unequaland ly distributed.INVESTMENT IN HUMAN CAPITAL: A THEORETICAL ANALYSIS 49 of earnings could be very unequal and even skewed. the empirical workreportedin this volume. This paper has concentrated on developing a theoryof investmentin human capital. or incompletely vested pension plans may be a means of insuring firms against a loss on their specificinvestments. empiricalusefulnessis the only justificationfor any theory. industries.and other traditional investments human capital. countriesusing the capital.and although I did not try to bringin even the quite limited evidence on the role of human capital. take anotherexample. We in argue that a relevant concept should cover all activities with identical effects and show that the total amount invested in a generalizedconceptofhuman capital and its rate of returncan be estimated from information on earnings alone.we showed that the well-known greaterun- employment amongunskilledthanskilled workersmay resultfromthe latter having morespecific capital. or These "specific"investments range from hiringcosts to executivetraining and are more importantthan is commonly believed. To learning. in Some investments human capital do not affect earningsbecause costs are paid and returnsare collected by the firms.

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->