secondary workers and the acquisition of negative...

Secondary Workers and the Acquisitionof Negative Human Capital

Joe B. Stevens

Frequent job-changing by secondary workers in a rural labor force is found to beconsistent with the human capital logic, even though negative human capital attributesare knowingly acquired in the process. The present value of future income streams froma "job-changing" strategy exceeds that from a "staying" strategy if the planning horizonfor wood products work is four years or less, which is not an unrealistic length for thisparticular labor force.

The theory of investment in human capitalhas become very much a part of orthodoxneoclassical thought within the past twentyyears. Dating back to the seminal writing ofSchultz and Becker, the published output inthis area is now voluminous. As with mostother broad and exciting theories, however,the passage of time and events is causing areappraisal of the usefulness of that body oftheory. Even though human capital theoryhas been extended to explain crime, child-bearing, and marital instability as economicphenomena, it has also come under attack forfailing to deliver within such conventional(and crucial) domains of interest as the per-sonal distribution of income and earnings.For example, Blaug asserts that causal rela-tionships, let alone functional relationships,are still far from clear despite two decades ofmeasurement on the relationship of humancapital to earnings.

The challenge to human capital theorywhich is of particular relevance for this papercomes from a loosely-knit group of laboreconomists who are associated with an evolv-

Joe B. Stevens is Professor of Agricultural and ResourceEconomics, Oregon State University, Corvallis. Techni-cal Paper No. 5014, Oregon Agricultural ExperimentStation, a contribution to Regional Research Project W-118R. The author is indebted to David Ervin, RichardJohnston, John Edwards, and Ronald Oliveira for helpand constructive criticism.

ing literature on "dual" or "segmented" labormarkets (DLM or SLM' for short). Caincategorizes the dissidents as dual (especiallyDoeringer and Piore), radical (especiallyBowles and Gintis and Wachtel), or jobcompetition (especially Thurow). My focushere is on the dualists, although there isconsiderable overlap among the groups. Inwhat follows, I attempt to (1) identify howone key dimension of the DLM argumentmight be reconciled with human capitaltheory, and (2) present empirical evidencefrom a rural labor force which is consistentwith that reconciliation. In short, I argue thatnegative as well as positive human capitalattributes may be acquired via the humancapital logic, and that recognition of this mayencourage an eventual synthesis of DLM andhuman capital approaches.

The Dual Labor Market(DLM) Approach

This particular challenge to orthodoxy de-veloped during the late '60's, when there wasa growing disillusionment with skill augmen-tation as a means of responding to concernsabout a persistent degree of income inequali-ty. In general, the critics argued that educa-tion and training programs had generallyfailed because neoclassical models of the

73

Western Journal of Agricultural Economics

labor market gave inadequate attention tothe structure of those markets. Dual labormarket theory, on the other hand, rests onthree broad premises about the structure oflabor markets [Wachter, p. 638]. The first isthat the key distinction for economic analysisis not between skilled and unskilled workers,as in human capital theory, but betweengood and bad jobs. The second is that labormarkets are segmented into a primary sectorand a secondary sector, with minimal mobili-ty between the two. As succinctly sum-marized by Zell [p. 8];

"The primary sector is characterized bygood jobs, high wages, satisfactory workingconditions, employment stability, and pros-pects for promotion. The secondary sector,its antithesis, is characterized by bad jobs,low wages, poor working conditions, layoffs,little chance for advancement, and highturnover."

The third premise is that secondary workersdevelop patterns of job instability whichreinforce their entrapment in the secondarysector. To again cite Zell [p. 8];

"According to this theory, while whiteadult males are usually employed in theprimary sector, women, teenagers and, inparticular, minority groups are generallyconfined to the secondary sector. But be-cause secondary firms provide little specificon-the-job training, and because a worker'scurrent wage is unlikely to differ widely fromthat available in a great number of othersimilar jobs, a worker finds little incentive toeither stay on the job or to perform particu-larly well at it. Hence, once a worker is in thesecondary sector, the unstable work environ-ment encourages the adoption of certain poorwork habits: 'casual devotion to job, report-ing for work late or not at all on some days,and quitting without good reason often with-in months of taking the job' [Hall, p. 683]. Itis these habits which most clearly distinguishthe primary and secondary sectors and whichmake movement into the primary sector somuch more difficult. In addition, this viciouscircle is reinforced as secondary-sector em-ployers are unwilling to invest heavily in thetraining of a work force which is prone tohigh turnover, and simultaneously, are lessreluctant to fire a worker in whom they havelittle invested. These factors thus tend toresult in entrapment in the secondarysector."

The labor market as seen by the dualist,then, is one in which earnings are demand-constrained by the structure of labor marketsrather than supply-constrained by insuffi-cient investment in human capital. They seejobs within the primary sector as distributedby custom and internal ladders rather thanby worker productivity, and the total numberof such jobs as largely determined by tech-nology and institutions. Within the second-ary sector, they argue that human capital is inample supply but largely irrelevant to wagedetermination. The most relevant form ofhuman capital, experience gained on-the-jobin the primary sector, is not available tosecondary workers and cannot be purchasedin schools or elsewhere. Their policy conclu-sion, in addition to a commitment to fullemployment policies and reduction of dis-crimination, is that primary sector character-istics should be imposed on the secondarysector through an expanded system ofminimum wages, an explicit system of indus-trial jurisprudence (e.g., unionization), andremovals of exemptions from social welfarelegislation of secondary employers.

The policy focus of this group, their rejec-tion of much neoclassical doctrine, and theirfailure to develop a full-fledged alternativetheory have prompted survey critiques byneoclassical economists, including Wachterand Cain. Both are quite critical of the DLMapproach and maintain that the neoclassicalapproach should be amended, not rejected.Elements of the DLM approach, however,are acknowledged as adding breadth to theneoclassical model. According to Cain [p.1248];

"The SLM economists' theoretical andmethodological criticisms of the neoclassicaltheory are not substantial and are oftenmisguided; nevertheless, a tradition of criti-cism of orthodox economics is sustained, andthis is healthy. The main theoretical contri-butions, which amount to modifications andadditions to orthodox theory, are (1) the ideasof the endogenous determination of attitu-dinal variables among workers, and (2) thehistorical and institutional dimensions of in-ternal labor markets..."

74

December 1980

Negative Human Capital

That is, the general DLM characterization ofsecondary workers is that their entrapment(if it exists) is due not just to the inferiornature of jobs and rewards in that sector, butmore basically to adaptations to circum-stances which further impede their move-ment up the economic ladder. These adapta-tions, as noted earlier, may include casualdevotion to the job, reporting for work late(or not at all), illegal activities, and frequentjob-changing. In short, the DLM thesis isthat worker "attitudes", "motivations","tastes for work", call it what you will, are notsimply given but are adaptations which areendogenous to the system. Moreover, theprospect is raised that these may be rationaladaptations in the sense that others (primaryworkers, employers, etc.) might make thesame decision if placed in the same circum-stance. To cite Cain [p. 1223];

"The interpretation of the SLM hypoth-eses about 'tastes for work' may be ex-pressed, however, in a way that largelyavoids the ideological controversies and thatfocuses on a major gap in neoclassical modelsof labor market behavior. Economists havetraditionally viewed "tastes" as exogenousand as one of the (unexplored) causal vari-ables explaining such labor market achieve-ments as employment, earnings, and occupa-tional achievement. The contribution of theSLM theorists lies not in reiterating thepotential importance of tastes in this role,but rather in pointing out how tastes may beendogenous and a result of one's labor mar-ket achievements. Thus, the effects of dis-crimination, other systematic factors, or evenrandom factors that start workers off in thesecondary sector (that is, in 'bad' jobs), canshape tastes in an antiwork direction andthereby reinforce the disadvantaged positionof low-wage workers. The model has anaspect of the 'vicious circle' or 'self-fulfillingprophecy' to it."

Wachter [p. 679] is also quite critical of theDLM approach, but he too notes the idea ofendogenous "tastes" as a contribution of thedualists;

"In addition, the dualists improve overmost neoclassical models by introducingfeedback effects into their model, althoughthe implications of these effects are not fully

integrated. The feedback hypothesis, an im-portant concept in sociological and psycho-logical models, views workers as alteringbehavior patterns (that is, undergoing shiftsin tastes or preferences) in response toexternal stimuli. For example, secondaryworkers adopt unstable work patterns as aconsequence of distaste for low-wage, dead-end jobs. They then become unacceptablefor employment in the primary sector, and,in a meaningful sense, inferior to primaryworkers - contrary to the dualist unemploy-ment hypothesis that workers in the twosectors are largely indistinguishable in theirskills, human capital, and the like."

This last quotation suggests the notion thatthe systematic "altering of behavior patterns"by secondary workers could also be viewed asthe acquisition of negative human capitalattributes. To this point, human capitaltheorists have considered the acquisition ofattributes which have positive payoffs; theDLM thrust now invites attention to condi-tions under which attributes with negativepayoffs might be systematically acquired.1

The following sections address this issueempirically in the context of a particularadaptation by production workers in Ore-gon's wood products industry. The adapta-tion, job-changing, is not only one of themore tractable behavioral outcomes (com-pared to "distaste for work" and "casualdevotion to job"), it is also one of the moredominant modes of behavior attributed tosecondary workers. In my empirical setting,two questions are asked. The first is, "Arenegative human capital attributes generatedby frequent job-changing?" If an affirmative

1This focus on negative development processes has alsobeen taken by Padfield and Young in their research onsocial marginalization at the Western Rural Develop-ment Center, Oregon State University. They definesocial marginalization as an integral, though negative,aspect of the total economic development processwhereby individuals, families, communities, classes,or cultures become isolated and enclaved relative tomainstream economic society. In their view, mar-ginalization is not an aberration, but is rooted inrational economic decison-making. This present paperarises out of interaction with that research project andits perspective.

75

Stevens


answer is found, the second question is, "Arethese negative attributes generated as a ra-tional response to market signals?" As it turnsout, both answers are "Yes". What is pre-sented below, then, strongly suggests thathuman capital theory (as well as the under-standing of a particular real-world problem)can in fact be improved by applying insightsfrom the dual labor market approach.

Problem Setting

The empirical estimates which followevolved during the course of research under-taken to assess the current economic welfareof Oregon's wood products workers and toidentify problems of future labor-release pre-cipitated by capital substitution and by sub-stantial declines in the availability of timber[Stevens]. Early in the research, it wasestablished that the 75,000 jobs (i.e., averagemonthly employment) in the industry in 1972were actually held by about 110,000 differentindividuals during the course of that year. Itwas further established that the 60,000 work-ers who had remained in the industrythroughout that year (usually with a singleemployer) are relatively secure in terms ofcurrent income and employment levels.They average in their early forties with six toten years of seniority with their currentemployer, they earn substantially more thantheir predicted opportunity earnings outsidethe industry, and their jobs are secure as longas their employer remains in business. In theevent of mill closures, they will be the firsthired by other mills. In the vernacular of theindustry, they are the "lifers".

Another group of 25,000 workers, howev-er, is in quite a different position.2 These areworkers who entered or left the wood prod-ucts industry during 1972; many alsochanged employers within the industry in

2In addition, about 14,500 college students were em-ployed on a seasonal basis during 1972, another 4,500workers left the labor force through retirement ordisability, and another 5,500 could not be accuratelyclassified.

that year. Primarily in their twenties, theiraverage job length is only about one year;most have held even shorter jobs than this.In industry terms, these are the "floaters"who work at less skilled jobs but nonethelessprovide a labor source to meet peak de-mands. Their employers, who express stronghiring preferences for more stable workers,generally view the floaters (but not theirwork environment) in DLM terms (e.g., highturnover, casual devotion to jobs, distaste forwork, etc.).

The existence of these two groups, then,tempts one to label them as primary andsecondary workers. To posit a strict one-to-one relationship between an imprecise theo-retical construct and the real world, howev-er, is obviously hazardous. Cain [p. 1231]notes, for example, that the DLM literaturehas been rather silent on discussion of ex-plicit criteria by which workers might beassigned to primary or secondary sectors. Ifthere is, in fact, a general duality to occupa-tional structure, means of assignment arecertainly not yet agreed upon. One can onlyargue, as I do, that there is a rough corre-spondence between the key DLM constructand reality, that these 25,000 younger, lessskilled job-changers at least have a markedresemblance to the secondary workers ofDLM theory.

The key issue, on the other hand, is notthe means of assignment or that of groupmembership, but whether job-changingmakes sense in the situation faced by job-changers, most of whom appear to be "sec-ondary" workers. 3 The implicit human capi-

3To be researchable, of course, the domain withinwhich it "makes sense" must be specified. One type ofadaptation is a dead-end street for analysis; someworkers find little satisfaction in mill jobs and will oftentake a lower paying job outside the industry to escapethe monotony of the mill. Little can be said about therationality of this type of decision; if it makes sense tothe worker, then it must make sense. Job-changingwithin the industry, on the other hand, is moreamenable to analysis. Discriminant analysis of intra-and inter-industry job-changers revealed that the for-

76

December 1980


tal assumption is that job-changers generateonly positive human capital; the contrarycase seems not to have been considered. Inthe reality of the Oregon wood productsindustry, however, job-changing does in factstigmatize the worker who is seen by em-ployers as violating social norms [Young andStevens]. If this type of adaptation doesreinforce one's status as a "secondary" work-er, as presumed by the DLM literature, thenthere must be some reason (besides masoch-ism) which triggers this action. For thosefrequent job-changers who are motivatedprimarily by income rather than work satis-faction, it could well be that the magnitude(or the timing) of the net earnings differentialfrom job-changing more than offsets thepenalties they might incur as a result of thisbehavior. 4 If this is the case, negative humancapital attributes are systematicallygenerated through rational adaptations tomarket circumstances.

Present Value Model

The above hypothesis can be tested, atleast for job-changing within wood products,by a relatively simple model for estimatingthe present value of future income streamsfor two alternative labor market strategies.The first strategy is to change jobs on a yearlybasis, which is modal behavior for the groupof 25,000 "secondary" workers. The secondstrategy is to not change jobs, or in otherwords, to stay with the current employer.(This is an option which was obviously notchosen by this group, at least during 1972.)

mer were more likely to find job satisfaction in millwork and that their decisions were far more likely toreflect monetary rather than non-monetary considera-tions [Stevens, pp. 140-146]. This suggests the appro-priateness of a present value model for evaluatingintra-industry job-changing.

4The DLM literature suggests that the net earningsdifferential from job-changing would be zero in a"pure" secondary sector; i.e., there would be nochance for advancement. Again, this is a polar theoreti-cal construct; in reality, some gains may be obtainable.

The present value (PVc) for "job-changing"would be:

J (1 p) n-p [Yo + (n)(AYc)](1) PVc =n

n=1 (l+r)n

whereYo = current earnings ($)

AYc = earnings differential from job-changing, assuming that a job canbe found ($)

p = reduction (due to frequent job-changing) in the probability offinding another job5 (0<p<l)

n = year (1, 2, ... j)

j = end of planning horizon

r = discount rate.

The present value (PVs) for "staying" wouldbe even simpler, since penalties for job-changing are not involved. That is,

( Yo + (n)(AYs)(2) PVs =

n=l (1+r)nwhere

AYs = earnings differential from remain-ing with the current employer foranother year ($).

Three parameters need to be estimated toimplement the model. The first, p, shouldreflect the (negative) impact of the negativehuman capital attributes which are acquiredby frequent job-changing. The second, AYc,is the gross earnings differential from job-changing. The third, AYs, is somewhat moresubtle in that the shadow price of "firm-specific" experience must be estimated.These tasks are dealt with in turn.

Penalties For Job-Changing

Early in the research, informal interviewswith workers and employers indicated that agreat premium is placed on worker "stabili-

5 Another alternative is to include (p) in the denominatoras additional discounting for risk, i.e., (1 + r + p)n.This procedure yielded results which were even morefavorable to job-changing than those shown here.

77

Stevens


ty" in the hiring process, even for unskilledentry-level jobs, and that instability reducesthe likelihood of getting a job. Perceptions ofrelevant hiring criteria for entry-level jobs(where most job-changing occurs) wereelicited on an individual basis from a group ofpersonnel managers and unemployed work-ers and ranked in order of importance bythese individuals. Personnel managers andworkers alike perceived that the hiringprocess favored those with stable work his-tories as measured by such criteria as "lengthof time worked for last employer", "reasonfor leaving last job", "whether or not fired byprevious employer", and "number of em-ployers in the past five years" [Young andStevens].

In order to verify the accuracy of theseperceptions, statistical analysis of actual hir-ings was also conducted by use of data fromjob application forms on file at four of thelarger firms whose personnel managers hadbeen interviewed. The sample consisted of46 workers who had recently been hired forentry-level jobs and 89 applicants who hadbeen considered but not hired. In view of therather large number of criteria (24) which hadbeen identified by personnel managers andworkers, discriminant analysis was used tosort out those criteria which best differenti-ated between workers who were hired andthose who were not hired.

Finally, the data set (now reduced in scopeby the discriminant analysis) was subjected tologit analysis since the goal was to isolate theeffect of job-changing on the probability ofgetting another wood products job at somelater point in time. According to the bestestimate (Equation I, Table 1), an additionalwood products job in one's work historywould reduce the probability of future (woodproducts) employability by about 14 percent.Extremes in job length were also important;decreases in the minimum length of jobwould reduce employability by 5 percent peryear at the margin (or about a 1/2 percentreduction in employability for having held a 6month job as contrasted to a 7 month job).Inter-industry job changers were also

78

penalized; those applicants who were at thattime in wood products had a 15 percentgreater chance of being hired than those whowere outside the industry.

Three conclusions emerged from this por-tion of the research [Young and Stevens].One is that workers are, in fact, penalized foradopting a "job-changing" strategy. The sec-ond is that positive effects of conventionalhuman capital attributes (education, experi-ence, health) are conspicuously absent fromTable 1, having been eliminated by thediscriminant analysis. The third is that thepenalties for job-changing arise less fromeconomic reasons than from social normswhich the more "stable" personnel managersapparently seek to impose on the pool ofyounger, less stable job applicants. Whilethis latter conclusion can be disputed, onlythe existence of penalties is important forimplementing the present value model. Ac-cordingly, the parameter p of Equation 1, thereduction in future employability associatedwith a "job-changing" strategy, is taken to be14 percent per job change, which occurs, forthese workers, on an annual basis.

Earnings DifferentialFrom "Job-Changing"

The expectation of gain from intra-industryjob changing (AYc) is essentially an empiricalquestion. Fortunately, this information wasavailable from interviews conducted with astate-wide random sample of 99 "primary"and 46 "secondary" workers in 1972. Fromwork history data, mean annual earningsdifferentials (for voluntary job changes) of$1,755 (87 cents per hour) and $860 (45 centsper hour) were identified for primary andsecondary workers, respectively. 6 Neitherconfidence interval contained the mean of

6 Most job changes were made within the three yearsprior to 1972. Historical unemployment rates of 3 and 7percent, respectively, were used in extending the dataon hourly wage increases into annual earnings differ-entials for primary and secondary workers.

December 1980


TABLE 1. Effects of Worker Instability on the Probability of Being Hired in Wood Products.a

Logit Modelsb Linear Modelsc

I II III IVIndependent Variable (n =79) (n = 53) (n =79) (n =53)

Number of wood products -. 1423 -. 1229 -. 0637jobs in the past (.0578) (.0753) (.0464)Veterans - .2864 -. 1447 -. 1594

(.1327) (.1466) (.1042)Length of shortest job .0543 .0226 .0500 .0356held in the paste (.0292) (.0298) (.0198) (.0251)Last job was in wood .1518 .0679 .2788 .2178productsd (.1250) (.1439) (.1128) (.1349)Complete application .1186 .0947 .2872 .3011formd (.1226) (.1347) (.1190) (.1419)

Adjusted R2.188 .095

Likelihood ratiof <.01 <.15Constant .0605 -. 0918

aStandard errors are in parentheses. All 79 workers had prior work experience in wood products; equations II andIV exclude 26 observations from one firm which discriminated against veterans, apparently for idiosyncraticreasons (Stevens, 1978, pp. 94-101).

bEstimates show effect (on probability of being hired) of a unit change in an independent variable (Xk); that is,ap

= r3kP(1 - P),aXk

where 1k is the logit estimate and P is the proportion of the sample which was hired (.34 in I, .32 in II) (Hanushekand Jackson, 1977, pp. 188-189).

CEstimated for comparison and to derive R2 values, even though the regression coefficients are subject to biasbecause of the discrete dependent variable (1 = hired, 0 = not hired). Equation IV estimated after deletion ofvariables which were non-significant in III.

dDiscrete variables (1 = yes, 0 = no).eln years.'Tests the hypothesis that all parameters are zero; asymptotically distributed as chi-square.

the other group (a = .05), thus tending tobolster the argument for duality of thisparticular labor force.

Earnings Differential From "Staying"

Whereas the earnings differential fromjob-changing can be quantified directly fromthe data, that which arises from the "staying"strategy must be estimated as the (partial)value accruing to another year's experiencewith that employer. To accomplish this,Ervin and Stevens and Ervin used cross-sectional income determination models toestimate the productivity of various humancapital attributes, including firm-specificwork experience (i.e., firm seniority) in wood

products. A priori, this latter magnitudewould exceed that derived for other forms ofwork experience only to the extent that firm-specific experience is necessary to performproduction-line tasks in this industry. If allforms of work are completely substitutablefor each other (all on-the-job experience is"general"), then the partial effects on earn-ings of each form of work experience shouldbe the same.

Most cross-sectional income determinationmodels use only a single equation and definethe dependent variable as hourly, monthly,or yearly earnings [Rees and Schultz andStoikov]. While these approaches may beuseful for predictive purposes, two problemsexist with respect to making structural infer-

79

Stevens


ences about the marginal returns to specifichuman capital attributes. Both problems aredealt with below, although the first is re-solved more satisfactorily than the second.The first is that labor earnings depend onboth remuneration per unit of time and thenumber of time units that the worker isemployed. Low earnings can arise from lowwages, frequent unemployment, or both.Moreover, the productivity of various humancapital attributes and/or the effects of institu-tions (e.g., the seniority system) may vary inthe two cases. 7 In the regression estimateswhich follow (Table 2), weekly earnings (W)and the number of weeks worked in a year(Q) (and hence, total yearly earnings (I)) arehypothesized to be determined by variableswhich embody human capital attributes ofworkers (Xi) or which otherwise affect thesupply of or demand for labor (Zi). 8 That is,

(3)

(4)

(5)

w = fi(Xi, Zi)

Q = f2(Xi, Zi)

I=W Q

(6) I = [fi(Xi, Zi)] [f2(Xi, Zi)]

If equations (3) and (4) can be estimated, it isthen possible to derive "marginal earnings

coefficients", aI-, which show the relation-ax 1

ship to yearly earnings of particular humancapital attributes. That is,

af1 af2(7) I/aX, = -af . Q + . W ,oraxi axi

7The effect on earnings of additional firm-senioritycannot be attributed solely to worker productivity ifcustom or union contracts mandate that more senioremployes have greater protection from lay-offs.

8Weekly earnings would depend upon both hourlywage rates and length of the work week. In theory,choices between income and leisure may affect thelatter. The assembly-line nature of production mili-tates against such choices in this case.

80

(8) aI/axi = W Q +Q. waxi axi

aw Q awhere - and are partial regression co-axi axiefficients from equations (3) and (4), and Qand W are selected values for appropriategroups of workers. The first term of equation(8) reflects the expected increment due tohigher weekly earnings, holding constant thenumber of weeks worked. The second term isthe increment due to an increase in numberof weeks worked, holding constant the levelof weekly earnings.

The second problem with using cross-sectional wage or unemployment regressionsto make structural inferences is that they aregenerally reduced-form equations [Blaug].Education, for example, may cause a workerto be more valuable to a wood productsemployer, but it may also cause him to beless likely to supply his labor to that industry.Either a rightward demand shift or a leftwardsupply shift (or both) would be consistentwith a positive relationship of education towage rates. This reduced-form problem canbe lessened (but not eliminated) by appropri-ate specification of variables and by carefuldefinition of data sets. As to variable specifi-cation, measures of different types of workexperience and other human capital variableswere available from the life-time work his-tories of 189 workers in the state-wide ran-dom sample. In addition, a variable wasincluded in the weekly earnings equation toreflect the predicted opportunity earningsoutside the wood products industry. 9 The

9Weekly earnings in non-wood products jobs wereregressed against age, education, vocational training,disability, total wood products experience, total non-wood products experience, experience in two specificoccupations and industries and four dummy variablesfor specific occupations and industries [Stevens, pp.67-70]. Squared terms were also included for age,education, total non-wood experience, and total woodproducts experience. Twelve (of sixteen) variableswere significant at ca S .05, with R2 = .50, n = 264jobs.

December 1980


appropriate interpretation of the coefficientitself will not be clear, since the variablecould reflect both supply and demand con-siderations. That is, those with higher oppor-tunity costs might be less likely to work inwood products (supply); they might also bemore productive workers in wood products(demand). In any event, the intent in specify-ing this variable is to allow supply considera-tions to surface and to better isolate thepartial effect of firm seniority.10

As to definition of data sets, the equationsfor weekly earnings (excluding fringe bene-fits, for which data were not obtainable) wereestimated from all the wood products jobs inthe lifetime work histories of the entiresample in order to allow the full range of datato be utilized (I, II, and III in Table 2). Theequations for explaining weeks worked (IVand V), however, utilize only data on thoseworkers who attempted to work full-time fora single wood products employer in 1972.That is, those with voluntary unemploymentperiods were excluded, as were those whoalso worked for another employer during thatyear. 11 The interpretation of coefficients,then, is that they reflect the expected out-comes for those who would attempt to remainwith a single employer, i.e., for those whowould follow a "staying" strategy.

Both the regression results (Table 2) andthe marginal earnings coefficients (Table 3)reveal that earnings are indeed influenced by

"0Although not shown here, deletion of predicted oppor-tunity earnings from the equations for weekly earnings(I, II, and III in Table 2) resulted in only a very slightincrease in absolute value of the other coefficients;very little change occurred in relative magnitudes.

lHad these workers been included, the statisticalresults would have a different interpretation. Workerswith little firm seniority, for example, might anticipatea lay-off by taking a job with another firm. In this case,the dependent variable (weeks worked) would reflecthow successfully they can execute a job-changingstrategy. While this is an interesting question, theimmediate focus is on implications of a "staying"strategy.

a variety of human capital attributes. 12 Forthe purpose at hand, however, we areprimarily interested in the size of the margin-al earnings coefficient for firm seniority,which should be an approximation to theexpected differential from the "staying"strategy. As indicated in Table 3, theseamounts are quite small at $170 for plywoodworkers, $77 for loggers, and only $11 forsawmill workers, or roughly $100 per yearover the entire industry.13 Moreover, theseamounts are not always larger than themarginal earnings coefficients which are de-rived for other forms of work experience,within or outside the industry. 14 Only in thecase of the plywood sector (with a morecomplex production process) is firm seniorityrewarded more highly than other wood prod-ucts experience. Even in this case, thedifferential for "firm-specific" rather than"general" experience amounts to only $65per year (about 3 cents per hour), which is

12A priori expectations included (a) positive returns tothe individual types of work experience, (b) diminish-ing returns to age, education and total wood productswork experience, and (c) substitutability between firmseniority and other wood products experience due tothe simplicity of many production line jobs.

13Additional support for this approximate magnitudecomes from a regression analysis of changes in weeklyearnings by 52 voluntary job-changers within woodproducts (R2 = .19). The (highly significant) partialregression coefficient for "years of wood productsexperience" was $2.16, which converts to $112 yearlyfor a fully employed worker [Stevens, pp. 146-148]. Inother words, workers do become somewhat morevaluable as they gain industry experience, but thisincrement can be captured equally well by job-changing or by staying with the current employer.

'4If these estimates reflected only demand considera-tions, one would not expect that firm-specific experi-ence would be less valuable at the margin than generalexperience. The explanation for this occurrence inlogging and sawmills appears to be that supply consid-erations have not been entirely removed. This isconsistent with observations by personnel managersthat more senior workers are often content with theircurrent jobs and do not "bid" on internal job openings(above entry level) which involve more pay and moreresponsibility.

81

Stevens


TABLE 2. Regression Analyses of Weekly Earnings (Work History Data) and Weeks WorkedDuring 1972.a

Independent Variables

General:

Age

Age (squared)

Education (yrs).

Education (squared)

Vocational trainingb

Disabilityc

Work experience (yrs.):

Firm seniority (current employer)

Other wood products firms

(Firm sen.)(other wood prod.)

All wood products (squared)

Non-wood products

Opportunity earnings:d

Job-specific:

Years agoe

Frequency of training f

Capital/labor ratio g

Weekly Earnings Weeks Worked

I 1 III IV VLogging Sawmills Plywood Logging Sawmills & Plywood

10.61(3.05)

-. 119(.004)

-. 40(.16)

.086(.030)

-2.99(2.02)

-5.93(.97)

28.41(13.18)

.010(.005)

20.97(9.19)

-. 72(.36)

- .99(.34)

46.44(21.73)

5.53(1.33)

3.73 4.49(1.32) (1.52)

2.78(1.21)

.32(.16)

-. 193(.006)

4.52(1.02)

.157(.008)

.008(.005)

-. 073(.049)

-. 87(.41)

-. 72(.38)

.054(.030)

-. 72(.46)

-2.15 -6.23(.74) (.66)

.062(.032)

Union membership h -3.45i ~ 'i ~ *i ~ (2.58)

Plywood job .69i i i i (.45)(.45)

Constant 31.81 -14.68 151.59 49.51 49.37R2 .372 .478 .552 .194 .220n 164 130 153 35 57

aStandard errors are in parenthesis. Coefficients are those computed after deletion of non-significant variables(which are left blank).

bWeekly earnings: months of vocational training in high school. Weeks worked: months of vocational training inhigh school and other types of vocational and on-the-job training.

CHours and/or types of work limited (1 = yes, 0 = no).dPredicted non-wood products earnings at time of wood products job."Time elapsed since end of job (means: logging = 4.3 yrs., sawmills = 4.4, plywood = 3.7).'Supervision received on at least a weekly basis (1 = yes, 0 = no).gAppraised value (000) of structures and equipment per worker, in thousands of dollars.hl = yes, 0 = no.Variable not included in regression.

82

December 1980


0

CCi

CIO

(aJ

(0c.'Jc~J

I

CM CM qCO) Co 0

N 0 CV)

O LO

(0 (0

( o co

cg0 8

c

00I

(D

CM 0)

00 (<CO 1<N C06Nq 8

o 3

0 0o Co

c0Jo0o

c1(\j(0

00

co 0

I CO

0

CO

co oLO00 O COCO 0, ,-Cv

cj

L(O0)

6s

0D

LOI

CD CD0) Lt0cv 't

o 00 o0( CD (

o Cj oo t .o> in

co

C > 0 Z

0 -§o a a2

xU)

w > oio z

Stevens

En

o

oa)' E

ao

a) 0

- 0)

) aa)cnC

onD0)

OC

a)

a)" a)0 0)

c.C 3Co rC>- r-

-I0~

U)

Cu

0d

CTC:.0)0)0-j

0

ro

0

._

0)0)0-J

0

0

CL

0,

E

0)

0-j

0aCO

a)CO -E -

c-

EU)toU)

0ca

a

Cr

oa)

Cu

0)a)

a) >

(.'

0

0) tCD

,

(0 tO

Ea) ,

E 0 ) CI-C) a

' O )

Ta oa

na) -

0c n I

C\1 cU)

cpu

0 -

0.E oC)

co 0

Cf) Q0

a 00

a) C

-0toCX

a CZ a,C,)

-0

0 0 oXo '.

C

0)

, t

CI'oaI -_ '

O c

_ocu

0.^

I-.. II

C W

a

- C )

W) IN

I a) -i o

a>

a0)

CZ

a

Ct

0

0)

a)0

a)a

U)a) ED CZ

(

0)

tog -i

E 0

0 cmW E

0> Cua

C N

"o 0Cr ..C0

II

0 CO

CD

(0Ch

C.)cl

C)0

0(I)0)

0)wc

w

a-

I-i

3=.jm

83

1.4

1

I

I

I

II

II

I

N

1

4

II

I

II

II

II

I

10


not a terribly large disincentive to job-changing.

Present Value Estimates

To this point, it has been documented thatsecondary workers in wood products canexpect increments to earnings from either"staying" ($100 per year) or "job-changing"($860 per year), but that the latter strategyreduces one's likelihood of future employa-bility within the industry by about 14 percentper year. This dilemma can now be quan-tified by solving equations (1) and (2) for thepresent values associated with the twostrategies. First, however, several qualifyingconsiderations should be examined. One isthat the true AYc value may be somewhatdifferent than that which was estimated ($860per year). For one thing, all secondary work-ers probably do not have the alternative ofgaining 45 cents per hour through annualjob-changing, even though this mean valuewas recorded for actual job-changers. Ac-cordingly, present values from "job-changing" (Table 4) are also computed for an

increment one standard error smaller thanthe mean value (i.e., $610 per year).

Secondly, by what interest rate should theincome streams be discounted? Havemanand Krutilla used a weighted rate whichreflected both marginal borrowing and lend-ing rates, by income class. My impression ofthis particular population is that they arebasically borrowers, not lenders. According-ly, a rate of 14.5 percent was computed torepresent the (weighted) average rate paid byU.S. consumers in 1972 on various forms ofinstallment credit. 15 As it turned out, thepresent value estimates were quite insensi-tive to the discount rate; use of an 8 percentrate changed the comparative values onlyvery slightly.

15By type, the credit was for auto loans (35 percent),personal loans (30 percent), other consumer goods (30percent), and repair and modernization loans (5 per-cent). By sources, the creditors were commercialbanks (46 percent), finance companies (26 percent),credit unions (13 percent), retailers (13 percent), andmiscellaneous (2 percent).

TABLE 4. Present Value of Future Income Streams.a

Planning Horizon

Strategy 3 years 4 years 5 years 10 years

Job-changing withinwood products

Mean earnings $19,327 $23,560 $26,979 $36,172differential(AYc = $860/yr.)Mean earnings $18,412 $22,275 $25,346 $33,351differential, re-duced by onestandard error(AYc = $610/yr.)

Staying with current $18,859 $23,746 $28,065 $43,188employer

(AYs = $100/yr.)

aBoth strategies assume initial earnings of $8,000 (Yo of text equations 1 and 2) and are incremented yearly byAYc or AYs with a discount rate (r) of .145. Job-changers are assumed to make the initial job-change in Year 1with p = 0, which makes the numerator of text equation 1 equal to [8,000 + (1)(860)] for the mean earningsdifferential. In Year 2, the expected earnings differential is still $860, but there is now a 14 percent chance thatthey will not be able to get a job. Thus, the numerator for Year 2 would be (1 - .14) [8,000 + (2)(860)], or $8,359.This computation is probably conservative in that it assumes a 14 percent chance of zero earnings, or that thewould-be job changer is completely unemployed during that year. A more realistic assumption would involve ajob search of extended, but not infinite, length.

84

December 1980


Finally, what is the length of the relevantplanning horizon for these workers? At a 14percent yearly reduction in re-employability,only the masochist would "job-change" if hewere locked into the industry for life. In thisparticular case, however, these youngerworkers are not locked into wood products;those who leave the industry do so primarilyon the basis of their feelings about worksatisfaction (Table 5). Contrary to what had

TABLE 5. Discriminant Analysis: Two GroProducts Jobsa

been anticipated, they are evidently notconstrained by inadequate education, age,lack of skill and experience, or rurality. Theirdecisions to leave wood products have beenrelatively free from these constraints; hence,their planning horizons within the industryhave not needed to be long. In contrast toolder workers, these younger workers maywell be exploring alternative occupationswith a somewhat less risk-averse outlook.

ups of Workers Who Voluntarily Left Wood

Those taking otherwood products jobs

(n = 52)

Those taking non-wood products jobs

(n = 27)

General

AgeEducationVocational trainingb

Experience (years)

Wood productsNon-wood productsConstructionDurable good mfg.Manager or craftsmanSemi-skilled laborer

Job left

Length of jobCity sizecValue of experiencedSkilled job (D)eWeekly earnings

Job taken

Inter-county migration (D)Inter-state migration (D)"Knew people" as source

of job information (D)f

More valuable experiencethan job left (D)

Change in weekly earnings

30.0111.548.72

.09

1.02

.04

.58

.13

5.521.07.16.11.27(**)

2.507.37

(*).68

$187.37

.25(***)

(***)

( *)+ $22.20

.38

0.74

.30

.15

.48

aMean values shown separately only if variable was significantly different between groups (*** = .01, ** = .05,* = .10). Variables with (D) are dummy variables (1 = affirmative, 0 = negative).

bMonths of vocational training in high school.Cln thousands.d"How much do you think the experience you gained on this job has helped you since then - a lot (2), a little (1),or not at all (0)?"

"Non-entry level job.'Friend, relative, or previous employer.

85

N' , -- /

Stevens


Having disposed of these considerations,the present value calculations in Table 4 bearout the general rationality of frequent (inter-nal) job-changing by secondary workers. Atthe estimated AYc and AYs values, job-changing "makes sense" if the planning hori-zon were about 4 years or less. At a some-what more conservative AYc value ($610 peryear), the job-changing strategy is superiorfor planning horizons up to about 3 yearsalthough it suffers badly in any event if the

'workers looks 10 years ahead. In view of theevidence in Table 5, however, "planninghorizons" of this length are relevant only ifworkers are locked into the industry byfactors other than choice, which in this casethey have not been.

The only substantive reservation to theseconclusions is that possibility that the pro-pensity to change jobs may also jeopardize aworker's future ouside the industry. There isno evidence that this is the case; indeed,regressions of the length of unemploymentperiod (following any type of employment) onvarious stability factors revealed nothing. Itmay be that non-wood products employersexpect these workers to have unstable workhistories because of the nature of the indus-try, which is highly sensitive to housing startsand interest rates.

Conclusions

Although neoclassical critiques of the duallabor market literature have come downrather hard on that approach, the endogen-ous determination of"tastes" or "attitudes" ofworkers appears to be a contribution whichshould either be assimilated into humancapital theory or form an integral part ofwhatever evolves to replace it. The outcomesof personal investments by workers need notalways be "good"; indeed, the DLM ap-proach forces one to ask whether "bad"outcomes might be systematically expectedin some situations. In the particular casewhich was documented here, job-changersacquired negative human capital in the formof unstable work histories, but they did so ina manner consistent with the human capital

86

logic. It appears that these particular workersare being marginalized (in the terms ofPadfield and Young) only in a partial sense;they may continue to fare reasonably well aslong as they have access to jobs outside thewood products industry (which is anotherquestion in itself, due to expected declines intimber supply in this timber-dependent re-gion). The prospect is raised, however, thatless desirable outcomes may occur in othersituations, and that the present directionalbias of human capital theory should be aban-doned in favor of a more generalized view oflabor adaptations.

References

Becker, Gary S. "Investment in Human Capital: ATheoretical Analysis." J. Polit. Econ. 70(1962):9-49.

Blaug, Mark. "The Empirical Status of Human CapitalTheory: A Slightly Jaundiced Survey." J. Econ. Lit.14(1976):827-855.

Bowles, Samuel and Herbert Gintis. Schooling in Capi-talist America. New York: Basic Books, 1975.

Cain, Glen G. "The Challenge of Segmented LaborMarket Theories to Orthodox Theory: A Survey." J.Econ. Lit. 14(1976):1215-1257.

Doeringer, Peter B. and Michael J. Piore. InternalLabor Markets and Manpower Analysis. Lexington,Mass.: Heath, 1971.

Ervin, David E. An Economic Analysis of IncomeDetermination for Production Workers in Oregon'sWood Products Industry: A Human Capital Ap-proach, Unpublished Ph.D. dissertation, OregonState University, 1974.

Hall, Robert E. "Prospects for Shifting the PhillipsCurve Through Manpower Policies." Brookings Pa-pers on Economic Activity 3(1971):659-701.

Hanushek, E. A. and J. E. Jackson. Statistical Methodsfor Social Scientists. New York: Academic Press,1977.

Haveman, R. H. and J. V. Krutilla. Unemployment, IdleCapacity, and the Evaluation of Public Expenditures.Baltimore, Md.: Johns Hopkins Press, 1968.

December 1980


Padfield, Harland and John A. Young. "InstitutionalProcessing of Human Resources: A Theory of SocialMarginalization." Rural Poverty and the Policy Crisis,ed. R. O. Coppedge and C. G. Davis. Ames, Iowa:Iowa State University Press, 1977.

Rees, Albert and George P. Schultz. Workers andWages in an Urban Labor Market. Chicago: Universi-ty of Chicago Press, 1970.

Schultz, T. W. "Investment in Human Capital." Amer.Econ. Rev. 51(1961):1-17.

Stevens, Joe B. The Oregon Wood Products LaborForce: Job Rationing and Worker Adaptations in aDeclining Industry. Oregon Agr. Exp. Sta. Spec. Rpt.529, 1978.

Stevens, Joe B. and David E. Ervin. "An IncomeDetermination Model with Implications for Occupa-tional Mobility in a Rural Labor Force." ContributedPaper, Amer. Agr. Econ. Asso. Mtgs., 1977.

Stoikov, Vladimir., "The Structure of Earnings in Japa-nese Manufacturing Industries: A Human CapitalApproach." J. Polit. Econ. 81(1973):340-355.

Thurow, Lester C. Generating Inequality. New York:Basic Books, 1975.

Wachtel, Harold M. "Capitalism and Poverty in Ameri-ca: Paradox or Contradiction?" Amer. Econ. Rev.62(1972):187-194.

Wachter, Michael L. "Primary and Secondary LaborMarkets: A Critique of the Dual Approach." Brook-ings Papers on Economic Activity 3(1974):637-680.

Young, John A. and Joe B. Stevens. "Job Rationing,Human Capital, and Normative Behavior." HumanOrg. 37(1978):29-37.

Zell, Steven P. "Recent Developments in the Theory ofUnemployment." Federal Reserve Bank of KansasCity: Monthly Review (September/October 1975):3-10.

87

Stevens


88

December 1980

secondary workers and the acquisition of negative...

Documents