 

 

 

 

. .X‘S'J-ZY.‘L‘-

n
2&3:
w

 

5

"RM:-
i
‘b’i‘
4.1 .

£71

'&

100‘)

LIBRARY
Michigan State

 

' ""’m

University

 

 

 

This is to certify that the
dissertation entitled

The Role of Occupation-Speciﬁc
Human Capital in Economic Analysis

presented by

Russell Alan Ormiston, Jr.

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in School of Labor and Industrial
Relations

 

 

AA 2 A%

Major Professor’s Signature

August 12, 2007

 

Date

MSU is an afﬁrmative-action, equal-opportunity employer

 

 

u—.-.-.-.--.--.—-.-.-.--n-.-n-- .--43-.---o-—o------.-4-.-.-n-o-n-agnqgg-.-.----.—.--uu..-

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/CIRC/DaleDue.indd-p.1

THE ROLE OF OCCUPATION-SPECIFIC
HUMAN CAPITAL IN ECONOMIC ANALYSIS

By

Russell Alan Ormiston, Jr.

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
School of Labor and Industrial Relations

2007

ABSTRACT

THE ROLE OF OCCUPATION-SPECIFIC
HUMAN CAPITAL IN ECONOMIC ANALYSIS

By

Russell Alan Ormiston, Jr.

This three-chapter compilation examines the theoretical and empirical implications of
occupation-speciﬁc human capital as it relates to current labor economics research. The
ﬁrst chapter demonstrates that acknowledging occupational speciﬁcity in the human
capital model allows for a reconciliation of a long-standing theoretical dispute regarding
the role of occupation in the labor market. The second chapter extends the literature by
estimating the cross-occupation transferability of human capital using data on the
knowledge, skills, and abilities utilized in each vocation. These estimates are then applied
to verify displaced blue-collar manufacturing workers as structural “victims” given lower
rates of human capital application in their new occupations compared to others displaced
in the labor market. The third chapter investigates the relationship between high school
employment and post-school economic outcomes, as it uses occupation-speciﬁc human
capital principles to dismiss the notion that in-school employment provides the

“marketable skills” necessary to stimulate post-school economic gains.

ACKNOWLEDGMENTS
Considerable appreciation is extended to the members of my dissertation committee—Dr.
Karen Roberts, Dr. Stephen Woodbury, and Dr. Richard Block—for their extensive time
and input in the formation of this work. Special thanks are particularly extended to Dr.
Dale Belman, the chairperson of my dissertation committee, who was instrumental in
. both this work and in my development as a scholar. Finally, I am grateful to my
colleagues in the School of Labor and Industrial Relations, especially Kyoko Kato, for

their support during my time at Michigan State University.

On a personal level, any acknowledgments would fail to be complete without recognizing
the love and encouragement provided by my parents, Russell Sr. and Barbara. It was their
support—and their plight—that set me on this path, and I am eternally appreciative.
Further, my siblings, Brian and Kate, have been with me every step of the way, and I am
proud of our relationship and hope it continues for the rest of our lives. Finally, my wife,
Mary, has been an incredible inﬂuence in my life, and I hope that this completed
dissertation represents the beginning, not the end, of a Wonderful, quixotic journey

together.

iii

TABLE OF CONTENTS

List of Tables ..................................................................................................................... vi
List of Figures .................................................................................................................. viii
CHAPTER 1
OCCUPATION AND THE HUMAN CAPITAL MODEL ................................................ 1
Introduction .............................................................................................................. 1
Literature Review ..................................................................................................... 4
Human Capital Model View of Occupation ................................................ 4
Institutional Approach to Occupation .......................................................... 8
Review of Occupation-Speciﬁc Human Capital ........................................ 11
Applications of Occupation-Speciﬁc Human Capital ................................ 20
Applications to the Wage Equation ........................................................... 27
Data and Model ...................................................................................................... 31
Results .................................................................................................................... 36
Discussion .............................................................................................................. 44
CHAPTER 2
DISPLACED MANUFACTURING WORKERS AND OCCUPATION-SPECIFIC
HUMAN CAPITAL ........................................................................................................... 46
Introduction ............................................................................................................ 46
Literature Review ................................................................................................... 48
Developing New Indices ........................................................................................ 52
Data and Model ...................................................................................................... 64
Results .................................................................................................................... 70
Discussion .............................................................................................................. 82
CHAPTER 3
HIGH SCHOOL EMPLOYMENT AND THE ACCUMULATION OF OCCUPATION-
SPECIFIC HUMAN CAPITAL ........................................................................................ 84
Introduction ............................................................................................................ 84
Literature Review ................................................................................................... 86
Data and Model ...................................................................................................... 89
Results .................................................................................................................... 97
Discussion ............................................................................................................ 103
APPENDICES ................................................................................................................. 106
Appendix A: Examining Occupation as a Mediatdr Between Education and
Wages ....................................................................................................... 106

iv

Appendix B: Nested Logit Results, Displaced Blue-Collar Manufacturing

Workers, Females, 2000-2004 DWS ....................................................... 113

Appendix C: Regression Analysis on ln(Hourly Wage), With Firm Controls,
Setting OSHC=0 for All Non-Employed High School Seniors ............... 115
REFERENCES ................................................................................................................ 117

LIST OF TABLES

TABLE 1. Sample Means .................................................................................................. 34
TABLE 2. Occupational Speciﬁcations ............................................................................. 35
TABLE 3. Wage Determination, Males, Full Sample, Series of Education Dummies ..... 37
TABLE 4. Wage Determination, Males, Full Sample, With Occupation*Education

Interactions, Series of Education Dummies ....................................................................... 43
TABLE 5. Knowledge, Skill, and Ability Categories Used by O*NET ............................ 54
TABLE 6. Comparing Sim Scores to Establish Applicability of Required KSAs .............. 5 8
TABLE 7. Demonstrating the Development of Transferability Scores ............................. 60

TABLE 8. 5x5 Fragment of T, the KSA Transferability Matrix (Values Represent tjj)....61

TABLE 9. 5x5 Fragment of Occupational Matrix, Presenting Qij and Tij ......................... 63
TABLE 10. Comparison of Blue-Collar Manufacturing Displaced Against Other
Displaced, 2004 Displaced Worker Survey ....................................................................... 68
TABLE 11. Re-Employment Occupation of Displaced Blue-Collar Manufacturing
Workers, 1984-2004 DWS ................................................................................................. 71
TABLE 12. Nested Logit Results, Displaced Blue-Collar Manufacturing Workers, Males,
2000-2004 DWS ................................................................................................................ 73
TABLE 13. Transferability of KSAs, Displaced Blue-Collar Manufacturing Versus All
Other Displaced, 2004 DWS .............................................................................................. 76
TABLE 14. Regression Analysis on Transferability of KSAs from Displaced to Re-
Employment Occupation, 2004 DWS ................................................................................ 79
TABLE 15. Regression Analysis on ln(Week1y Earnings), Among Those Re-Employed
and Working at Least 20 Hours/Week, 2004 DWS ........................................................... 81
TABLE 16. Summary Statistics ......................................................................................... 96

vi

TABLE 17. Mean Hourly Wage by Occupational and Employer Status .......................... 97
TABLE 18. Regression Analysis on 1n(Hourly Wage), Without Firm Controls ............... 98
TABLE 19. Regression Analysis on 1n(Hourly Wage), With Firm Controls .................. 101

APPENDIX TABLE Al. Top-Scoring Occupations and Knowledge Categories
According to O*NET Rankings ....................................................................................... 110

APPENDIX TABLE B1. Nested Logit Results, Displaced Blue-Collar Manufacturing
Workers, Females, 2000-2004 DWS ............................................................................... 113

APPENDIX TABLE C1. Regression Analysis on 1n(Hourly Wage), With Firm Controls,
Setting OSHC=O for All Non-Employed High School Seniors ....................................... 115

vii

LIST OF FIGURES

FIGURE 1. Partial Dimensions of Human Capital in Bachelor’s Degrees in Engineering
and Liberal Arts ................................................................................................................. 18

FIGURE 2. Diagram of Hierarchical Decision Tree Applied by Nested Logit Model of
Occupational Choice .......................................................................................................... 69

viii

Chapter 1

OCCUPATION AND THE HUMAN CAPITAL MODEL

INTRODUCTION

With standard applications ranging from discrimination to education, the wage equation
has evolved into the most fundamental analytical tool utilized by labor economists. With
origins tracing back to Mincer (1974), its prevalence in the ﬁeld is evidenced by the fact
that no less than forty articles featured a wage equation in the Journal of Labor
Economics between 2000 and 2004. However, despite the pervasiveness of its usage—and
perhaps because of it—cross-sectional approaches are typically nonchalant in specifying
occupation in the model, even absent data restrictions. For example, efforts to “include
occupation” in a regression can run the gamut, ranging from a simple white-collar vs.
blue-collar demarcation (Arai 2003) or a series of ﬁve occupational controls (Wheeler
2001) to expansive sets of controls, whether it be thirteen l3 vocational variables
(Bratsberg, Ragan, and Nasir 2002) or other more sophisticated techniques (Hirsch and
Macpherson 2004). This inconsistency has failed to be rigorously examined—or even
denoted—in labor economics literature despite the fact that it plays a signiﬁcant role in

shaping research ﬁndings offered through the wage equation.

While this paper will investigate this issue, the inconsistency of occupational

speciﬁcation in the wage equation mirrors a larger, unresolved ambiguity in vocation’s

role in the human capital model, the underlying theory from which the wage equation
was derived. Thus, while the human capital model envisioned by Mincer (1958) and
Becker (1962) sparked the “human investment revolution in economic thought”
(Bowman 1974) that serves as the foundation for much of today’s labor market analysis,
the failure of the model to predict certain occupational phenomena has fueled an
undercurrent of dissent over the validity of the theory, which, in turn, has manifested
itself in alternative research streams such as segmented labor markets (SLM) and sorting
models. This dissension has been typically steeped in institutional thought, with
supporting literature questioning the human capital model’s theory of wage determination
(Fogel 1979), explanation of educational wage premiums (Spence 1973), its neoclassical
economic assumptions (Gottschalk 1978), and to some extent, its deﬁnition of the labor
market itself (Osterrnan 1975; Kerr 1977). While debate over these issues reached its
climax in the 19703, the role of occupation in the model has never been resolved, as
reﬂected in the adversarial relationship between human capital theorists and institutional

scholars that exists to the present day (Budd 2005; Vedder and Gallaway 2005).

While scholars have typically recognized the concurrent inﬂuences of human capital and
institutional thought at work in the labor market (Becker 1993), attempts to bridge the
theoretical gap between the two schools have been minimal. However, a small, but
recently revitalized literature on occupation-speciﬁc human capital may represent a
simple elaboration of human capital principles upon which to establish a reconciliation of
the occupation-based failings of the human capital model. Originally advanced by Shaw

(1984), it has been only recently that scholars have begun to examine the implications

associated with occupational speciﬁcity within the human capital framework (Milgrom
2004; Kambourov and Manovskii 2004a, 2004b, 2005; Sullivan 2005a, 2005b; Groen
2006). Given these conﬁrmations of occupation-speciﬁc human capital’s role in
determining labor market outcomes, the opportunity now arises to revisit these
occupation-based critiques of the human capital model to examine how this elaboration
of the theory may harmonize the seemingly disparate arguments of institutionalists and

human capital theorists.

To examine the questions surrounding occupation’s role within the wage equation and,
more broadly, the human capital model, this paper will begin by laying out the human
capital theory and its suggestions for occupation’s role. Next, this study will synthesize
occupation-based critiques of the human capital model into a cohesive reference, an
absence in the literature that may have been partly responsible for the inherent ambiguity
surrounding vocation’s place in the theory. Moving on, the next section of the paper will
examine the evidence on occupation-speciﬁc human capital before laying out the bridges
this reﬁnement may provide between institutional and human capital thought. Finally,
this paper will apply these lessons to examine the initial topic advanced by this paper—the

speciﬁcation of occupation in the wage equation.

L] T ERA TURE RE VIEW

Human Capital Model View of Occupation

In response to wage analyses of individual occupations, the original goal of the human
capital model was to serve as a means through which researchers and practitioners could
explore the economic impact of investing in people from a broader, more general
perspective (Becker 1993, 29-30). While the founders of the model may have been
apprehensive about dubbing their theory “human capital” (Becker 1993, 16), the theory
recognizes that people, like factories and businesses, are entities that can be invested in to
increase their productive capacity. Through such “investments” as education, training,
and work experience, it is argued that an individual’s knowledge, skills, and abilities

expand to generate greater productivity at the workplace.

These investments are hypothesized in the early literature as taking one of two forms—
ﬁrm-speciﬁc human capital, or expertise useful to a single employer, and general human
capital, which is useﬁil to all employers. It was theorized that ﬁrms had the economic
incentive to fund the former, as workers would be more productive at that speciﬁc
company and could not parlay this investment to advantage elsewhere in the labor
market. Further, it was argued that the onus would be on individual employees to ﬁnance
general training, as ﬁrms would not want to run the risk of losing their investment upon

employee turnover.

Regardless of the type of investment made, the augmented productivity wrought through
human capital investments is posited to be experienced through increased wages to the
individual. This is based on the neoclassical economics approach that suggests that ﬁrms
will hire workers such that the wage will equal each individual’s marginal productivity,
with some adjustments made for the cost of training within the ﬁrm (Becker 1993, 31-
33), among other factors. Strengthening this argument, the human capital model contends
that a person’s investment choice is based on the predicted long-term earnings stream

associated with each available investment option (Becker 1993, 60).l

Attempts to econometrically explore the relationship between investments, such as
education, and earnings were signiﬁcantly boosted by the work of Mincer (1974), who
operationalized cross-sectional analyses through his advancement of the human capital
earnings function. The base of Mincer’s model, which has subsequently served as a

foundational model in labor economics, is as follows:

. . . 2
1n (wager) = [30 + 01*educatroni + [32*exper1encei + [33*exper1encei + 8i_

While recognizing the possible need to include marketplace controls, Mincer justiﬁed this
simple econometric approach on the basis of ﬁrndamental hypotheses conveyed in the

human capital model:

 

A subtle inconsistency in the literature illustrates that when talking about the actions of workers, some

theorists suggest that individuals act to maximize earnings (e.g., Shaw, 1984, 322), while others discuss
maximizing utility (e.g., Kumar and Coates 1982, 443). While the former is often used to proxy the latter
given the near impossibility of quantifying “utility,” theoretical completeness requires one to recognize the
inherent differences between the two.

The model of worker self-investment as the basic determinant of earnings might be criticized as
giving undue weight to the supply of human capital while ignoring the demand side of the market.
Certainly, demand conditions in general, and employer investments in human capital of workers in
particular, affect wage rates and time spent in employment, and thereby affect earnings. It should
be clear, however, that the earnings ﬁinction in this study is a “reduced form” equation, in which
both demand conditions and supply responses determine the levels of investment in human capital,
rates of return, and time worked. The present approach is an initial and simple one, and greater
methodological SOphistication is clearly desirable. There is a need to relate employers’ behavior
both as demanders of and direct investors in human capital to the observed distribution of
eamrngs.

The investment-eamings relation in this study is in reduced form also in the sense of describing
equilibrium loci in the (human) capital market as well as in the labor market in which human
capital is supplied as a factor of production. As Becker describes in his analysis, the cross-
sectional earnings function results from two simultaneous structural relations in the (human)

capital market. These are demand functions (Di), which relate individual investments to marginal

rates of return, and supply functions (Si), which relate to the volume of funds that can be obtained
for human capital investment to the marginal “interest” costs. Of course, worker demand for self-

investment (Di) is, in part, derived from employer demand for the workers’ human capital. (p.

137)
As stated previously, Mincer’s “reduced form” approach—the wage equation-is inherently
built on the neoclassical economics assumptions required for equilibrium within the
human capital and labor markets. It is important to recognize that these assumptions
maintain that disequilibria along racial, gender, or occupational lines should not persist,
as worker and investment ﬂows should be altered to achieve arbitrage, or the equalization
of payments across particular groups. Any such differences in payoffs to occupations are
thus attributed to compensating wage differentials (Smith 1976; Kumar and Coates
1982), or non-pecuniary aspects of the job, maintaining a form of “utility arbitrage”

across vocations.

While the origins of human capital thought were generally silent on vocational concerns,2
the model’s general stance on occupational choice is best illustrated by an example used

by DeBeyer and Knight (1989):

Once education is determined... workers have a preference-ordering among occupations. There
will be one occupation in which a worker with a certain education is more productive, and thus
receives a higher wage, than any other. On the one hand, the educated will wish to avoid the
unskilled jobs in which their education is less productive and is therefore less highly rewarded. On
the other hand, the uneducated will wish to avoid the skill-intensive jobs in which their
productivity and pay are lower than in unskilled jobs. (p. 596-97)

While based on economic assumptions, it should be noted that the wage equation and the
human capital model’s approach to occupational choice-taking the job that maximizes
earnings following completion of self-investment—collectively fed into a broader
ideology that devalued the role of occupation in terms of explaining wage variation.
Duncan (1961) typiﬁed this stance by suggesting that occupation was merely an
“intervening” variable through which educational differences operated to create earnings
differentials. Thus, given the “ﬁrst education, then occupation” choice mechanism and
supposed arbitrage in the labor market, the human capital model advanced a philosophy
espousing human capital as the driving force in wage determination given the

mechanisms of the market.

 

6‘

As an illustration of the human capital model’s silence” on occupational concerns, the central book on
the human capital model-Becker (1993)—fai|s to include the word “occupation” even once. In Mincer’s
(1974) fundamental book in deriving the human capital earnings function (i.e., the wage equation), the
word “occupation” appears only one time.

Institutional Approach to Occupation

Institutional thinkers resisted the doctrine of human capital given its assumptions that the
labor market operated in a fashion similar to capital markets,3 as well as its “ad hoc”
approach to explaining various phenomena (Blaug, 1976). In particular, a speciﬁc
undercurrent of dissent arose in regard to the theory’s silence on occupational issues and
the failure of the theory to predict certain vocational outcomes. Ideologically, Mayhew
(1971) explicitly rejected Duncan’s assertion that occupation serves as an “intervening”
variable between education and wages, urging wage analyses to include occupation for
“the correct analysis of the economic importance of education” (p. 225). Fogel (1979)
took this one step farther, labeling the human capital approach to wage determination
“misleading” (p. 26). Fogel thus challenged the notion that wage rates were the result of
the interaction of a worker’s marginal productivity with the short- and long-run
framework of market supply and demand, instead suggesting that one’s occupation—not
one’s personal characteristics—was the key component of the wage-setting process,
calling occupational choice “the most ﬁmdamental” decision an individual makes in the

determination of wages:

Buyers and sellers of labor services organize their exchange communication around a set of work
duties, that is, an occupation (or job, which is simply a speciﬁcally deﬁned occupation); thus the
forces of wage determination must necessarily be directed to occupations, although wage changes
may frequently be applied to entire administrative or bargaining units. Furthermore, as previously
stated, individual wage contracting takes place with reference to a speciﬁc occupation, and the
associated occupation wages-however determined—constrain the contract zone within which
agreement is reached. (p. 26).

 

For discussion of the dissimilarities between the two markets, see Prasch (2004).

Beyond the ideological issues surrounding the true nature of earnings differentials, a
number of alternative research streams identiﬁed the existence of labor market
imperfections that were not easily explained using the human capital framework. Thus,
while ﬁndings of persistent occupational wage disequilibria (Reder 1955; Eckaus 1973)
called into question neoclassical economic assumptions of arbitrage, segmented labor
market theorists (Doeringer and Piore 1971) illustrated that different portions of the labor
market operated in vastly different manners.4 In particular, it was noted that the return to
education and other forms of human capital was markedly varied across primary and

secondary occupational groups (Osterman 1975).5

Kerr (1977) justiﬁed the presence of these segmentation effects by recognizing that the
idea of a “labor market” as examined by the human capital model was a misnomer.
Instead, Kerr surmised that the labor market was balkanized along occupational (and
geographic) lines due to institutional barriers, such as schooling requirements. This
premise gave rise to the notion of a fragmented labor market consisting of an inﬁnite set
of non—competing groups, even if those groups are never sharply deﬁned. Arguing that

“Painters do not compete with bricklayers, or typists with accountants, or doctors with

 

4
As summarized by Hagner (2000), primary-sector workers are theorized to enjoy high wages, fringe

beneﬁts, satisfactory working conditions, reasonable levels of employment security, possibilities of
promotion, and some degree of worker autonomy. Secondary-sector workers, meanwhile, are suggested to
experience low wages, minimal or nonexistent beneﬁts, less-than-desirable working conditions, greater
probability of layoff and displacement, and harsh work rules or supervisory styles.

Cain (1976) and Kruse (1977) criticized Osterman’s methodology, suggesting that within-segment

regressions suffered from truncation bias of the dependent variable. These criticisms argued that by
grouping individuals into “primary” and “secondary” jobs, Osterman was classifying people on the basis of
their wage-earning ability. Osterman (1977) responded that this technique was appropriate, for it was the
job, and not the individual, that served as the selection variable.

lawyers,” (p. 23), Kerr recognized the inﬂuence of occupation-speciﬁc market forces and

wage structures on the determination of wages.

As theorized by Kerr (1977) and Fogel (1979), a consequence of such occupational
barriers was that they served to insulate particular labor markets, thereby restricting
worker ﬂows and, making them noncompetitive. This allows the persistence of
nonequilibrium outcomes, which contradicts the arbitrage assumptions of the human
capital model. Fogel, in particular, noted that occupational wage disequilibria were able
to exist because some workers could not forgo present costs in order to make the
necessary investments to enter particular vocations, leading to noncompetitive
occupational labor markets. Thus, different occupational environments are able to exist
such that returns to human capital vary not only by labor market segment, but also by
individual occupation (Rumberger 1987; Freeman and Hirsch 2001; Quinn and Rubb

2005)

Corresponding to this, a primary assertion of the human capital model rationalizes that
the positive relationship between education and earnings is attributable to the enhanced
productivity gained through schooling. That said, research on sorting models, or the more
general name for “signaling” and “screening” effects, has illustrated that educational
premiums extend beyond payment for increased productivity (Spence 1973; Layard and
Psacharopoulos 1974; Riley 1979; Weiss 1995; Lofstrom 2001; Frazis 2002). This theory
suggests that employers, with incomplete information regarding an applicant, will use

education as a signal of productivity or other personal attributes in making their hiring

10

decisions. As such, institutional theorists have hypothesized that the positive relationship
between schooling and wage is not due to enhanced productivity, but rather to incomplete

information in the hiring process.6

Review of Occupation-Speciﬁc Human Capital

To review, two schools of thought have emerged to explain wage determination in the
labor market. Proponents of the human capital model suggest that education, experience,
and training are the primary drivers of earnings, as individuals select occupations that
deliver the greatest payoff to their investments. Human capital theorists cite neoclassical
economic assumptions to suggest that augmented productivity wrought through such
investment decisions will yield increased wages as established by the supply and demand
conditions of the human capital and labor market. Meanwhile, institutionalists contend
that occupation is the primary determinant of wages, with human capital investments
merely representing a means of entry given the presence of vocational barriers, such as
schooling requirements. Proponents of this school of thought justify their position by
pointing to the shortcomings of the human capital model in explaining such phenomena

as occupational wage disequilibria, segmented labor markets, and sorting models.

While forty years of debate between the two schools may have polarized the issues

(Leonard 2000), even the most ardent advocate recognizes the validity of each position as

 

It should be noted here that efforts to obtain an unbiased return on schooling have been relentless, with

the primary critique concerning the inability to measure—and thus to include—ability in the wage equation
(Griliches 1977). For further discussion, see Card (1998), among others.

11

it applies to labor market operations (Becker 1993). Thus, it is important to recognize that
a simple reﬁnement of the term “human capital” as it applies to the labor market may
allow the human capital model to incorporate some of the occupation-based
considerations brought forth by institutionalist thought. In particular, in reexamining the
original composition of “human capital”—both general and ﬁrm-speciﬁc—as advanced by
Becker (1993),7 a small, but recently revitalized literature has conﬁrmed the presence of a
subtle variation therein, occupation-speciﬁc human capital, that may be the appropriate
lens through which the human capital model can consider the occupational phenomena

that lie at the heart of the institutional critiques discussed earlier.

While vocational speciﬁcation of skill has long been established in economics (Smith
1976; Weiss 1971), Shaw (1984) was the ﬁrst to examine this occupational speciﬁcity as
it related to the assumed composition of human capital. Deﬁning occupational investment
as “the accumulation of skills an individual acquires to perform work within an

’9,

‘occupation (p. 320),8 Shaw presented a nuanced view of human capital, suggesting
that general human capital maintained a distinct occupation-speciﬁc component that

could be carried across employers.9 However, Shaw noted that, unlike the case of general

human capital, individuals would be able to transfer only a certain amount of their

 

A number of studies have challenged the perfect competition assumptions of the human capital model

which suggest that ﬁrms will not fund general training, as it is theorized that labor market imperfections
allow employers to earn rents on “generally-trained” workers (e.g., Acemoglu and Pischke 1998, 1999);

Shaw (1984) deﬁnes an “occupation” as “a homogeneous skill classiﬁcation within which individuals

are perfect substitutes in demand and/or have inﬁnite cross elasticities of substitution in supply. Cabinet-
makers and engineers, for example, have occupation-speciﬁc skills,” (pg. 320).

Given the portability of occupation-speciﬁc human capital as a component of general human capital, the

tenets of the model suggests that individuals, and not ﬁrms, would bear the brunt of funding occupation-
speciﬁc human capital. Exceptions do exist, however, as illustrated by construction apprenticeship
programs, as ﬁrms and unions collectively fund training programs despite worker mobility.

12

occupation-speciﬁc human capital stock in the event of a vocational switch, with the
amount inherently depending on the “transferability” between the two respective
occupations. In other words, the skills acquired as an art teacher may not be transferable

to a career in medicine; in contrast, the training received by an actuary may have

considerable transferability to a job as a statistician.10

Empirically, Shaw demonstrated the prominent role that occupation-speciﬁc human
capital played in the determination of wages, as it served as a predictor of wage “far
superior” to that of the general experience variable as advanced by Mincer (1974).
However, despite this ﬁnding, the implications of Shaw’s work were largely ignored until
very recently, as additional research has conﬁrmed occupation-speciﬁc human capital’s
role as an important predictor of wages. Kambourov and Manovskii (2005), for instance,
suggest that an additional ﬁve years of occupational tenure is expected to increase wages
by 12 to 20 percent. Further, the authors noted that while displaced males between the
ages of twenty and sixty suffered, on average, a 15 percent reduction in weekly earnings,
those who stayed in the same occupation experienced only a 6 percent drop compared to

the 18 percent loss for those who switched vocations.

Additional studies have conﬁrmed the inﬂuence of occupation-speciﬁc human capital on
labor market outcomes. Kambourov and Manovskii (2004a), for instance, suggest that the
occupational speciﬁcity of human capital, along with increased variability of productivity

shocks to those skills in the labor market, explains the increase in wage dispersion and a

 

Dolton and Kidd (1998) note that the degree of transferability is inherently an empirical question
dependent upon how broadly one treats occupational titles.

13

decline in wage stability over the last thirty years. Dolton and Kidd (1998) and Gibbs,
Ierulli, and Milgrom (2002) illustrate the role occupation-speciﬁc human capital has in
explaining vocational mobility,ll while Sullivan (2005b) ﬁnds substantial wage inﬂuence
of occupation-speciﬁc human capital through a new approach-a dynamic structural model
of career choice that nests job search within a human capital model of occupational and
educational choices. Finally, Groen (2006) ﬁnds a signiﬁcant relationship between the
size of one’s local economic environment and the outcomes associated with occupation-

speciﬁc human capital.

Therefore, given the growing acceptance of both occupation-speciﬁc and industry-
speciﬁc (Neal 1995; Parent 2000) human capital, it is necessary to revisit the assumed
composition of human capital in the basic model. As mentioned earlier, the original
formulation of the human capital theory was built upon on the concept of two-
dimensional human capital—general and ﬁrm-speciﬁc-an assumption that has increasingly
been called into question (see, among others, Carrington 1993; Thomas and Ong 2002).
However a reexamination of productivity in the labor market is suggestive of ﬁve distinct
categories of human capital, segmented on the basis of worker-ﬁrm attachment critical to

basic human capital theory:

 

Gibbs, Ierulli, and Milgrom (2002) suggest that if a typical worker ﬁrst gains attachment to an

occupation and stays with it while searching across ﬁrms, then many of the characteristics of internal labor
markets may apply as well to “occupational labor markets (OLMs).” Milgrom (2004) examined internal
labor markets and found that, especially at higher ranks, there exists more evidence of occupation-based
ports of entry than ﬁrm-based ports of entry. In other words, a ﬁrm is more likely to hire someone in the
same occupation from a different ﬁrm than promote from within.

14

MARKET-WIDE EMPLOYER-SPECIFIC

(1) General (4) Firm-Speciﬁc

(2) Occupation-Speciﬁc (5) Job-Speciﬁc

(3) Industry-Speciﬁc
The “market-wide” category contains the components of human capital that are portable
across employers, thereby putting the theoretical burden on employees to acquire these
skills on their own. Meanwhile, the “employer-speciﬁc” classiﬁcation includes the two
components that are unique to the Worker-ﬁrm relationship and would be lost upon its
termination. While the research has yet to identify “job-speciﬁc” components, this
inherently allows for the inclusion of the productivity increases attributable to knowing
the best methods, or “shortcuts,” available at speciﬁc tasks in one’s job that would be lost
if the individual was reassigned elsewhere in the ﬁrm but with a similar role. For
example, an insurance claims agent responsible for Michigan may have signiﬁcant
familiarity with all of the necessary forms and procedures after working in that role for
ﬁve years. However, if that person were moved across the hall and put in charge of
Georgia claims, there would be a loss of job-speciﬁc capital that affects that person’s
productivity given his or her initial unfamiliarity with the corresponding procedures for
the new state. Therefore, failure to identify job-speciﬁc components excludes an

important dimension of one’s productivity on .the job and the corresponding utility

inherent in the position.
Drawing clear distinctions between the ﬁve classiﬁcations of human capital represents a

subtle theoretical variation from Shaw (1984) and Thomas and Ong (2002), among

others, who inherently advance an intertwined notion of occupation-speciﬁc and purely

15

general human capital.12 In contrast, this clear partitioning allows for greater theoretical
ease in explaining certain labor market phenomena using human capital principles (see
following section), and distinctions between each type are made in light of well—deﬁned
occupational and industrial barriers to entry that are responsible for the balkanization of
the labor market (Kerr 1977).13 In other words, there are clear occupational effects at
work in the labor market operating through these barriers to entry, such as requirements
for a master’s in business administration or a Ph.D. in economics. In addition, Becker’s
(1962) original deﬁnition of “general” human capital suggests that it can be transferred
across all employers; in contrast, occupation-speciﬁc (and industry-speciﬁc) human
capital is not applicable to all employers, leaving notions of the intertwined nature of
occupation-speciﬁcity contradictory to the basic theory. Further, research has illustrated
markedly contrasting effects of each of the forms of human capital on wage estimates,
with occupation-speciﬁc human capital generally outpacing other forms, including
general, in terms of importance to establishing wages (Shaw 1984; Milgrom 2004;

Kambourov and Manovskii 2005; Sullivan 2005a, 2005b).”

 

12

Speaking on the application of occupational investment across vocations, Shaw (1984) suggested that,

“Transferability is always greater than zero because a portion of postschool investment is completely
general (for example, learning to communicate with others)” (p. 321).

A controversial theoretical alternative would be to suggest that, in light of occupation-speciﬁc and

industry-speciﬁc human capital, the idea of “general” human capital contains an empty set. This is because
even commonly referenced examples of “general” human capital, such as general literacy or the ability to
interact with people, inherently are applied only within the context of a given job. Literacy may not be
required for agricultural workers. Being able to interact with people may not affect the productivity of
computer programmers. While this point is not made in this paper, it could be alternatively argued that
“human capital” is deﬁned only by how it is applied to the characteristics and duties of a particular job.

Shaw’s (1984) illustration of occupation-speciﬁc human capital’s dominance over a measure of general

human capital has been found to be consistent with later studies demonstrating the prominent role that
occupation-speciﬁc human capital plays in the determination of wages. However, the impact of other forms
of human capital has been a source of debate. In particular, while Kambourov and Manovskii (2004a, 2005)
suggest industry-speciﬁc human capital had little effect, Sullivan (2005a) found the opposite, indicating
that the former’s failure to predict signiﬁcant industry-speciﬁc effects was due to the failure to include

16

An important extension of the reclassiﬁcation of human capital recognizes that all human
capital investments—and thus, an individual’s human capital stock-can be partitioned into
each of the ﬁve categories above. While vague boundaries between classiﬁcations and
the impossibility of quantifying human capital investments makes this a theoretical
exercise,'5 it serves as a means of understanding the impact of various investments on
labor market outcomes such as wage and mobility. For instance, consider the following
hypothetical comparison between a bachelor’s degree in engineering and a bachelor’s

degree in liberal arts:

Figure 1 illustrates that each investment—in this case, a bachelor’s degree—can be
examined as to the amount of human capital imparted to each occupation, industry, ﬁrm,
and general category. While Figure 1 provides only a partial list of human capital
dimensions, it illustrates that an engineering degree may instill an extraordinary amount
of knowledge and skills within engineering occupations, as well as some exposure to
speciﬁc ﬁrms and industries. However, an engineering curriculum may be of much less
use in introducing the necessary expertise to teach a high school class or work as a
newspaper reporter. Meanwhile, a broader liberal arts degree may provide more pure

general training, as well as equip one with some proﬁciency across a wide range of

 

within-ﬁrm occupational switches. While the effect of industry-speciﬁc human capital is an issue of debate,
the role of ﬁrm-speciﬁc expertise in wage determination has increasingly been seen as marginal (Lazear
2003; Kambourov and Manovskii 2004a, 2005; Milgrom 2004; Poletaev and Robinson 2004).

15

As noted in Kambourov and Manovskii (2005), the deﬁnition of “occupation” plays a signiﬁcant role
in determining the accumulation—and applicability-of human capital in that vocation. For instance, if the
occupational spectrum were divided to include a category for “professors,” this would inherently suggest
that physics and economics professors are perfect substitutes. Thus, greater reﬁnement would be needed to
accurately model labor market phenomena.

17

FIGURE 1. Partial Dimensions of Human Capital in Bachelor’s Degrees in
Engineering and Liberal Arts

 

 

General

 

 

134‘ . R
Omupation: Electrical ‘ ‘ , _
Engineer '0,’ ,
Occupation: High School ,.
Teacher .iiﬂ.
Occupation: Newspaper .. .
Reporter ‘ *r '* F"

 

Industry: Automotive [i 6 ”Fifth“.

 

 

 

 

 

 

 

 

 

 

Firm: General Motors 4 ‘

 

 

 

EIBS in Engineering .88 in Liberal Arts

 

 

 

 

occupations and industries, but might not provide as great a depth of training in any one

ﬁeld.

The recognition of the multidimensional nature of human capital within this ﬁve-category
system represents a simple elaboration on the transferability of occupation-speciﬁc
human capital as denoted by Shaw (1984). She contended that time spent within an

occupation, i, could be applied to a separate occupation, j, by the degree of transferability

between the two occupations, or yij (p. 321).16 This suggests, for example, that a year

 

In an effort to establish a matrix of transferability between occupations, Shaw (1984) examined
occupational mobility-movement across occupations-using the Current Population Survey. However, this

18

spent as an accountant not only expands one’s occupation-speciﬁc human capital stock as

an accountant, but also expands one’s speciﬁc stock as an actuary, engineer, and physicist

by the degree of transferability between the respective occupations given a nonzero Yij.

In light of the ﬁve partitions of human capital advanced earlier, the implications of the
multidimensionality of investments suggest that each person inherently maintains a stock

in each of V vocations, I industries, F ﬁrms, and J jobs, as well as a general category—or a

(IF VxIxeJxl human capital matrix. Each component of that matrix, (bvifjl, is

determined by the sum of education, experience, training, and other forms of learning
applicable to that cell of the matrix, depreciated by some measure of obsolescence.l7

Thus, while this may be quantitatively intractable, it is important to note that individuals

use only a single component of that matrix, (bvifj 1, in any one job. 18

It is important to note a couple of intricacies in constructing measures of the components

of the human capital matrix, ¢viijl- First, while researchers will typically be limited to

duration of schooling or within a particular occupation as a means of estimating speciﬁc

 

approach introduces a signiﬁcant amount of noise given the occupational coding errors of the CPS that bias
vocational switching studies (Kambourov and Manovskii, 2004b).

The rate of obsolescence as applied to this theoretical matrix would allow for heterogeneity given

varying degrees of technological change, among other factors, across occupations, industries, ﬁrms, and
jobs. For instance, the rate of obsolescence of one’s skill in the occupation of “high school algebra teacher”
may be quite different than the rate of obsolescence of one’s expertise as a “computer programmer”
(lgishop, 1998).

As a matter of precision, it is therefore important to recognize that other (l’viﬁl in the matrix are not

lost or “destroyed” (Kambourov and Manovskii 2005) if unapplied. They remain inherent to the person,
just not used.

19

human capital, one must recognize the inherent differences in intensity that may exist
across investment options (Shaw 1984; Sullivan 2005a). In other words, one might expect
a higher intensity level of investment for one’s freshman year at an Ivy League school
compared to a local community college. Second, this intensity will not be linear over
time, as there may be diminishing marginal returns to additional years spent in a
particular position. In other words, once someone has mastered the tasks speciﬁc to a
particular job, there may not anything left to learn, making additional time spent in that

role insigniﬁcant to the accumulation of additional human capital.

Applications of Occupation-Speciﬁc Human Capital

While the full labor market implications of the redeﬁnition of human capital extend
beyond the bounds of this paper, the introduction of occupational speciﬁcity into the
human capital model represents a subtle reﬁnement of the theory capable of explaining
certain occupational phenomena noted by institutionalist critiques. Of most importance
has to do with the underlying source of discord between institutionalists and human
capital theorists regarding wage determination in the labor market. As discussed earlier,
human capital theorists contend that self-investment is the primary determinant of
earnings while institutionalists argue that occupation in the principal factor. On the
surface, these stances seem to conﬂict; however, they have been merely addressing
opposite sides of the same coin. In other words, when viewed together through the lens of

occupation-speciﬁc human capital, the two positions can be uniﬁed to provide a simple,

20

eloquent framework within which to understand the relationship between self-investment,

occupation, and earnings.

The capacity to fuse these seemingly conﬂicting theories rests upon the recognition of the
ﬁve different dimensions inherent in each human capital investment, as previously
described. What this suggests is that when individuals make investment decisions, they

are inherently choosing to augment each respective cell of their human capital matrix,

(l’vitjl- However, different investment options alter each ‘l’vitjl in dissimilar ways, as

individuals make choices to acquire human capital along speciﬁc dimensions (for
example, work experience in the automotive industry, schoolwork in psychology, etc.).
Upon this conscious or unconscious accumulation of human capital along occupational,
industrial, ﬁrm, job, and general dimensions, individuals will then seek out employment

that provides the best utilitarian returns on their investment.

Generalizing this discussion to a speciﬁc investment (education) and dimension of human
capital (occupation), the reconciliation of the two schools of thought relies on the
recognition of the occupational components of schooling options, as acknowledged by
Fogel (1979) and Orazem and Mattila (1991). As such, when individuals make schooling
choices, they are, consciously or not, altering their human capital stocks along speciﬁc
occupational dimensions. For instance, extensive coursework in statistics builds
substantial occupation-speciﬁc human capital for particular vocations, such as actuaries,
statisticians, economists, and computer programmers, among others. Therefore, upon

completion of schooling, an individual will seek employment in professions that will

21

inherently yield the largest payoff to the human capital stock, which, after the person’s

education, is largest within certain occupational dimensions.

This advancement suggests that establishing the primary determinant of wages between
education and occupation is inherently a chicken-or-egg proposition. This perspective is
consistent with the importance the human capital approach places on education, as it
corroborates the notion that individuals will select the occupation that offers the
maximum return on their investment. At the same time, this hypothesis conﬁrms the
institutional emphasis on occupation in wage determination, as the occupational
components of the respective schooling options play a prominent role in occupational
selection decisions given the inherent wage implications of occupation-speciﬁc human
capital across the vocational spectrum (Kambourov and Manovskii 2005; Sullivan

2005a).'9

While different educational options will instill varying levels of occupation-speciﬁc
human capital, it is important at this point to denote the presence of vocational schooling
requirements that have moved some to suggest that education merely serves as a means
of entry to occupations (Fogel 1979). These formal and informal schooling barriers
undoubtedly play a role in the labor market (Taubman and Wales 1973), and individuals
often make occupational decisions before educational decisions, contrary to the “ﬁrst

education, then schooling” example cited by DeBeyer and Knight (1989). For instance,

 

19

This suggestion not only is conﬁrmed by ﬁndings of the dominant role played by occupation-speciﬁc

human capital in the determination of wages, but also allows for variability in the payoff to vocations and
their respective skills, or the “productivity shocks to occupations” noted by Kambourov and Manovskii
(2004a)

22

medical students do not push through four years of medical school and three years of
residency and only then decide to become physicians—the occupational selection was

what drove educational choice, making it, in some cases, a simultaneous outcome.

The reﬁned theory of human capital composition expressed here can account for the
earnings impact of vocational schooling requirements on educational and occupational
decisions.20 In particular, the presence of schooling requirements recognizes that

individuals must necessarily acquire a certain level of occupation-speciﬁc human capital

before they are able to reap a substantial payoff. In other words, the return to each ¢v

may not be linear; instead, in the presence of educational barriers, the return to schooling
within a given occupation may resemble a stepwise function. Not only does this explain
the educational premiums attached to degrees (Belman and Heywood 1991; Belzil and
Hansen 2002), but it also explains the common sense recognition that taking an
introductory class in physiology as a college freshman would have zero return as
occupation-speciﬁc human capital in the vocation of “physician” absent further pursuit of
a medical degree.” Thus, individuals may pursue educational choices based on the
acknowledgment of signiﬁcant payoff only to speciﬁc levels of occupation-speciﬁc

human capital.22

 

Rumberger (1987), for instance, demonstrates diminished return to education beyond the schooling
requirements of the particular job.

21

Students may maintain a level of uncertainty regarding the schooling requirements of occupations
(Robst and Cuson-Graham 1999). .

The introduction of potential step-wise returns to self-investment inherently contradicts the

productivity-wage relationship assumed within the human capital model. This assumption, based on
neoclassical economic assertions, suggests that increases in applicable productivity (such as that wrought
through a introductory physiology class) would be met with an increase in wages in the labor market.

23

From the labor demand perspective, the very existence of these schooling requirements
has been a source of contention in the debate between these two schools, as their presence
has typically been explained by the incomplete information held by ﬁrms in the hiring
process. While this represents the basic modus operandi of sorting models, the
recognition of occupation-speciﬁc human capital requires a subtle reformulation of this
approach.23 In their quest to select the most productive employee in light of incomplete
information, ﬁrms base their decisions on a variety of signals, including an applicants’
completion of speciﬁc educational tracks. Employers then utilize these signals to estimate
the occupation-speciﬁc human capital stock imparted to the applicant given the ﬁrms’

knowledge about the curriculum and intensity of the school.

Such an explanation also lies at the heart of occupational licensing issues in vocations
requiring individuals to obtain speciﬁc educational degrees in order to become certiﬁed
in a speciﬁc profession, such as physicians and teachers. These educational mandates for
vocational entry, such as a bachelor’s degree in education, are based on the belief that the
required schooling curriculum imparts, at a minimum, a speciﬁc, known amount of

occupation—speciﬁc human capital that would allow one to adequately perform in a given

 

While this may be true in general, this obviously does not hold within occupational markets given the
presence of vocational requirements (such as schooling barriers). Such a ﬁnding was stressed in Gottschalk
(1978); however other efforts (e.g., Hellerstein, Neumark, and Troske 1999) have maintained the
applicability of the productivity-wage correspondence.

As noted in Weiss (1995), it is important to recognize that sorting models inherently do not contradict

human capital principles, as they are built upon the same basic principles—proﬁt—maximizing ﬁrms and
utility-maximizing workers. Sorting models inherently are the ripple in the equation proposed by the human
capital model created by incomplete information in the labor market between ﬁrms and workers.

24

profession. Regardless of the motivation,24 the formal and informal requirements of
known levels of occupation-speciﬁc human capital (education, work experience)
represent the inherent barriers denoted by Kerr (1977) that balkanize the labor market,
thereby producing noncompeting groups that restrict worker ﬂow and allow for the

persistence of nonequilibrium outcomes, including the failure of labor market arbitrage.

While never referenced using these terms, occupation-speciﬁc human capital has
inherently been presented as an explanation for one of these contested outcomes, namely
segmented labor markets. Picard (1993), in particular, suggests that segmentation is a by-
product of differences in the type of labor required in various employment situations; in
particular, jobs in the secondary market involve only a few skills, thus denying workers
the ability to accumulate a repertoire of transferable skills necessary to gain entrance into
a primary-sector job. While others have referenced secondary work as “mini-skilled”
(Reiter 1991) and “deskilled” (MacDonald and Sirianni 1996), this argument has
suggested that workers are stuck in the secondary labor market because their work

experience fails to render them more marketable over time.

While a full explanation of segmented labor markets may involve a complex interaction
of supply- and demand-side factors (Bailey and Waldinger 1991), Picard’s explanation
corroborates the expected Outcome associated with occupation-speciﬁc human capital in

a labor market rife with vocational barriers and incomplete information. Essentially,

 

While some occupations, such as teachers and physicians, may use educational requirements as a

- means of ensuring public health and safety, other groups may insist on such mandates as a means of
restricting competition in order to increase market power.

25

Picard’s description suggests that workers in the secondary labor market have limited
ability to move into the primary labor market given their inability to develop occupation-
speciﬁc human capital due to the restricted set of skills necessary for secondary work and
the correspondingly low degree of transferability of those minimal skills into primary

occupations. Thus, considering that ﬁrms will hire the applicant with the presumed

maximum value of the appropriate ‘l’vitjl for a given job, the lack of any known

investment that yields signiﬁcant transferability in an environment of incomplete
information will lead such applicants to be continually passed over for work in the

primary sector regardless of their intrinsic capacities.

Given this explanation, the known accumulation of occupation-speciﬁc human capital
represents the balkanizing “barrier” not only between individual occupations, but also
between the primary and secondary sectors of the labor market. Not only would the skills
of primary-sector jobs maintain a higher degree of transferability that allows for greater
movement of workers 1 across vocations within the sector (e.g., from accountant to
teacher), but the accrual of occupation-speciﬁc human capital would also intrinsically
serve as a signal of unobservable factors regarding given applicants. Either way, the
accumulation of occupation-speciﬁc human capital serves as the foundation upon which
vocational and earnings outcomes are determined. As such, given the occupational
implications of educational investments, the underlying question regarding the primary
determinant of eamings—whether education or occupation—inherently fails to see the

forest for the trees; the correct answer is “both.”

26

Applications to the Wage Equation

A phenomenon commonly referenced in the segmented labor market theory literature
recognizes the stark contrasts in the return to education between the primary and
secondary sectors (Osterman 1975; Dickens and Lang 1985; Sakamoto and Chen 1991).
However, such a determination is only a part of the story. Thus, while it has become
increasingly apparent that the return to education is minimal among a certain cluster of
jobs in the secondary market, the return to education actually varies greatly within the
primary market, as illustrated by the varying returns to schooling across the entire

occupational spectrum (Eckaus 1973; Quinn and Rubb 2005; Sullivan 2005a).

While dissimilar educational returns by occupation may have represented a source of
institutional critique, such results fall directly out of a. human capital model incorporating
the occupational speciﬁcity of education, or, more broadly, the ﬁve dimensions inherent
in each investment choice. Too often, data restrictions lead researchers to view education
as a homogeneous duration variable while ignoring other vital aspects—including the
occupational components of the schooling choice (Orazem and Mattila, l99l)—that are
inherently responsible for generating certain labor market outcomes. However, education
is far from homogeneous, engendering markedly different responses within an
individual’s human capital matrix depending on the curriculum, intensity, and other
factors associated with the schooling choice. Correspondingly, one would expect a degree
rich in occupation-speciﬁc human capital (e.g., engineering) to producelarge wage

returns when applied within the respective vocation, as compared to a broad-based degree

27

(e.g., liberal arts). Further, the theory of occupation-speciﬁc human capital would expect
that any vocational switch would generate a signiﬁcant loss of rents, a ﬁnding conﬁrmed

by Kambourov and Manovskii (2005).

Given the inﬂuence of occupation-speciﬁc human capital on schooling outcomes in the
labor market, it is imperative to recognize how the characteristics of occupations and

schooling options interact to affect the potential level of human capital accumulation, or

the increase in ¢vitj 1, across investments. Thus, beyond duration effects of schooling, the

ability of education to impart speciﬁc knowledge along an occupational dimension of the
theorized human capital matrix depends heavily on the duties and tasks required by the
position. For instance, some jobs, such as those of economists or physicists, rely heavily
on educational investments to impart the expertise necessary to succeed in the position; in
other words, the accumulation potential through schooling along these occupational
dimensions of the matrix is quite signiﬁcant. However, for other vocations, such as
electricians or bricklayers, traditional educational programs may have scant applicability
in increasing productivity given that the duties of the position are difﬁcult, if not
impossible, to teach through classroom learning; instead, human capital accumulation

within those cells of the matrix relies heavily on work experience and on-the-job training.

While it is important to recognize that the characteristics of an occupation may dictate the
varying success of speciﬁc investments (e.g., education vs. experience), it is also
imperative to note that marginal productivity of any investment is not necessarily linear;

rather, the capacity of such investments to inspire growth in one’s expertise is, once

28

again, based on the characteristics of the job. For example, while basic math skills
learned in high school may be suitable for one to run a cash register, additional
mathematics training may yield little, if any, gains in productivity. Also, occupational
barriers may restrict marginal productivity, as demonstrated by truck drivers: after
learning the “tricks of the trade” in their ﬁrst year, movement from one’s sixteenth to
seventeenth year on the job may yield little, if any, additional productivity gains given the
presence of speed limits and hours-of-service regulations. Taken together, this suggests

that additional years of schooling may not yield signiﬁcant productivity gains within the

applicable (l’vitjl value within one’s human capital matrix. Rumberger (1987) conﬁrms

these ﬁndings, illustrating declining marginal returns to education beyond the schooling

requirements of one’s respective job.

Given these outcomes of the model using occupation-speciﬁc human capital, the failure
to ﬁnd a sizeable education premium in the secondary market relies on the fact that
secondary jobs, by deﬁnition, do not require a signiﬁcant amount of occupation-speciﬁc
human capital to adequately perform the required tasks of the position. As such, the
applicability of additional schooling beyond a necessary level would yield diminishing
marginal returns given the productivity restrictions of a particular occupation that fails to
provide ﬁrm incentives to hire educated workers at a premium rate. Further, worker
investment in work experience in these secondary-sector jobs only perpetuates the cycle,
as the failure of these jobs to provide an accumulation of occupation-speciﬁc human
capital leaves such workers “stuck” in these roles given primary-sector skill

requirements. On the contrary, it should be noted that education and experience produce

29

comparatively greater marginal earnings gains given occupational characteristics that

allow for larger productivity ranges in the primary sector.

Summarizing to this point, it is apparent that the ﬁve-dimensional composition of
investment options and the characteristics and productivity restrictions of occupations
justify occupation-speciﬁc returns to such investments. However, the standard
application of the human capital model—the wage equation—fails to allow for this
vocational speciﬁcity, as typical wage regressions implicitly assume equality of
investment return across the vocational spectrum (Freeman and Hirsch, 2001). Therefore,
given the implications of this paper, it is imperative to test the speciﬁcation of the

standard wage equation and its common exclusion of education-occupation interactions.

Beyond the use of interactions, the speciﬁcation of occupation itself in wage equations
represents an unexplored, yet subtly disruptive inﬂuence on research ﬁndings scattered
throughout labor economics research. Given the pervasiveness of the wage equation in
the ﬁeld, the failure to examine how occupational speciﬁcation affects other coefﬁcients
in the model is surprising. In particular, efforts to “include occupation” in the standard
wage model are extremely inconsistent, as the attempts in the literature range from
minimal (Arai 2003; Wheeler 2001) to sophisticated (Hirsch and Macpherson 2004).
However, how one speciﬁes occupation can produce unintended biases on other

coefﬁcients.

3O

As an example, consider a standard wage equation that includes an occupational control
for “legal occupations,” which would include such things as lawyers, judges, and
paralegals. However, upon estimation, the model will attribute the earnings differentials
between individual occupations within the category—such as lawyers and paralegals—to
any systematic difference between individuals within the two speciﬁc professions. In the
example of lawyers and paralegals, for instance, a regression model will attribute the
earnings differential between the two professions to the effect of education considering
the former will have substantially more. While this begs the chicken-or-egg question
regarding the source of the payoff—the occupation or the education—the example
illustrates, more broadly, that systematic differences across vocations within a category
may be misattributed, whether it be to education, race, citizenship status, or other
potential variables. In other words, simplistic efforts to “control for occupation” may
produce biased coefﬁcients given concealed relationships between vocation and other
control variables. Given this potentiality for bias, this study will not only examine
education-occupation interactions, but also address how occupational speciﬁcation

affects other coefﬁcients in the model.

DA TA AND MODEL

To examine occupational speciﬁcation and the use of vocation-education interactions in
the wage equation, this paper will utilize data extracted from the 2000 United States
Census One-Percent Public Use Micro Sample (PUMS) Files. As its title indicates, the

sample is drawn from the decennial census, with a standard battery of questions aimed at

31

demographics, income, occupation, and work habits for the year 1999. This cross-
sectional data set, while failing to contain a union variable, is otherwise ideal due to its
enormous sample size and its expansive occupational classiﬁcation system, allowing for a

very detailed exploration of particular labor market questions.

Since hourly wage is not an item in the PUMS ﬁle, it is calculated by dividing one’s
annual income by the total number of hours worked in the previous year. Considering
. that the goal of this paper is to examine basic features of occupation in the wage
equation, and not to isolate precise coefﬁcients, the fact that this calculation of wage
contains a number of statistical caveats (Borjas 1980) is not particularly worrisome. Also,
the derivation of hourly wage is limited only to work performed outside of self-
employment given the different environments faced by employees and the self-employed
(Fuchs, 1982; Gill, 1988; Carrington, McCue, and Pierce 1996). Those with wages of less

than $3 per hour were discarded.

Given the importance of the occupation and education variables in this study, it was
imperative to eliminate all potential bias generated through allocated, or “hot-carded”,
entries for these observations (Hirsch and Schumacher, 2004). Thus, for inclusion in this
analysis, individuals within the PUMS ﬁle must have reported their own occupation and
education. In addition, the sample was pared down to include only those who worked full
time (35+ hours per week, 48+ weeks per year) and were between twenty-ﬁve and sixty-
four years old. Finally, the analysis was limited to men in order to remove any potential

gender effects (Oaxaca 1973; Macpherson and Hirsch 1995). The resulting sample

32

included 395,244 males who met the eligibility criteria. The data is described in detail in

Table 1.

The baseline wage model applied in this study utilizes common human capital variables
included in standard wage equations, as shown below. It should be noted that industry
and occupation, as well as the occupation-education interactions, will be excluded from

initial models to establish a baseline comparison:

log(wage) = [30 + [31 education + Bzrace + B3marital_status + [34citizenship_status +
. . . 2 . .
Bsdzsabtlity_status + Béstate + B7age + Bgage + Bgmdustry + Blooccupatlon +

[3] 10ccupation *education + 8

In regard to the education variable, there has been a stream of research debating the
applicability of the linear schooling variable, which was the original suggestion of
Mincer (1974). While the linear variable continues to be employed by a number of
studies (e.g., Anderson and Shapiro 1996), one of the limitations of the linear

speciﬁcation is that it assumes an equal effect of schooling across all education intervals.

33

TABLE 1. Sample Means

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Human Capital Variables Mean Human Capital Variables Mean

Education Industry
Less than 9 years 0.032 Agriculture 0.016
High school, no diploma 0.066 Mining 0.009
High school diploma 0.276 Utilities 0.019
College: Less than 1 year 0.073 Construction 0.101
College: One or more years 0.156 Manufacturing 0.227
Associate’s ggree 0.075 Wholesale Trade 0.053
Bachelor’s Degree 0.196 Retail Trade 0.093
Master’s Degree 0.070 Transportation 0.065
Professional Degree 0.026 Information 0.032
Doctorate Degree 0.015 Finance, Insurance, Real Estate 0.054

Race Professional, Management 0.059
White 0.842 Administrative Support 0.027
Black 0.074 Education 0.044
Native American 0.012 Health Care 0.044
Asian 0.038 Recreation, Entertainment 0.013
Paciﬁc Islander - Hawaiian 0.002 Food Service 0.029
Other 0.050 Service - Other 0.040

Marital Status Public Administration 0.065
Married 0.729
Divorced or Separated 0.1 16
Widowed 0.005
Never Married 0.149

Age 41.810

Non-Citizen 0.064

Disability Status 0.140

Wage $21.24

 

 

 

 

This presupposition, however, has been rejected in favor of using a series of dummy

variables to signify different educational attainment (Belzil and Hansen 2002). This

allows for the possibility of “diploma effects” that are generally supported in the

signaling literature (Hungerford and Solon 1987; Belman and Heywood 1991). The

rejection of the linear schooling speciﬁcation also allows this study the ﬂexibility to

explore how occupation interacts with education at different levels of schooling. The ten

mutually exclusive educational categories used in this study, and their respective means,

are also listed in Table 1.

34

 

TABLE 2. Occupational Specifications

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

22 Major Occrﬁations (SOC #9 Count Education/ Training Categories Count
Management (1 1) 51,31 1 Professional Degree 10,326
Business and Financial Ops. (13) 16,111 Doctor’s Degree 3,595
Computer and Math. Science (15) 14,536 Master’s Degree 3,057
Architecture and Engineering (17) 16,685 Degree Plus Experience 33,183
Life, Physical, Social Sciences (19) 4,764 Bachelor’s Degree 58,951
Community and Social Services (21) 5,039 Associate’s Degree 8,192
Legal (23) 4,01 1 Vocational Award 17,522
Education, Training, and Library (25) 8,681 Work Experience 65,021
Arts, Design, Entertain, Media (27) 5,906 Long-Term On-the-Job Trainigg 47,330
Healthcare Practitioner and Tech. (29) 9,594 Moderate-Term On-the-Job Training 85,907
Healthcare Support (31) 1,418 Short-Tenn On-the-Job Training 62,160
Protective Service (33) 14,022
Food Preparation and Serving (35) 6,994
Building and Grounds Cleaning, (37) 11,412
Personal Care and Service (39) 2,600
Sales (41) 38,960
Ofﬁce and Admin. Support (43) 26,025
Farming, Fishing, and Forestry (45) 3,514
Construction and Extraction (47) 35,157
Installation, Repair, and Mainten. (49) 32,169
Production (51) 49,598
Transportation and Material Mov. (53) 36,737

 

Since the inclusion and speciﬁcation of vocational controls in the wage model lies at the

heart of this discussion, some mention of the occupational coding structure is required.

The PUMS data categorizes workers according to a three-digit, Census 2000 occupational

coding system. From there, the PUMS ﬁle converts these classiﬁcations into their

respective six-digit occupational icodes within the 2000 Standard Occupational

Classiﬁcation (SOC) system. Originally introduced in 1977, the SOC system covers all

vocations in which work is performed for pay or proﬁt, and is based on the work

performed within the job in addition to the similarity of skills, education, training, or

credentials required for the profession.25 The resulting structure allows for twenty-three

major occupational categories (including one military grouping), each of which is

 

Bureau of Labor Statistics. Revising the Standard Occupational Classiﬁcation System. June 1999.

http://www.bls.gov/soc/socrpt929.pdf.

35

 

denoted by a distinct ﬁrst-two-digit code. For example, “1 l-XXXX” denotes all
management occupations, while “l3-XXXX” signiﬁes all business and ﬁnancial
professions. These classiﬁcations become narrower as one moves to the third, fourth,
ﬁfth, and sixth digits of the coding structure, producing a classiﬁcation hierarchy that can
be applied to group occupations together on a number of different levels. As a result of
the SOC classiﬁcation, the PUMS data utilized here includes the 22 two-digit groupings
(excluding military) that contain each of the distinct 92 three-digit and 505 six-digit
classiﬁcations explored in this study. A listing of each of the 22 major categories, along

with their respective observation tallies, is provided in Table 2.26

RES UL TS

Occupational Speciﬁcation

The initial examination of the wage equation explores occupational speciﬁcation without
addressing the inclusion of occupation-educationinteractions. These results, presented in
Table 3, provide six models that analyze varying speciﬁcations of vocation in the wage
equation. Model 1 presents the base wage equation minus industry and occupation, with

Model 2 incorporating industry (but not occupation) to provide an intermediate step in

 

26

Because of minor incongruences between the occupational structures of the 2000 Census and the SOC,
31 low-population occupations were aggregated into 13 aggregated categories.

36

TABLE 3. Wage Determination, Males, Full Sample, Series of Education Dummies

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Model I Model 2 Model 3 Model 4 Model 5 Model 6
Education
Less than 9 years -0.244 -O.215 -0.194 -O.183 ‘ -0.172 -0.164
(45.99) (41.45) (37.89) (36.12) (34.19) (32.79)
High school, no -0.1 10 -0.101 -0.094 -0.088 -0.083 -0.079
diploma (30.07) (28.30) (26.77) (25.25) (24.20) (23.15)
High school diploma Base Base Base Base Base Base
(No college)
College: Less than 1 0.101 0.094 0.079 0.065 0.059 0.054
year (28.99) (27.37) 423.44) (19.54) (17.91) (16.34)
College: One or more 0.150 0.142 0.118 0.096 0.086 0.076
years (56.55) (54.28) (45.41) (36.96) (33.43) (29.91)
Associate’s Degree 0.184 0.171 0.146 0.107 0.095 0.086
(52.99) (50.22) (43 .32) (31.51) (28.36) (25.84)
Bachelor’s Degree 0.435 0.409 0.349 0.306 0.275 0.251
(173.55) (160.54) (134.39) (111.78) (99.36) (90.14)
Master’s Degree 0.541 0.540 0.461 0.430 0.400 0.362
(149.76) (146.57) (124.09) Q1106) (102.85) (92.02)
Professional Degree 0.805 0.775 0.752 0.653 0.548 0.410
(147.13) (136.61) (133.94) (94.66) (73.90) (49.52)
Doctorate Degree 0.622 0.681 0.627 0.598 0.537 0.482
(88.57) (95.81) (89.18) (81.97) (69.76) (61.28)
Race
White Base Base Base Base Base Base
Black -0. 127 -0.117 -0.099 ‘ -0.082 -0.073 -0.064
(38.16) (35.94) (30.70) (25.72) (23.14) (20.61)
Native American -0.1 14 -0.106 -0. 100 -0.094 -0.090 -0.086
(14.88) 44.15) (13.52) (12.84) (12.46) (12.10)
Asian -0.061 -0.064 -0.047 -0.071 -0.067 -0.069
(12.86) (13.93) (10.31) (15.77) (14.94) (15.57)
Paciﬁc Islander -0.037 -0.035 -0.033 -0.024 -0.022 -0.017
(2.06) (1.99) (1.90) g .42) (1.28) (1.01)
Other Race -0. 131 -0.1 15 -0. 101 -0.091 -0.086 -0.079
430.46) (27.31) (24.33) (22.29) (21.27) (19.62)
Marital Status
Single, Never Married Base Base Base Base Base Base
Married 0.190 0.175 0.159 0.155 0.148 0.142
(74.27) (70.09) (64.36) (63.72) (61.13) (59.1 1)
Separated, Divorced, 0.061 0.050 0.043 0.042 0.039 0.038
Widowed (17.63) (14.91) (12.93) (12.96) (12.12) (11.74)
Age 0.052 0.050 0.048 0.048 0.047 0.046
(70.63) (69.65) (68.12) (68.78) (68.25) (68.01)
Age2 -0.0005 -0.0005 -0.0005 -0.0005 -0.0004 -0.0004
(58.90) (57.45) (56.40) (56.58) (55.96) (55.75)
Non-Citizen -O.161 -0.145 -0. 135 -0.136 -0. 129 -0. 120
(40.85) (37.64) 95.46) (36.35) (34.65) (32.39)
Disabled -0.079 -0.073 -0.066 -0.062 -0.058 -0.054
(31.96) (30.38) (27.81) (26.43) (24.84) (23.38)

 

 

 

 

 

 

 

37

 

TABLE 3. Wage Determination, Males, Full Sample, Series of Education Dummies

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(cont)
State X X X X X X
Industry X X X X X
Occupation (7 X
controls)
Occupation (22, two- X
di it SOC)
Occupation (92, X
three-digit SOC)
Occupation (505, six- X
ﬂjgit SOC)
R-squared 0.2733 0.3089 0.3282 0.3450 0.3584 0.3755
Adjusted R-squared 0.2732 0.3087 0.3281 0.3448 0.3581 0.3746
Observations 395,244 395,244 395,244 395,244 395,244 395,244
F -Test Against Model X 1 197.01 1 137.48 489.39 *
1
F -Test Against Model X X 1035.99 338.35 *
2

 

Note: Absolute value of t-statistic in parentheses.

the analysis.27 Model 3 takes a ﬁrst attempt at including occupation, as it enveIOps seven
vocational controls in an effort to simulate crude occupational categorizations that arE
often applied in the literature. Model 4 expands this vocational speciﬁcation to twenty-
two classiﬁcations based on commonality of the ﬁrst two digits of each occupation’s

respective SOC code. Model. 5 magniﬁes this effect even further, including ninety-two

 

27

Wage equations that exclude job characteristics, such as occupation and industry, typically encompass

attempts to isolate the return to education. The method of estimating the return to schooling through the
wage equation has run into a number of issues, however its failure to include occupation has yet to be
addressed head-on. In particular, the simultaneous relationship between education, occupation, and wages
produces biased schooling coefﬁcients in light of the above conversation of occupation-speciﬁc human
capital and the concurrent decision-making between schooling and career decisions. As such, even a
rudimentary examination of the issue at least calls into question the practice of trying to isolate educational
returns absent occupational controls: using the principles of Baron and Kenny (1986), it can be illustrated
that occupation serves as a mediator between education and wages, an effect that is statistically signiﬁcant.
While it is undeniable that education plays a key role in occupational attainment, the disentanglement of
occupation from the retum-to-schooling has yet to be fully examined. For a full discussion of mediation
results, see Appendix A.

38

occupational controls based on SOC three-digit cormnonality. Finally, Model 6 presents

the maximum speciﬁcity—including controls for each unique six-digit SOC code.

While the results of individual models provide some insight, the more important results
stem from an examination of the trends across models. An initial analysis recognizes the
inﬂuence of occupational concerns in explaining wage variation in the model; not only
does the inclusion of job characteristics (industry and occupation) boost the adjusted r-
squared from 0.2732 (Model 1) to 0.3746 (Model 6), but the mere elaboration of
vocational speciﬁcation from crude (Model 3) to reﬁned (Model 6) was able to explain an
additional 5 percent of the variation in wages. The inclusion of occupational controls is

undoubtedly signiﬁcant.

In turning to individual variable trends, it is important to note that Table 3 presents the
coefﬁcient on the variable in the log(wage) equation followed by the absolute value of
the t-statistic below it in parentheses. Given the earlier discussion regarding the inherent
relationship between education and occupation, it should be rather unsurprising that the
reﬁnement of vocational controls in the wage equation considerably inﬂuences schooling
returns. For example, turning to the effect of a bachelor’s degree against a high school
diploma, Model 1 suggests that someone with a four-year degree is expected to earn 54.5
percent more than a high school graduate without any college experience.28 However,
upon adding simple controls (Model 3), that effect falls to 41.8 percent; elaborate

controls (Model 6) deﬂate the effect of a bachelor’s degree to 28.5 percent. Similarly, all

 

Note that percentages drawn from the wage equation are determined by e -1.

39

education effects dampen in light of additional vocational controls, systematically

moving toward zero as occupational categorization expands.

While the effect of occupational reﬁnement on schooling returns may be expected, its
inﬂuence on other parts of the wage equation represents the subtle means through which
speciﬁcation acts to shape research ﬁndings. In particular, given the wage equation’s
utilization as a means to explore racial discrimination, it is more than a little eye-opening
to examine the trend on the “black” variable in Table 3, especially spanning from Model
3 to Model 6, as the estimated white-black earnings gap falls from 10.4 percent to 6.6
percent, respectively. As mentioned previously with the lawyer-paralegal example, this
demonstrates that, in Model 3, the wage equation misattributed earnings variation to race
when it was instead due to differing occupational compositions between whites and
blacks. While this may represent another form of discrimination, it is nevertheless vital to
recognize that analyses of discrimination that utilize the wage equation produce results
that are inherently dependent upon their speciﬁcation of occupation given the suspected
correlation between race and the occupational composition within broad vocational

groupings.

While all coefﬁcients in the model ﬂuctuate somewhat given the elaboration of
occupation between Model 3 and Model 6, another effect to particularly note comes in an
examination of the estimated discrimination between whites and Paciﬁc Islanders. In
particular, Model 3 indicates a statistically signiﬁcant effect while a more reﬁned

approach to vocation leads to an insigniﬁcant coefﬁcient. What makes this ﬁnding

40

particularly interesting is that it occurred using a sample of nearly 400,000; given that
most efforts in the literature will be working with a drastically smaller sample, the
possibility of occupational speciﬁcation inﬂuencing the statistical signiﬁcance of an
outcome grows exponentially. Therefore, wage regressions should take special precaution
to examine the robustness of their efforts to the vocational speciﬁcation utilized in the

model.

Occupation-Education Interactions

As discussed earlier, there are a number of reasons that imply occupational—speciﬁcity of
educational returns, extending from the vocational components of schooling options to
the intrinsic productivity (and thus wage) ranges offered by the profession. However, as
noted in Freeman and Hirsch (2001) and Sullivan (2005a), typical wage equations
inherently assume homogeneity of investment returns across the vocational spectrum by
failing to include interactions with occupational controls. While data from the PUMS
fails to allow content differences in schooling options, it does allow the potential to
examine the vocational speciﬁcity to the duration of schooling, an important option given

that most data sets express educational attainment only by its length.

Analysis of occupation-education interactions is presented in Table 4, which, for the sake
of simplicity, utilizes a set of seven occupational controls interacted with each of the ten
educational variables. While Table 4 lists only a portion of those interactions, the

individual signiﬁcance of the presented coefﬁcients pales in comparison to an

41

examination of their joint signiﬁcance—that is, whether such interactions should be
included in the ﬁrst place. While an F -test on these seventy interactions is indicative of
their signiﬁcance (F = 36.96, p < 0.0001), an LM test-employed given the large sample
size utilized in this study—also is suggestive of the overwhelming signiﬁcance of

including interactions in the model (LM = 4084.97, p < 0.0001).

In relation to speciﬁcation of the wage equation, this paper has examined three basic
speciﬁcations of the wage model in relation to its efforts to control for education and

occupation:

Table 3, Models 1-2: Wageij = [3i*educationi
Table 3, Models 3-6: Wageij = Bi*educationi + aj*OCCupationj

Table 4, Model 7: Wageij = Bi*educationi + aj*occupationj + yij*interactionij

While the ﬁrst two speciﬁcations represent restricted versions of the latter (i.e., aj = 0 and

Yij = 0), the results in Table 4 demonstrate that such restrictions are misspeciﬁcations of

the wage model. These ﬁndings—combined with the results of Freeman and Hirsch (2001)
regarding the signiﬁcance of occupation-experience interactions—suggest that full
speciﬁcation of the wage model requires that interactions of self-investment and

occupation be utilized. Through this relaxation of homogeneity assumptions, the model

42

TABLE 4. Wage Determination, Males, Full Sample, With Occupation*Education
Series of Dummies

    
   

 

year

01' more years

Interaction Degree Degree

t-statistic race,
status, age, age squared, citizenship status, disability status, state of and industry.

43

would allow the ﬂexibility required to explore wage determination in a balkanized labor

market wrought with short-run disequilibria due to occupational barriers.

DISCUSSION

While the wage equation developed by Mincer (1974) continues to represent a foundation
of labor economic analyses, there has nevertheless been an exhibition of nonchalance in
the literature regarding the speciﬁcation of occupation in the model. While data
restrictions are often to blame, the results of this paper demonstrate that how one
speciﬁes vocation can have subtle, yet signiﬁcant, inﬂuence on research ﬁndings. In
addition, this paper—along with Freeman and Hirsch (2001)—demonstrates that the
implicit assumption of the wage equation regarding homogeneity of returns to human
capital across occupations clearly fails. A more detailed examination of wage

determination requires the inclusion of occupation-investment interactions.

These issues of occupational speciﬁcity in the wage equation have only masked a larger,
unresolved ambiguity in the role of vocation from the underlying theory upon which the
wage equation was based—the human capital model. Thus, this paper has indicated that
reﬁning the basic composition of human capital—beyond general and ﬁrm-speciﬁc—may
provide the necessary theoretical framework upon which to understand some of the labor
market phenomena for which it has been criticized in institutionalist literature. In
particular, reﬁning human capital to include occupation-speciﬁc elements in light of

distinct vocational balkanization in the labor market may provide an explanatory

44

framework for the model in regard to sorting models, segmented labor markets, and other
phenomenon, thus providing the means to explain and predict the outcomes that have

served as the basis for forty years of dissent in certain circles.

While this paper represents a ﬁrst start at recognizing the broader implications of reﬁned
human capital deﬁnition, the recent revitalization of work on occupation-speciﬁc human
capital reveals that much work remains in understanding how occupation and self-
investment interact in the labor market to produce various outcomes. The role of
occupational speciﬁcity permeates such outcomes as wage determination, occupational
selection, mobility, turnover, ﬁrm investment, and other basic labor market conditions. In
particular, understanding the simultaneous nature of wages, education, and occupation—
and the aspirations therein-represents a key inﬂuence in employment situations that has
yet to be fully examined or explained. Further research on these issues may provide
further reﬁnements on the human capital framework that will allow researchers a more

powerful, sophisticated lens through which to view the labor markets and their outcomes.

45

Chapter 2

DISPLACED MANUFACTURING WORKERS
AND OCCUPATION-SPECIFIC HUMAN CAPITAL

INTRODUCTION

As of January 2006, manufacturing employment in the United States had sunk to its
lowest level since 1950, and it has remained relatively ﬂat since a record three-year
decline of 2.8 million workers between 2001 and 2003. This rapid deterioration has left
an indelible mark on the public consciousness, especially in light of near-daily headlines
announcing plant closings at companies once thought to be the foundation of American
industry. The concurrent explosion of the service sector as a part of the economy’s well-
publicized structural transformation has engendered the social myth that laid-off blue-
collar manufacturing workers have little option but to retire early, stay in the
unemployment line, or end up working at a “big box” retailer. However, despite volumes
of research examining post-displacement wage outcomes, scant interest has been directly
paid to the occupational outcomes associated with displacement. This lack of attention is
surprising given public attention on retraining initiatives undertaken to assuage the

inherent mismatch of skills and jobs associated with structural shifts in employment.

At its foundation, understanding occupational movement in the midst of a structural

industrial transformation requires an examination of the potential applicability of the

46

knowledge, skills, and abilities (KSAs) utilized by blue-collar workers in the
manufacturing industry to other vocations in the labor market. A recently revitalized
literature on occupation-speciﬁc human capital has particular relevance, as research has
begun to demonstrate its importance in determining labor market outcomes (Milgrom
2004; Kambourov and Manovskii 2004a, 2004b, 2005; Sullivan 2005a, 2005b; Groen
2006). This current rejuvenation of occupation-speciﬁc human capital literature comes on
the heels of two decades during which progress lay dormant following the work by Shaw
(1984), who originally proposed such an aspect of human capital by deﬁning
occupational investment as “the accumulation of skills an individual acquires to perform

9”

work within an ‘occupation (p. 320). Shaw also theorized that occupation-speciﬁc skills
possessed a certain degree of “transferability” across vocations and across employers.
Applying this deﬁnition to the plight of displaced workers, one would expect that such
individuals would seek to enter occupations in which they could apply-and receive a

return on-the greatest percentage of their KSAs; in other words, occupations in which

they would have the greatest degree of transferability.

Despite the potential importance of occupation-speciﬁc human capital as it relates to
labor market outcomes, research to this point has failed to develop an applicable
construct to estimate cross-occupation transferability of KSAs. This paper attempts to
correct this shortcoming, using data from O*NET to develop a 900 x 900 matrix
estimating the transferability of KSAs across occupations. Condensing this into a 500 x
500 matrix capable of matching Census occupational codes, this study utilizes this

transferability matrix to examine post-displacement outcomes using the 2004 Displaced

47

Worker Survey. The results demonstrate that displaced) blue-collar workers from the
manufacturing industry suffer signiﬁcantly greater occupation-speciﬁc hmnan capital
losses than any other labor market group, thereby verifying that the structural shifts in the
economy are more prominently harming this sector, as experienced by signiﬁcantly more

pronounced mismatches of skills with available jobs.

Examining occupational movement by laid-off blue-collar workers from the
manufacturing industry, this paper will be structured as follows. First, this piece will
provide a review of prior literature on post-displacement outcomes, as well as
background on the small but growing literature on occupation-speciﬁc human capital.
Next, this piece will discuss O*NET and the development of the transferability matrix
and its potential role in the literature. Finally, the paper will examine occupational
outcomes associated with displacement from the manufacturing sector, looking not only
at summary trends but also applying the transferability matrix to examine the movement

of human capital across vocations upon displacement.

LIT ERA TURE RE VIEW

Despite extensive literature on the topic, what deﬁnes a “displaced worker” has been a
source of contention, given its implications of government responsibility for
compensating the losers in structural economic change (Fallick 1996). To build
consensus, Jacobson, LaLonde, and Sullivan (2005) identiﬁed three common

characteristics of most descriptions of diSplaced workers: “(1) They have not been

48

discharged for cause; (2) they have permanently separated from their former employer or
have only a very small likelihood of being recalled to their old jobs; and (3) they have.
had strong prior attachment to the industry of their pre-displacement employer,” (p. 48).
Given the importance of workers’ strong prior attachment to their pre-displacement job in
determining post-displacement outcomes, the Bureau of Labor Statistics typically deﬁnes
displaced workers as persons having at least three years of tenure upon their permanent

job loss.

A majority of the research on displaced workers has focused on earnings losses
attributable to displacement, as studies have demonstrated the large and persistent costs
that accompany displacement over an individual’s work history. The literature suggests
that, ﬁve years after displacement, earnings typically remain somewhere between 15 and
30 percent lower than a worker’s pre-displacement level (among others, Topel 1991;
Ruhm 1991; Jacobson, LaLonde, and Sullivan 1993; Schoeni and Dardia 1996; Stevens
1997; Kletzer 1998; Kambourov and Manovskii 2005; Jacobson, LaLonde, and Sullivan
2005).29 A speciﬁc stream of the literature has consistently identiﬁed signiﬁcantly larger
earnings losses associated with workers’ re-employment attachment to a different
industry than that of their former employer (Addison and Portugal 1989; Jacobson,
LaLonde, and Sullivan 1993; Carrington 1993; Neal 1995; Parent 2000). While a
considerably smaller literature, recent work has also identiﬁed similar effects within
occupations, as Kambourov and Manovskii (2005) denote that, among males aged twenty

to sixty, those switching occupations experienced an 18 percent drop in earnings

 

29

For a more extensive review of displaced worker literature, see, among others, Fallick (1996), Kletzer
(1998), F arber (2005), and Jacobson, LaLonde, and Sullivan (2005).

49

compared to a 6 percent decline for those who stayed in their pre-displacement

occupation.30

The above results have engendered a more sophisticated View of the human capital
interpretations of displacement, as some have begun to contend that human capital
maintains occupation- and industry-speciﬁc aspects that blur the boundaries between
“general” and “ﬁrm-speciﬁc” characteristics. While the recognition of occupational
speciﬁcity and its inﬂuence on the labor market dates back to Adam Smith (1776), Shaw
(1984) was the ﬁrst to apply such principles to the human capital framework. Deﬁning
“occupational investment” as the “accumulation of skills an individual acquires to
perform work within an ‘occupation,’” (p. 320),31 Shaw suggested that general human
capital maintained an occupation-speciﬁc component that could be carried across
employers. While differing opinions may exist on the potential dissection of human
capital into distinct components (as discussed in Chapter 1), the inﬂuence of occupation-

speciﬁc human capital on labor market outcomes has been demonstrated by a growing

 

Fox (1994) presents the logical explanation, suggesting that earnings drive occupational decisions: if

workers can fare better in a different occupation, they are more likely to switch. As such, the idea of
“occupational choice” may be a misnomer, given the inﬂuence of local labor market environments,
resource availability, or other restrictions that dictate a restricted set of outcomes to an individual.
Carrington (1993) also examines the role of occupation in earnings losses, however the utilization of one-
digit occupational codes represents an overly crude measure that likely conceals signiﬁcant within-control
effects.

Shaw (1984) deﬁnes an “occupation” as “a homogeneous skill classiﬁcation within which individuals

are perfect substitutes in demand and/or have inﬁnite cross elasticities of substitution in supply. Cabinet-
makers and engineers, for example, have occupation-speciﬁc skills,” (p. 320).

50

literature (Milgrom 2004; Kambourov and Manovskii 2004a, 2004b, 2005; Sullivan

2005a, 2005b; Groen 2006).32

Central to the notion of occupational speciﬁcity, Shaw (1984) recognized the importance
of transferability in establishing particular labor market outcomes. Distinguishing
occupation-speciﬁc from ﬁrm-speciﬁc human capital, Shaw noted that the return to the
stock of the former was not completely foregone upon an occupational switch; instead,
the return to one’s human capital stock depended upon the transferability of skills from
one occupation to the next. For instance, the knowledge, skills, and abilities procured
through one’s time as an accountant may be extremely applicable to becoming an
actuary. In contrast, the KSAs acquired. during one’s time as a dental hygienist may be
practically useless as a carpenter. Transferability thereby depends upon the commonality
of the KSAs necessary in one occupation as it compares to others across the vocational
spectrum; it thus serves a vital role in understanding occupational movement given the

varying rates of return of one’s acquired skills within different professions.

While cross-occupational transferability of human capital has particular application to

occupational choice, retraining initiatives, and a variety of other labor market

 

32

It is important to note that while Shaw (1984) established the dominance of occupational investment in

determining wages compared to “general” human capital, there has been some issue in regard to the
superiority of occupation-speciﬁc versus industry-speciﬁc human capital. While Kambourov and
Manovskii (2004a, 2005) suggest that industry-speciﬁc human capital plays little role in wage
determination upon the inclusion of occupation-speciﬁc human capital in the model, Sullivan (2005a)
contends that industry-speciﬁc human capital indeed plays a signiﬁcant role after including within-ﬁrm
occupational switches. 1f the former is accurate, however, it begs the question regarding early displaced
worker examinations that isolated industrial effects absent occupational concerns were ﬁndings of
signiﬁcant effects of industry-switching attributable to a proxy for occupational distribution changes?
While this effect of industry-speciﬁc human capital is an issue of debate, it should be noted that the role of
ﬁrm-speciﬁc expertise in wage determination has increasingly been established as marginal (Lazear 2003;
Kambourov and Manovskii 2004a, 2005; Milgrom 2004; Poletaev and Robinson 2004).

51

phenomena, research establishing the vocational commonality of characteristics has been
limited: Shaw (1984) represents the only known attempt to estimate values of cross-
occupation transferability. Shaw proxied such numbers by examining occupational
switches, primarily through a retrospective vocational question in the 1970 Census,
arguing that there will be greater occupational mobility between jobs that have higher
rates of vocational skill transferability. In other words, one would expect to see higher
occupational mobility between psychologists and counselors than between psychologists
and electricians, given the differences in acquired skills, or variation in applicable

occupation-speciﬁc human capital.

DEVELOPING NEW INDICES

While Shaw’s work laid the foundation for future work on occupation-speciﬁc human
capital, the use of occupational switching incidence to proxy transferability fails to
explicitly examine the knowledge, skills, and abilities actually transferred across
occupations, leading to a ﬂawed set of estimates. In other words, while experience as a
sports writer may have signiﬁcant impact on one’s productivity as an economics
professor through the development of one’s writing skills, the lack of actual labor market
mobility between the two ﬁelds would result in an estimated transferability approaching

zero using Shaw’s method.33 Thus, given the workings of the labor market and the

 

Without careful analysis, proxying transferability through occupational switches can ignore the

intermediate career steps between vocations. For instance, while there may be transferability between
sports writers and economics professors, the occupational switches in the data would reveal non-zero
transferability between (a) sports writers and graduate assistants and (b) graduate assistants and professors,
but not (c) sports writers and professors. While the solution to this problem would be to, say, multiply the
transferability in switch (a) and the transferability in switch (b) to establish the transferability in switch (c),

52

inherent occupational barriers that restrict mobility (Kerr 1977), the development of
cross-occupational transferability estimates would be considerably better served by
comparing the actual KSAs utilized within each vocation and developing a corresponding

metric based on their applicability across the occupational spectrum.

This paper attempts to rectify this shortcoming in the occupation-speciﬁc human capital
(OSHC) literature by using data from O*NET, the authoritative national database on
vocational information coordinated by the United States Department of Labor. O*NET
data characterizes 900 distinct professions along 120 standardized knowledge, skill, and

ability categories (listed in Table 5), and is applied in this paper to generate two indices

of occupational comparability. The ﬁrst, tij, represents an estimated proportion of

occupation-speciﬁc human capital that would be transferred from one’s origin occupation

of (occi) to one’s destination occupation (och'). Corresponding to the deﬁnition of

“transferability” established by Shaw (1984), the estimates of tij are derived by the

computation of a ratio of shared KSAs between two occupations to the KSAs required
within the origin occupation. As demonstrated by Shaw (1984), higher rates of
transferability, or the ability to apply more of one’s acquired OSHC in a new job, should

yield higher earnings, all else equal.

 

such tactics would be particularly thorny, given the near-inﬁnite career paths available and amount of data
needed to estimate such values.

53

TABLE 5. Knowledge, Skill, and Ability Categories Used by 0*NET

 

Knowledge

Skill

Ability

 

Administration and Management

Active Learning

Arm-Hand Steadiness

 

Biology

Active Listening

Auditory Attention

 

Building and Construction

Complex Problem Solving

Category Flexibility

 

Chemistry

Coordination

Control Precision

 

Clerical

Critical Thinking

Deductive Reasoning

 

Communications and Media

Equipment Maintenance

Depth Perception

 

Computers and Electronics

Equipment Selection

Dynamic Flexibility

 

Customer and Personal Service

Installation

Dﬂiamic Strepgth

 

Design

Instructigg

Explosive Strength

 

Economics and Accounting

Judgment and Decision Making

Extent Flexibility

 

Education and Training

Learning Strategies

Far Vision

 

Engineering and Technology

Management of Financial
Resources

Finger Dexterity

 

English Language

Management of Material
Resources

Flexibility of Closure

 

Fine Arts

Management of Personnel
Resources

Fluency of Ideas

 

Food Production

Mathematics

Glare Sensitivity

 

Foreign Lamguage

Monitoring

Gross Body Coordination

 

Geography

Negotiation

Gross Body Equilibrium

 

History and Archeology

greration and Control

Hearing Sensitivity

 

Law and Government

Operation Monitoring

Inductive Reasoning

 

Mathematics

Operations Analysis

Information Ordering

 

Mechanical

Persuasion

Manual Dexterity

 

Medicine and Dentistry

Programmﬂ

Mathematical Reasoning

 

 

 

 

 

Personnel and Human Resources Quality Control Analysis Memorization
Philosophy and Theology Reading Comprehension Multilimb Coordination
Physics Repairing Near Vision

Production and Processing Science Night Vision
Psychology Service Orientation Number Facility

 

Public Safety and Security

Social Perceptiveness

Oral Comprehension

 

Sales and Marketing

Speaking

Oral Expression

 

Sociology and Anthropoltgy

Systems Analysis

Originality

 

Telecommunications

Systems Evaluation

Perceptual Speed

 

Therapy and Counseling

Technology Design

Peripheral Vision

 

Transportation

Time Management

Problem Sensitivity

 

Troubleshooting

Rate Control

 

Writing

Reaction Time

 

Response Orientation

 

Selective Attention

 

Sound Localization

 

Spatial Orientation

 

Speech Clarity

 

Speech Recognition

 

Speed of Closure

 

Speed of Limb Movement

 

Stamina

 

Static Strggth

 

Time Sharing

 

 

 

 

 

Trunk Strepgth
Visual Color Discrimination

 

54

 

TABLE 5. Knowledgg, Skill, and Ability Categries Used by O*NET (cont)

 

Visualization

 

Wrist~Finger Speed

 

Written Comprehension

 

 

Written Expression

 

 

 

The second index established in this paper, qij, represents a “qualiﬁcation” rate, or the

estimated proportion of occupation-speciﬁc human capital required in the destination
occupation that can be instilled through work in the origin occupation. A second index is
required to understand movement within the labor market because while workers in
entry-level occupations may be able to transfer most of their acquired OSHC into higher-
order professions (i.e., high transferability), it is an individual’s level of qualiﬁcation that
is most important in determining whether an individual secures entry into these higher-
order vocations. In other words, while an entry-level retail salesperson may be able to
transfer almost all of his or her OSHC into a role as a Sales engineer, work in the former
vocation alone represents woefully inadequate training for the latter profession, making

entry into this higher-order occupation unlikely without further vocational or educational

training. Previously ignored in the occupation-speciﬁc human capital literature, qij is

estimated as the ratio of shared KSAs between two occupations to the KSAs required

within the destination occupation.

The derivation of these two occupational indices requires additional explanation of the
O*NET data and the methodology involved with creating these measures. In regards to
the former, it is necessary to note that O*NET scores the KSAs utilized in each

profession along two separate indices: (a) the level of each KSA required of one to

55

 

perform adequately, scored on a 0-7 scale, and (b) the importance of each KSA to the
responsibilities of the occupation, scored on a 1-5 scale.34 This two-dimensional approach
is important in capturing both the breadth and the depth of KSAs required in each
profession. For example, while knowledge of “Medicine and Dentistry” may be equally
important to the duties of a paramedic and, a physician, the level of such knowledge
required is vastly. different across the two professions.3S Therefore, as a means of

consolidating the two scores into a meaningful, unitary metric within each KSA category

for a given occupation (necessary for future steps), this paper advances the use of 5im or

the product of the importance and level scores for each KSA dimension, m, within each
occupation, i. By weighting the importance of a KSA by its level, this step generates a

“point” total encapsulating the breadth and depth of that characteristic within an

occupation. This produces 120 values of 5im for each occupation, one for each of the 33

knowledge, 35 skill, and 52 ability categories employed by O*NET to characterize each

vocation. Ranging from 0-35,36 51m “point” totals within knowledge categories for

human resource managers, for example, include: administration and management

 

While O*NET lists the level and importance of each KSA on a lOO-point scale in the on-line version,

this represents a mere transformation of the developer format that scores data on a 0-7 (level) and 1-5
(importance) scale. Note that these scales are not restricted to integers; for example, the score for a level
may be 5.16.

As an opposing example, consider writing skills for psychology professors and newspaper reporters.

While the two occupations may require similar levels of expertise, the importance of writing across the two
vocations is quite different. While the reporter’s entire job may revolve around his or her writing skill,
psychology professors may succeed through the utilization of other skills, such as instructing students,
providing counseling services, and organizing research projects.

The maximum is rarely met in a speciﬁc category, as only selected occupation-KSA categories are
scored at 35 (e.g., surgeons and knowledge of medicine and dentistry).

56

(21.735), biology (0.000), clerical (5.033), personnel and human resources (30.800), and

telecommunications (1.538).

Upon computing each of the 120 51m values for all 900 occupations scored by O*NET,

the next step is to make pair-wise comparisons of 5im values across vocations in order to

establish the number of shared KSA “points” between two occupations; this value

represents the numerator in the computation of both tij and qij. This process is
demonstrated in Table 6, as four of the 5im knowledge scores are compared across two

occupations: economist (occi) and human resources manager (och'). Looking within the

“Computers and Electronics” category, economists use 8.84 points while human
resources managers utilize 5.00 points. Within this standardized category, it can be
deduced that the two occupations “share” 5.00 points; while HR managers can apply all
5.00 of their points in their role as an economist, given that it requires greater knowledge

in that category, economists can apply only a portion of their 8.84 points as an HR

manager, given the latter’s more limited requirements. As a result, deﬁning ‘l’ijm as the

shared point total between occupations i and j within a given KSA dimension, m, the

formula becomes:

(1) ‘l’ijm = min(5im, 5jm)

57

TABLE 6. Comparing 5im Scores to Establish Applicability of Required KSAs

 

 

 

 

 

 

6;," Scores
Personnel &

Computers & Human Communications Economics &
Occupation (SOC) Electronics Resources & Media Accounting
i: Economist
(193011.00) 8.84 6.72 3.24 24.64
j: Human Resources
Manager (1 “304000) 5.00 30.80 3.87 6.20
Shared Point Value 5.00 6.72 3.24 6.20

 

 

 

 

 

 

Upon the calculation of the 120 (l’ijm values for each ij occupation combination, this

paper ﬁrst turns its attention to the derivation of tij, the ﬁrst metric described above

estimating transferability, or the proportion of OSHC from one’s origin vocation that can
be applied in one’s destination occupation. This value is estimated by the ratio of the
number of shared points between occupations i and j to the total number of points within
the origin occupation i. While the direct summation of each value-would assume equal
weighting between the 120 KSA categories, this paper advances a slight variation, instead
giving equal weight to the knowledge, skill, and ability categories as a whole considering

the unequal number of dimensions scored within each category by O*NET:37

 

37

As of June 2006, version 9.0 represented the latest release of O*NET data. However, upon comparing
data across occupations, it became clear that there were nonsensical, systematic outcomes that derived from

working with this data; for example, upon summing 5im scores, version 9.0 data listed “steamﬁtters” as the

occupation requiring the most knowledge across the 900 occupational categories used to deﬁne the labor
market, higher than such vocations as physicists and physicians. Considering that this represented one of
many nonsensical results, this paper utilizes O*NET version 4.0, which solely applies expert analysis of
occupational characteristics. While no known paper has outlined the reasoning behind such inconsistencies .
across occupations in O*NET’s later versions, the likely cause lies with the data collection process
underlying O*NET. In particular, to update their data-a key feature of the system as advertised by O*NET
itself-O*NET surveys a random sample of workers within given occupations to produce “updated”
numbers regarding the level and importance of speciﬁc knowledge, skills, and abilities within the given
vocation. However, given that the developer version of O*NET 7.0 shows update sample sizes of ﬁfteen to

58

 

 

2 t0. =avg
” 5,, 6,- 6.

 

 

To understand how this works, consider the abbreviated example in Table 7, which

revisits the four knowledge scores (61m) for two occupations-economists and human

resources managers. With the sum of shared points (21.16) between the two vocations

serving as the numerator (‘l’ijml and the total points of the origin occupation (43.44)
serving as the denominator (Sim), the resulting value represents the percent of KSAs that

would be transferable from one occupation to the next, or tij=0.4871. By summing this

within all knowledge, skill, and ability categories, and then averaging the totals to
provide equal weighting, the resulting value represents the estimated proportion of OSHC

transferability as pursued by Shaw (1984).

 

thirty within each occupation to serve as the means of updating speciﬁc vocational totals, it very well could
be that small sample sizes are creating hidden, yet substantial, inconsistencies in the comparison of
characteristics across occupations. Further investigation of the causes of these nonsensical outcomes would
represent a signiﬁcant boon to researchers, practitioners, and developers.

59

TABLE 7. Demonstrating the Develtmment of Transferability

Scores

 

 

 

 

 

 

5im Scores
Occupation til _
(SOC) Computers Personnel & Communi- Economics .
& Human cations & & Sum Sum ¢i~lm/
Electronics Resources Media Accounting (5im) (Him) 61m
i: Economist
(19-30] 1.00) 8.84 6.72 3.24 24.64 43.44
j: Human
“mm“ 5.00 30.80 3.87 6.20 45.87
Manager (1 1-
3040.00)
Shared Point
5.00 6.72 3.24 6.20 21.16 0.4871
Value (¢iim)

 

 

 

 

 

 

 

 

While the preceding example represents an examination of two occupations and four

KSA dimensions, it should be noted that the full expansion of the model produces a 900 x

900 matrix of transferability scores, T. While this matrix is too cumbersome to be

presented in its entirety within this paper,38 Table 8 advances a 5 x 5 snippet of T in order

to provide some clarity on the structure of the larger matrix. As an example, t15 suggests

that only 7.72 percent of the KSAs utilized by a government service executive (occl) can

be applied to one’s position as a refuse and recyclable material collector (occ5).

Similarly, t24 indicates that 18.39 percent of the KSAs utilized by an industrial

production manager (occz) can be applied to work performed by a bread and pastry

bakers (OCC4). Conversely, t42 suggests that 96.48 percent of the KSAs used by a bread

 

The full matrix is available in electronic form from the author.

60

 

TABLE 8. 5x5 Fragment of T, the KSA Transferability Matrix (Values Represent

- /
.00

— I.
100

        

are as government service executives — )
production managers ([1305 I 00), procurement clerks (43- 3061. 00). bakers bread and pastry (5]-
3011.01) and refuse and recyclable material collectors (53- 708 I 00).

and pastry baker can be applied to work as an industrial production manager. It is

important to note that, while t24 and t4; have the same numerator (shared points), the

denominators (total points within the origin occupation) differ. Thus, considering that

industrial production manager jobs are more demanding than baker positions (and thus

have a larger denominator), t24 does not equal t42; this mirrors the larger trend that tij

does not equal tji, thereby accounting for the lack of symmetry in the matrix. Table 8 is

shaded along the diagonal to reinforce this asymmetry.

The derivation of the qualiﬁcation rate, qij, represents a slight variation from the

computation of the transferability rate, tij. While the two values maintain the same

numerator-the shared point value between occupations i and j-the denominator of the
qualiﬁcation rate is the total point value of the destination occupation, j. Therefore, the

formula for the qualiﬁcation rate, can be represented as follows:

61

Z ¢ijk Z ¢ijs a2 fbra

 

 

 

 

(3) qij =avg ’ 5135 ’ 52
261k 251's 02510
K k: 1 s: l a)

Just as with the transferability rates, the development of qualiﬁcation rates produces a
900 x 900 matrix for each occupational pair-wise combination. This qualiﬁcation matrix,

Q, provides the degree to which work in occupation i prepares one for work in occupation

j across the entire vocational spectrum. However, given the deﬁnition of each Qij

comprising the matrix, it is important to note the following:

(4) (lij: tji

This follows from the fact that both terms, by deﬁnition, use the same numerator (shared
KSA points between occupations i and j) and denominator (total KSA points in

occupation j). Therefore, generalizing this to matrices created, Q is the transpose of T, or:

(5)Q’ =

62

TABLE 9. 5x5 F of and T

       
   

T and (A cross)

Occupation T and Q,‘ (Down) ”401] 1_

]_

      

are as government service executives - ),
production managers (II- 305 I 00), procurement clerks (43- 3061. 00), bakers, bread and pastry (51-
3011. 01); and refuse and recyclable material collectors (53- 7081. 00).

To help explain this, Table 9 re-examines the 5 x 5 snippet of the matrix presented in
Table 8. This version allows one to access the qualiﬁcation and transferability rates

between any two occupations i and j. To interpret this table, one can ﬁnd the appropriate
tij by locating the column containing occupation i across the top of the table and then
moving down the matrix to ﬁnd the row housing occupation j; similarly, one can
determine qij by ﬁnding the row of occupation i down the side of the table and then

moving across the matrix to locate the column containing occupation j. For instance, if

one wants to know t13, or the transferability rate of OSHC from government service

executives (occl) to procurement clerks (occ3), the appropriate number is found by

moving over to the ﬁrst column and then down to the third row to note that 32.74 percent

of the KSAs used in the former occupation are applicable in the latter occupation.

Conversely, if one wants to ﬁnd q13, or how much working as a government service

executive “qualiﬁes” one to work as a procurement clerk, the result is found by looking at

63

the ﬁrst row and following it to the third column to denote that a government service

executive utilizes 91.50 percent of the skills required of a procurement clerk.

To recap, this paper represents the ﬁrst known attempt to develop a full OSHC
transferability matrix based on occupational characteristics and introduces the concept of
a qualiﬁcation rate while generating a similar matrix. Using data from O*NET, this paper
generates a 900 x 900 matrix that makes pair-wise comparisons of the KSAs used within
each vocation to estimate the two occupational indices. While an imperfect estimator of
occupational comparability, it nevertheless represents a signiﬁcant advancement in
understanding occupational movement in the labor market and (is a tool that will be

utilized in later sections to examine the effects of structural unemployment.39

DA TA AND MODEL

This investigation of post-displacement occupational outcomes utilizes data drawn from
1984-2004 Displaced Worker Surveys (DWS), a supplement to the January edition of the
Current Population Survey (CPS) in even-numbered years. Displaced workers are

identiﬁed through a question in the CPS that asks, “During the last three calendar years...

 

39

Using data on occupational characteristics to estimate transferability requires a number of caveats.

First, this effort ignores occupational barriers that would prohibit the direct transfer of KSAs between two
particular vocations. For instance, while it is estimated that 71.1 percent of the KSAs applied by
kindergarten teachers can be utilized in one’s job as a pediatrician, there are educational barriers (i.e., a
medical degree) that would prohibit occupational mobility from the former to the latter in the absence of
such schooling. Second, it is important to note that applying aggregate occupational requirements to study
individual occupational movement implicitly fails to capture potential differences of individuals’ actual
KSAs within occupations. For instance, a master electrician would stand to lose signiﬁcantly more rent to
his or her accumulated occupation-speciﬁc human capital compared to an apprentice. Finally, the
development of the transferability/qualiﬁcation matrix fails to include speciﬁc work activities or other
potential occupation-speciﬁc features that might also inﬂuence the estimates.

64

did (you) lose a job or leave one because: (your) plant or company closed or moVed,
(your) position or shift was abolished, there was insufﬁcient work, or another similar
reason?”40 If the respondent replies “yes” to that question, he or she is asked a series of
questions regarding the lost job, including some information on the reason for the
person’s displacement, as well as a series of questions regarding his or her prior job,
including his or her tenure, union status, full-time status, earnings, and so on. This data is
obtained in addition to details on the respondent’s current job, earnings, and demographic

information as acquired through the main CPS survey tool.

While the DWS represents the “industry standard” (Seitchik 1991, 51) for studying

displacement, given its overwhelming strengths, namely its large, random sampling of R
US. households weighted to represent the country’s workforce, it is important to note the
potential biases attributable to utilizing this data. First, research has suggested that the
DWS signiﬁcantly understates the number of displaced workers due to its “snapshot”
nature (Kletzer 1998) and its failure to collect information on multiple job losses (Stevens
1997). In addition, studies have suggested that retrospective biases may further understate
the number of displaced workers over the preceding ﬁve-year period by approximately
one-third, and may not be randomly distributed across groups (Evans and Leighton

1995).41 Finally, as noted by Kambourov and Manovskii (2004b), there are signiﬁcant

 

0 Prior to 1996, the DWS had a ﬁve—year recall period. For example, the 1992 DWS asked respondents if
they had been displaced at any time in 1987-1991. There is a slight discontinuity in the deﬁnition of
displaced worker in this study from 1984 to 1992 and 1994 to 2004, as earlier surveys failed to inquire
whether a person was expecting recall from his or her displaced job. Under current BLS deﬁnitions, those
individuals would not be considered “displaced.”

41
For more discussion on the retrospective biases of the DWS, see Topel (1991), Farber (1993, 1997),

Fairlie and Kletzer (1996), and Kletzer (1998), among others.

65

coding errors attributable to occupation within the Current Population Survey that may
hinder studies of occupational mobility. While the imputation of occupation by coders
exists, those observations represented less than 1 percent of the current sample and were

subsequently dropped from this analysis.

Caveats aside, given that the questions at hand focus on those displaced from blue-collar
occupations in the manufacturing industry, this study identiﬁes “blue-collar” vocations as
those in the following occupational categories: transportation and material moving;
production; installation, maintenance, and repair; construction and extraction; and
farming, ﬁshing, and forestry. Restricting the sample to those aged twenty and older,
Table 10 provides means for the resulting data drawn from the 2004 DWS. As exhibited
in the table, much of the current stereotype of displaced blue-collar manufacturing
workers is accurate, namely that they are typically predominantly male, older, less
educated, more likely to be a union member, and more concentrated in the Midwest when
compared to the entire set of displaced workers across all industries and occupations.

These results are unsurprising, as they are consistent with the ﬁndings of Kletzer (2001).

This study will utilize a two-pronged approach to examine the occupational outcomes of
displaced blue-collar manufacturing workers. First, considering that vocational outcomes
of displacement largely have been ignored in the literature, this study will begin by
providing summary trends of occupational movement over the twenty-year period in
question. To investigate the potential causal factors of the occupational placement of

those displaced from blue-collar manufacturing positions, this paper then applies a

66

TABLE 10. Comparison of Blue-Collar Manufacturing Displaced Against Other
Displaced, 2004 Displaced Worker Survey

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Displaced
Displaced Displaced Blue-collar
All Displaced Blue-Collar Manufacturing manufacturing
Male 0.554 0.788 0.634 0.658
Hige 40.82 40.30 43.17 42.55
Married 0.556 0.572 0.599 0.572
# of Children, Age<18 0.709 0.773 0.702 0.732
Education
No HS Diploma 0.111 0.211 0.144 0.207
HS Graduate, No College 0.319 0.443 0.393 0.493
Some College, No Degree 0.215 0.199 0.177 0.166
Assoc. or Voc. Degree 0.087 0.082 0.078 0.069
Bachelor’s Degree 0.198 0.056 0.170 0.058
Graduate Degree 0.070 0.009 0.037 0.006
Race
White 0.831 0.853 0.824 0.808
Black 0.097 0.101 0.101 0.128
Other 0.05] 0.046 0.064 0.064
Region
Northeast 0.217 0.191 0.252 0.229
Midwest 0.247 0.255 0.270 0.291
South 0.256 0.275 0.257 0.284
West 0.281 0.279 0.220 0.196
CiQStatus
Central City 0.240 0.210 0.208 0.197
Balance of City 0.416 0.365 0.405 0.357
Non-Metro 0.190 0.266 0.231 0.290
Other 0.154 0.160 0.154 0.156
Union Member 0.087 0.185 0.148 0.219
Citizen 0.934 0.904 0.927 0.909
Veteran 0.099 0.141 0.127 0.131
Received U1 Beneﬁts 0.494 0.541 0.651 0.663
Exhausted Ul Beneﬁts 0.219 0.239 0.313 0.322
Observations 4,774 1,566 1,162 704

 

67

 

variant of the nested logit model of occupational choice originally advanced by
Soopramanien and Johnes (2001). The choice structure demonstrated in Figure 2
represents an improvement on standard multinomial logit models (Schmidt and Strauss
1975), as it allows for non-employment options that are prevalent among displaced
workers (15.1 percent of the 2004 DWS sample) and must be considered in any study of
the employment status of displaced workers.42 The nested logit structure utilized in this
study advances a two-stage decision process; ﬁrst, the respondent “chooses” an
employment outcome (employed, not employed, not in the labor force) before selecting
an occupation conditional on choosing employment. With six occupational categories, the
nested logit model is estimated using a standard battery of demographic variables allowed
by the CPS-age, tenure, race, marital status, number of children, union membership,
citizenship status, education, region, city status, notice of layoff, and unemployment
insurance status. The model is estimated for males ’within a pooled data set extending
from 2000 to 2004 to obtain a suitable sample size while attempting to minimize across-

year differences in the labor market.43

The second avenue pursued in this piece will address the relationship between

occupation-speciﬁc human capital and vocational outcomes following displacement. In

 

42

Attempts to include non-employment as an outcome of a multinomial logit (MNL) model of

occupational choice impose an unrealistic choice structure on the respondent. Further, merely including
non-employment outcomes in a MNL structure runs the risk of violating the independence of irrelevant
alternatives (IIA) assumption necessary for the proper estimation of MNL models, which implies that
adding another alternative or changing the characteristics of a third alternative does not affect the relative
odds between two alternatives within the available set of choices. For a further discussion of the [IA
assumption, see Wooldridge (2001) or the classic “red bus, blue bus” example.

The nested logit model is estimated separately by gender given documented differences in labor market

behavior (see, among others: Oaxaca, 1973; MacPherson and Hirsch, 1995; Soopramanien and Johnes
2001). The results for females between 2000 and 2004 are given in Appendix B and are not highlighted in
the text due to the signiﬁcantly smaller sample size.

68

particular, this analysis examines two hypotheses. First, manufacturing shed 2.8 million
workers in the three-year span between 2001 and 2003, the largest such decline in the
industry since the BLS began recording industry employment in 1939 and a priori
evidence of a structural economic shift away from blue-collar manufacturing work in the
United States. Displaced workers from this sector should experience this structural shift
by a greater mismatch of skills to available jobs when compared to those displaced from
other segments of the labor force. This should produce lower re—employment rates
following displacement among displaced blue-collar manufacturing workers, as well as
lower degrees of transferability of knowledge, skills, and abilities from their previous job
upon ﬁnding employment. These hypotheses can be tested using data from the 2004
DWS (covering those displaced between 2001 and 2003) and the transferability matrix as
a means of conﬁrming and examining the effects of the structural economic shift on those
displaced from blue-collar manufacturing jobs.

FIGURE 2. Diagram of Hierarchical Decision Tree Applied by Nested Logit Model
of Occupational Choice

Not in Labor
Force

Unemployed

 

Occupational
Employed Categories

69

Corresponding to vocational outcomes, this paper will also examine the wage impact of
occupation-speciﬁc human capital within the displaced worker framework. Human
capital theory suggests that capturing more of one-’s speciﬁc human capital following
displacement should result in a higher wage. Thus, using the transferability matrix
described in this paper, one would expect that those experiencing higher degrees of
transferability of acquired KSAs from their lost job should exhibit higher wage rates,
ceteris paribus. In essence, this becomes a test of the theoretical underpinnings of
occupation-speciﬁc human capital and the transferability matrix employed in this paper.
Given a change in occupational coding structure and differing economic environments

across years, this examination will also be limited to data from the 2004 DWS.

RES UL TS

Table 11 provides a summary of occupational outcomes for displaced blue-collar
manufacturing workers from 1984 to 2004 who were re—employed at the time of the
survey, offering a number of interesting trends despite the use of broad occupational

44 Most signiﬁcantly, rates of re-employment within the blue-collar

classiﬁcations.
segment of the occupational spectrum have declined; while roughly 67-70 percent of
displaced blue-collar manufacturing workers were re-employed in blue-collar jobs from

1984 to 2000, that percentage had fallen to 64 percent in the 2002 and 2004 DWS. Much

lower rates of re-employment in blue-collar manufacturing jobs have fueled much of this

 

The 2004 CPS (and DWS) utilizes a different occupational coding structure than previous years.

Because of this, Table 11 utilizes broad vocational categorization in order to provide cross-year consistency
in occupational deﬁnition.

70

TABLE 11. Re-Employment Occupation of Displaced Blue-Collar Manufacturing
Workers, 1984-2004 DWS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rate

 

 

 

 

 

 

 

 

 

 

 

 

1984 I986 [988 1990 1992 1994 1996 1998 2000 2002 2004
Executive,
Administrative, 3.0% 3.2% 4.1% 3.2% 2.4% 2.9% 2.9% 2.7% 3.3% 2.8% 3.2%
and Managerial
gmf‘isswnal 2.4% 1.4% 2.4% 2.4% 1.6% 3.1% 2.9% 2.4% 1.7% 3.3% 4.8%
pec1alty
Tecm‘c‘ansand 1.5% 1.8% 0.5% 1.7% 1.4% 2.2% 0.9% 1.4% 1.7% 2.0% 1.4%
Related Support
Sales 5.4% 5.9% 5.3% 5.7% 6.8% 6.0% 5.0% 5.8% 2.7% 9.2% 5.1%
Administrative
Support Including 6.6% 6.6% 6.7% 5.4% 6.2% 8.0% 6.7% 6.9% 9.0% 4.9% 9.1%
Clerical
Protective Service 1.6% 1.4% 2.2% 2.3% 1.6% 1.1% 0.9% 1.0% 1.0% 2.0% 1.8%
semcerExcep‘ 11.2% 12.8% 9.8% 9.0% 13.3% 8.0% 10.2% 10.3% 9.7% 11.7% 10.3%
Protectlve
glue'C°l'a’ 68.3% 67.0% 69.2% 70.4% 66.7% 68.6% 70.5% 69.5% 70.9% 64.1% 64.2%
ccupatrons
Blue-Collar 7
Occupations- 39.3% 41.6% 45.3% 47.5% 46.3% 49.8% 48.1% 46.9% 45.2% 35.0% 38.8%
Manufacturing
Blue-Collar
Sgg‘jpamns' 29.0% 25.3% 23.9% 22.9% 20.4% 18.8% 22.4% 22.6% 25.8% 29.1% 25.5%
Manufacturing
Obsewat'ons 1,272 1,235 837 756 795 448 343 292 299 392 436
(Employed)
Obsewat‘ons 892 727 406 356 582 295 177 125 119 335 294
(Unemployed)
Obsewa‘mns 2,164 1,962 1,243 1,112 1,377 743 520 417 418 727 730
(Total)
Re‘Emp'Wmem 58.8% 63.0% 67.3% 68.0% 57.7% 60.3% 66.0% 70.0% 71.5% 53.9% 59.7%

 

71

 

trend, as just 38.8 percent of those displaced from such jobs between 2001 and 2003 were
re-employed in blue-collar manufacturing positions. While such re-employment and
industry absorption rates follow economy-wide and industry expansion/contraction

cycles, respectively, the ﬁgures post-2000 represent signiﬁcant downturns.

While the results of Table 11 may offer initial evidence of a quickening of structural
change in the economy away from blue-collar employment, the volatility of other sectors
makes it difﬁcult to necessarily pinpoint growth areas. Sales and service occupations, in
particular, have not demonstrated increased absorption rates of displaced blue-collar
manufacturing workers over the last two decades, standing contrary to the social myth
regarding such individuals landing as burger ﬂippers or stock handlers at big-box
retailers. On the contrary, administrative support and professional specialty occupations
have seen general increases in their absorption rates of displaced workers from this
sector. These results may be marginally indicative of retraining success, the expansion of

particular segments of the labor market, or a combination of other factors.

The causal inﬂuences of this occupational choice are examined in Table 12 through the
estimation of the nested logit model. While the coefﬁcients are not directly interpretable
and thus take on the role of an index of dissimilarity (Greene 2002), the results indicate

that having a college degree represents a statistically signiﬁcant feature of those entering

72

TABLE 12. Nested Logit Results, Displaced Blue-Collar Manufacturing Workers,

Males, 2000-2004 DWS

 

Participation Choice

Occupational Choice

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Brandy (Twig)
Profession Blue-
Unem- NILF -al, Service Sales Clerical Collar,
ployed Manage- Non-Man.
ment
Age
20-24 -0.149 -0.481 -1 .061 * 0.090 -0.627 -0.097 0.479
(0.340) (0.447) (0.572) (0.470) (0.717) (0.683) (0.348)
25-34 -0.065 -0.653 ** -0.840** -0.625 -0. 145 -0.207 -0.091
(0.223) (0.327) (0.354) £392) (0.444) (0.539 (0.245)
35-44 Base Base Base Base Base Base Base
4554 0.363 -0.061 -0.881** 0.249 -0.451 —0. 185 -0.096
(0.237) (0.309) (0.400) (0.399) (0.514) (0.549) (0.276)
55=64 0.490 0.823" -0.485 0.317 * -l.864 -0.305
(0.310) (0.359) (0.546) (0.530) (1.141L (0.380)
65 & up 0.564 2225*" -0.142 0.710 * 0.352 0.723
(0.778) (0.690) (1.306) (1.289) ‘ (1.392) (0.906)
Job Tenure -0.009 0.013 -0.014 -0.040* 0.019 0.033 0.010
(0.01 1) (0.012) (0.022) (0.024) (0.025) (0.023) (0.012)
Race
Black 0.160 0.154 -0.531 -0.301 -0.768 0.028 -0.224
(0.292) (0.359) (0.554) (0.555) (0.806) (0.833) (0.340)
Other 0.148 0.654 -0.316 -0.4 l 7 * 0.919 -1.370**
(0.359 (0.418) (0.577) (0.611) ' (0.619) (0.582)
Married -0.3 83* -0.521** -0.410 -0.750*“‘ -0.386 -0.052 -0.277
(0.198) (0.250) 10.332) (0.337) (0.433) (0.475) (0.233)
Children <18 -0.060 -0. 193 -0.044 0.024 -0.374* -0.506** 0.000
(0.085) (0.126) (0.140) (0.139) (0.222) (0.257) (0.094)
Union, Old Job 0.133 0.533M -0.804** -0.044 -0.841 -0.394 -0.054
(0.206) (0.251) (0.389) (0.377) (0.519) (0.542) (0.232)
US Citizen 0.016 -0.012 -0.289 -0.455 -1 .250" -0.863* -0.229
(0.272) (0.341) (0.384) (0.387) (0.488) (0.510) (0.288)
Education
Less than HS 0.502“ -0.190 -0.228 . 0.598 -0.521 -0.203 0.572"
(0.243 (0.325) (0.495) (0.377) (0.673) (0.678) (0.261)
HS Diploma Base Base Base Base Base Base Base
Some 0.160 0.124 0.829“ 0.428 0.164 0.597 0.168
College (0.230) (0.286) (0.363) (0.366) (0.522) (0.492) (0.257)
Vocational or «0.333 -0.088 1600*" -0.057 1.243" 0.300 0.71 I"
Assoc. Deg. (0.399) (0.448) (0.448) (0.675) (0.572) (0.825) (0.356)
College Deg. 0.553 0.148 2.027*** 0993* 1235* 0.604 0.142
(0.388) (0.503) (0.485) @568) (0.639) (0.780) (0.477)
Region
Northeast -0.254 -0.757** —0.619 -0.702 -0.353 -I.532** -0.048
(0.252) (0.325) (0.404) (0.456) (0.584) (0.646) (0.298)
Midwest —0.069 -0.640** -0.782* 0.182 0.393 -0.683 0.180
(0.238) (0.303) (0.402) (0.374) (0.507) (0.505) (0.281)
South -0.202 -0.430 -0.357 -0.162 0.061 -1.448** 0.222
(0.252) (0.310) (0.383) (0.416) (0.547) (0.618) (0.290)

 

 

 

 

 

 

 

 

73

 

TABLE 12. Nested Logit Results, Displaced Blue-Collar Manufacturing Workers,
Males, 2000-2004 DWS (cont.)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

West Base Base Base Base Base Base Base
City Status
In City -0.189 -0.373 -0.369 -0. 124 -0.279 -O.899 -0.181
(0.240) (0.301) (0.415) (0.393) (0.518) (0.600) (0.277)
Balance of -0.137 —0.659** -0.112 0.140 0.046 -0.234 -0.082
City (0.201) (0.260) (0.346) (0.327) (0.434) (0.463) (0.227)
Non-Metro Base Base Base Base Base Base Base
Other -0.367 -0.414 0.548 -0.863 -0.126 0173 -0.088
(0.262) (0.319) (0.401) (0.547) (0.592) (0.640) @286)
Notice of -0.175 -0.245 0.194 -0.437 -0.079 -0.220 0.125
Layoff (0.172) (0.221) (0.281) @302) (0.380) (0.413) (0.193)
U1 Exhaustee 0.452" 0451* 0.304 0.005 0.414 0.364 0.354
(0.199) (0.244) (0.334) (0.366) (0.442) (0.489) (0.230)
Inclusive Em- 0.6707 Unem— 0.5000 NILF 0.5000
Values ployed ployed
Observations l , 1’00
Log-likelihood -1,880.3354
LR ch12(169) 814.1007
Prob > chi2 0.0000

 

 

Note: Standard error in parentheses.

professional/management and sales occupations following displacement. Beyond the role
of education, the model also serves to conﬁrm a number of common beliefs, namely that
older males are considerably more likely to drop out of the labor force and married men
exhibit signiﬁcantly higher rates of not only staying in the labor market but also ﬁnding a

job.45

 

45

A comparison of multinomial logit and nested logit procedures encompassing the eight possible

outcomes (six occupations, unemployment, nonparticipation) yields statistically signiﬁcant evidence
verifying the latter’s dominance in modeling occupational choice. As noted in Soopramanien and Johnes

(2001), the test statistic comparing the two models, -2(LRm—LRn), is distributed as a x-square with the
degrees of freedom equaling the number of restrictions that are imposed in order to reduce the nested logit
to a simple multinomial logit (results of the multinomial logit model are available from the author upon
request). Therefore, with the current models yielding a test statistic of 18.0329, the results of this study
indicate the preference for utilizing nested logit procedures when working with occupational choice of

74

While results of the preceding occupational analyses have only indirectly spoken to
structural effects in the economy, a more direct approach recognizes that within a
structural shift in the economy, those displaced from occupations deemed most
vulnerable would ﬁnd fewer openings in comparable professions. Those displaced from
these segments of the labor market would be expected to be forced to accept positions
utilizing less of their acquired occupation-speciﬁc human capital, given its relative
obsolescence when applied to expanding portions of the labor market. When translated
through the lens of the transferability matrix developed earlier in this study, the supposed
effects of economic restructuring on the hypothesized group in question-blue-collar
manufacturing workers-should be demonstrated by those displaced from this segment
applying lower percentages of the KSAs acquired in their pre-displacement vocation to
their post-displacement occupation. Therefore, a direct examination of structural effects
on blue-collar manufacturing workers leads the analysis to apply a condensed, 46 500 x
500 version of the transferability matrix (to match the DWS occupational coding

structure) to data from the 2004 DWS that captured a time of great upheaval in the

 

displaced workers. Thus, while this study employs only a rudimentary set of explanatory variables, which,
outside of age and education, have little theoretical implications for the labor market outcomes of displaced
workers, the results of this study indicate that occupational choice models incorporating more advanced
data (such as retraining initiatives) would be well advised to apply nested logit to more accurately model
the decision-making process.

By far, the most prevalent inconsistency between the occupational structure of the DWS/Census and the
SOC was that of multiple SOC codes matching onto one DWS/Census vocational code. To alleviate this

problem, the Yij are averaged over all matching SOC codes to generate the appropriate DWS/Census 79.

While this technique has been utilized by those working with O*NET data (Hirsch 2005), one must be
wary of the effects of averaging SOC classiﬁcations that have grossly unequal number of job holders. For
instance, the matrix scores for the DWS/Census occupation code of “Chief Executives (0010)” represents
the average of matrix scores for the SOC occupation codes for “Government Service Executives (11-
1011.01)” and “Private Sector Executives (1 1-1011.02).” While the number of individuals holding such
jobs are not likely to be equal across the economy, methods that utilize averaging are inherently producing
biased results. However, the absence of detailed statistics on the number of job holders by SOC/O‘NET
occupational categories mandates the continued utilization of cross-SOC averages in order to merge
O*NET data with Census survey data.

75

TABLE 13. Transferability of KSAs, Displaced Blue-Collar Manufacturing Versus .

All Other Displaced, 2004 DWS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Pct. 71'1". 71.1".

Obs. Employed Transfer Qualified
Displaced, Overall
Blue-CollgaryOccupations, Manufacturing 345 60.80% 68.01% 71.19%
All Other Displaced 2481 67.99% 75.36% 76.78%
t-Statistic 5.852 4.637 '
Displaced, By Sector
Professional/Technical Occupations 893 71.18% 74.28% 82.45%
Service Occupations 283 61.12% 75.11% 72.46%
Sales Occupations 294 68.39% 74.02% 73.52%
Clerical Occupations 435 64.81% 75.31% 70.93%
Blue-Collar Occupations, Manufacturing 345 60.80% 68.01% 71.19%
Blue-Collar Occupations, Non-Manufacturing 574 69.54% 77.95% 76.26%

 

 

 

 

 

 

manufacturing industry, namely those displaced from that sector between 2001 and

2003.47

The results of applying the transferability matrix to data from the 2004 DWS are posted
in Table 13, and are distinctly illustrative of the structural obsolescence of blue-collar
occupations within the manufacturing industry. In addition to lower re-employment rates,
those displaced from this sector exhibit a signiﬁcantly lower percentage of transferability
(68.01 percent) of their knowledge, skills, and abilities to their post-displacement
occupation than those displaced from the balance of the labor market (75.36 percent).
Comparing sectors in the bottom half of Table 13 further reinforces this conclusion, as all

other categories of workers exhibit transfer rates over 74 percent, far exceeding the rates

 

The condensing of SOC/O*NET occupational classiﬁcations to DWS/Census codes also resulted in a

number of observations in the sample not matching into the matrix, given the utilization of “all other”
occupational categories in the sample. For instance, a worker’s pre- or post-displacement occupation may
have been categorized by the DWS as “Engineers, all other.” However, the SOC/O*NET system does not
have comparable “all other” categories that yield a suitable‘match; as such, a small percentage of the
sample do not ﬁnd an appropriate home in the transferability matrix given the incongruent nature of their
DWS/Census occupation when compared to possible SOC/O*NET vocational codes.

76

exhibited by those from blue-collar manufacturing jobs. Strikingly, blue-collar workers
displaced from nonmanufacturing industries display transfer rates of 78 percent-nearly 10
percentage points higher than those displaced from manufacturing-further cementing the

recognition regarding the relative obsolescence of such positions in the labor market.

Turning to “qualiﬁcation” rates in Table 13, those displaced from blue—collar occupations
within the manufacturing industry exhibit lower qualiﬁcation rates (71.19 percent) in
their post-displacement profession compared to their displaced brethren (76.78 percent).
This trend generally holds when examining different sectors of the labor market as well,
except for the marginally lower rate of clericals (70.93 percent). Nevertheless, conﬂicting
reasons may be responsible for higher qualiﬁcation rates: a person will exhibit higher
rates for ﬁnding work in a comparable (or the same) occupation as well as in one in
which they are overqualiﬁed. In other words, if a plant foreman has to take a job as a
manual laborer, he or she will display a high degree of “qualiﬁcation.” However, this
would likely represent a negative post-displacement outcome, and the resulting loss of
human capital upon this “downward” movement would be captured only through the use

of transferability rates.

Limiting the focus to transferability rates, given its importance in establishing labor
market effects of economic restructuring, Table 14 displays the estimation of a regression
model using transferability rates as the dependent variable. The results indicate that, even
after factoring out heterogeneity (e.g., education, age) between occupational groups,

individuals displaced from blue-collar manufacturing jobs experience a greater loss of

77

occupation-speciﬁc human capital than do their displaced counterparts. Beyond the role
of occupational groupings, Table 14 reveals a number of other interesting trends. First,
transferability is lower for younger workers, as they may not have had the necessary
experience to develop their human capital both across the labor market and within a given
occupation. While the effect of a college education has a predictably positive effect on
one’s ability to ﬁnd a high-transferable occupation following displacement, it should also
be noted that displaced union members tend to exhibit higher transferability rates, ceteris
paribus, that may speak to differences in individual skill levels. Further, those who
exhaust their unemployment insurance (U1) beneﬁts tend to experience signiﬁcantly
lower rates of transferability, potentially due to the failure to ﬁnd a comparable job
during the UI period and the subsequent scramble to ﬁnd any job following the

exhaustion of beneﬁts.48

To establish transferability as a key determinant of post-displacement outcomes, Table 15
provides the results of a regression equation analyzing the effect of transferability rates
on weekly earnings of those re-employed for at least twenty hours per week. Controlling
for a multitude of variables, the use of ordinary least squares (OLS) in the ﬁrst model
demonstrates a positive effect of transferability on earnings, with the effect signiﬁcant in
greater than a two-sided, one-percent signiﬁcance test. The coefﬁcient suggests that as
the transfer rate goes up by 10 percentage points, earnings will be expected to increase by

6 percent.

 

48

A speciﬁcation of the regression equation separating all occupations into manufacturing and
nonmanufacturing groups found comparable results.

78

TABLE 14. Regression Analysis on Transferability of KSAs from Displaced to Re-

Employment Occupation, 2004 DWS

 

[If Transfer

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Model I Model 2
Displaced Occupation
Professional/Technical Occupations -0.032*
(0.017)
Service Occupations Base
Sales Occupations -0.013
(0.019)
Clerical Occupations 0.008
(0.017)
Blue-Collar Occupations, Manufacturing -0.065 * * *
(0.019)
Blue—Collar Occupations, Non—Manufacturing 0.029“
(0.017)
Age
20-24 -0.046*** 0048*"
(0.017) (0.017)
25-34 -0.013 -0.015
(0.011) (0.011)
3544 Base Base
45-54 -0.022* -0.023*
(0.012) (0.012)
55-64 ‘ 0.015 0.012
(0.016) (0.016)
65 and up 0.032 0.035
(0.043) (0.042)
Education
No High School Diploma 0.023 0.022
(0.016) (0.016)
High School Diploma, No College Base Base
Some College, No Degree 0.015 0.014
(0.012) (0.012)
Vocational or Associate’s Degree _ 0.005 0.008
(0.015) (0.015)
Bachelor’s Degree or More 0045*“ 0.059***
(0.011) (0.013)
Demographics
Male 0.0001 -0.004
(0.009) (0.009)
Black -0.013 -0.011
(0.016) (0.016)
Other Non-White Race 0.027 0.030"'
(0.018) (0.018)
US. Citizen -0.044** -0.043**
(0.018) (0.017)
Married 0003 -0.003
(0.010) (0.010)

 

79

 

TABLE 14. Regression Analysis on Transferability of KSAs from Displaced to Re-

Employment Occupation, 2004 DWS (cont.)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Number of Children (< age 18) -0.003 -0.003
(0.004) (0.004)
Region
Northeast 0042*“ 0.045***
(0.012) (0.012)
Midwest 0.005 0.010
(0.012) (0.012)
South 0.029M 0032*”
(0.012) (0.012)
West Base Base
City Status
In City 0.011 0.010
(0.013) (0.013)
Balance of City 0.015 0.015
(0.012) (0.012)
Other City Status -0.012 -0.013
(0.014) (0.014)
Non-Metro Base Base
Job Characteristics
Union Member at Old Job 0066*" 0068*"
(0.015) (0.016)
Tenure at Old Job 0002’” -0.001*
(0.001) (0.001)
Received Notice of Layoff 0.008 0.012
(0.009) (0.009)
Exhausted U1 Beneﬁts -0.080*** -0.074***
(0.011) @011)
Constant 0.770 0.772
(0.025) (0.027)
Observations 2,698 2,698
R-Stprared 0.0497 0.0660
F-Statistic on Overall Model 5.58 6.28
(p<0.0001) (p<0.0001)
F-Statistic on Occupational Variables 9.34
(p<0.0001)

 

Notes: Standard errors in parentheses. Significance levels noted as follows: *** - I percent, ** - 5 percent,

* - 10 percent.

80

 

TABLE 15. Regression Analysis on 1n(Weekly Earnings), Among Those Re-
Employed and Working at Least 20 Hours/Week, 2004 DWS

 

 

 

 

 

 

 

 

 

 

ln(weekly earnings)
OLS ZSLS
Transfer Rate of KSA" 100 0.006*** 0.048*
(0.0008) (0.027)
Observations 2,405 2,405
R-sguared 0. 1507 *
F(31, 2373) 13.59 5.64
(p<0.0001) (p<0.0001)

 

 

Notes: Other variables include: Male, age (6 categories), education (5), race (3), citizen, union member at
lost job, marital status, tenure at lost job, number of children under eighteen, region (4), city status (4),
received notice of layoﬂf exhausted UI beneﬁts, origin occupation (6). Second-stage equation identified by
the state’s WIA spending per displaced worker in 2001 and an indicator denoting whether the individual
lost his or her job via plant closure versus other reasons (slack work, shift abolished). Standard errors in
parentheses. Significance levels noted as follows: *** - I percent, ** - 5 percent, * - 10 percent.

The results of the OLS model are biased, however, by the inherent simultaneity between
transferability of OSHC and post-displacement earnings. To resolve this, a two-stage
least squares (ZSLS) approach is utilized. In the ﬁrst stage, transferability is predicted by
the basic set of explanatory variables used previously along with two others: (a) the
amount of state spending per displaced worker through the Workforce Investment Act
during the year 2001 as the IV, and (b) an indicator variable denoting whether the
individual lost his or her job because of a plant closing as opposed to some other reason
(e.g., shift abolished).49 While the use of instrumental variables calls into question the

magnitude of the effect of transferability on earnings, the results nevertheless conﬁrm the

 

State-level spending through the WIA is used by local agencies to match the characteristics of displaced

workers with the needs of the local labor market, with emphasis on “a quick return to employment”-and not
necessarily higher post-displacement wages. Further, the reason behind an individual’s displacement speaks
to the availability of different types of jobs within the local labor market, as plant closings may indicate a
more downtrodden environment for an occupation or industry. Data drawn from the Government
Accounting Ofﬁce report, GAO-02-274: “Workforce Investment Act: Better Guidance and Revised
Funding Formula Would Enhance Dislocated Worker Program,” February 2002,
http://www.gao.gov/new.items/d02274.pdf.

81

previous ﬁndings that higher transferability of OSHC upon re-employment has a positive,

statistically signiﬁcant effect on post-displacement earnings.

DISCUSSION

While the manufacturing sector is at its lowest level of employment since 1950, evidence
of structural unemployment affecting blue-collar, manufacturing workers has been
largely circumstantial. As a part of a broader analysis of post-displacement occupational
outcomes within this group, this paper provides evidence that directly conﬁrms the
“structural victim” status of those displaced from blue-collar, manufacturing jobs.
Signiﬁcantly lower rates of both re-employment and transferability of occupation-
speciﬁc human capital afﬂict those displaced from this sector. However, the occupational
analyses presented here demonstrate that typecasting laid-off workers from this sector as
future low-level service or sales associates is largely unrealistic. In contrast, twenty years
of data reveals marginally rising absorption rates of displaced blue-collar manufacturing
workers by professional specialty and administrative support occupation groups while
absorption rates by sales and service occupations have remained ﬂat. Further, given the
importance of the transferability of OSHC in establishing post-displacement earnings,
future research should go beyond its typical focus on wages to examine why certain
groups fare better upon re-employment and how public policy can best aid those who are

the casualties of structural shifts in the economy.

82

Beyond these results, the greater long-term contribution of this paper may be the
development of a 900 x 900 occupation transferability matrix of KSAs based on
occupational descriptions provided by O*NET. This matrix represents a signiﬁcant
improvement from Shaw’s (1984) original attempt to estimate cross-occupation
transferability of human capital using occupational switching proxies. With a growing
literature awakening to the impact of occupation-speciﬁc human capital, the development
of this matrix may serve as a valuable statistical lens through which to examine a wide
range of labor market questions, including the identiﬁcation of comparable-and
expanding-occupations for displaced workers that may lead to greater percentages of

retraining success.

83

Chapter 3

HIGH SCHOOL EMPLOYMENT AND THE ACCUMULATION
OF OCCUPATION-SPECIFIC HUMAN CAPITAL

INTRODUCTION

While labor market participation rates among high school students in the United States
dominates that of most other industrialized countries (Alsaker and F lammer 1999; F uligni
and Stevenson 1995), research efforts have yet to completely untangle the positives and
negatives associated with in-school employment amongst secondary school students. On
one hand, working during high school is posited to negatively affect academic
performance and attainment, a stance generally borne out in the literature (Marsh and
Kleitman 2005; Rothstein 2007). On the other hand, nearly three decades of research
have consistently demonstrated that high school employment has a positive effect on
post-school economic outcomes such as earnings, employment, and occupational

attainment.

Despite the pervasiveness of the research demonstrating a positive relationship between
in-school work and future economic gains, the literature is silent in regards to why this
association exists. While studies establishing this positive relationship hypothesize that it
is attributable to the role of high school employment as a positive net inﬂuence on human

capital (DeSimone 2006; Light 1999; Ruhm 1997), such a statement is made without

84

empirical evidence and lacks speciﬁcity given the complex nature of human capital (as
discussed in Chapter 1). In particular, does high school employment develop “marketable
skills” within the student that represent speciﬁc human capital? Or could it be that in-
school employment builds character and socializes one to the workplace, thus providing
general human capital? Or could it be that the positive association between high school
employment and future economic outcomes is not attributable to human capital

differences at all?

Given the empirical silence regarding these questions, this paper makes the ﬁrst known
foray to investigate the potential rationale behind the positive relationship between high
school employment and post-school economic outcomes. To such ends, this paper
addresses the ﬁrst possibility above, namely that in-school employment instills
marketable skills—deﬁned in this paper to be occupation-speciﬁc human capital (OSHC)—
that represent the causal inﬂuence explaining why high school employment is suggested
to yield post-school economic gains. By the tenets of the human capital model, it is
expected that greater applicability of one’s learned occupational skills from high school
employment to one’s adult (post-school) vocation should result in higher productivity,
and thus higher wages. To test this hypothesis, this study utilizes eight rounds of data
from the 1997 National Longitudinal Survey of Youth (NLSY97) and multiple measures
of OSHC transferability, including the OSHC matrix provided in Chapter 2. If it is
conﬁrmed that higher transferability of skills from one’s senior-year vocation to one’s

adult profession yields positive wage effects, then this would stand as at least part of the

85

reason for the consistent ﬁnding of a positive relationship between in-school work and

post-school economic outcomes.
L1 TERA TURE RE VIE W

The research agenda on high school employment has historically had a heavy focus on
the potentially deleterious effect of in-school work on academic performance, as
measured by grades, test scores, or school completion rates. No consensus currently
exists about whether student employment improves or impairs high school academic
performance, however the evidence seems to suggest that any effects—positive or
negative—are minimal at low and intermediate levels of work intensity. Meanwhile,
there is overwhelming evidence that heavy work commitments have strong negative
effects on academic outcomes. While this has generally been viewed as a threshold effect
(D’Amico 1984; DeSimone 2006; McNeal 1995, 1997; Schill, McCartin, and Meyer
1985), others have suggested that the negative effect of work hours on academic

performance is linear (Marsh and Kleitman 2005).

Contrary to the above, research has consistently demonstrated that high school
employment has a positive association with future economic outcomes. D’Amico (1984)
and Marsh (1991) demonstrated that holding a job while in high school was associated
with lower post-school unemployment rates. Stephenson (1981), Meyer and Wise (1982),
and Mortimer and Finch (1986) showed that in-school employment was related to higher

future incomes. Mortimer and Finch (1986) illustrated that high school job holding was

86

correlated with higher post-school occupational attainment. Despite their consistent
conclusions, these studies suffered from two primary ﬂaws: each only examined the
immediate period following school completion and they treated in-school employment as
exogenous; by doing so, the studies ignored the selection process of individuals into the

workforce.

In an attempt to rectify both issues, Ruhm (1997) utilized data from the NLSY79 to
demonstrate that hours worked during one’s senior year was positively correlated with
higher earnings, fringe beneﬁts and occupational status six to nine years following school
completion. A breakthrough work for its application of sample-selection corrections as a
means of identifying and resolving the inherent endogeneity issues, Ruhm estimated that
working twenty hours per week as a senior is associated with a 9 percent boost in hourly
wages and an 11 percent increase in hourly compensation as an adult. Light (1999)
addressed the underlying question of whether job-holding in high school served as a
“value added” experience, or whether high school employment raised future labor market
productivity beyond any losses attributable to lowered academic performance. Light
analyzed a set of male terminal high school graduates to examine the inﬂuence of high
school employment, high school curriculum, and future work experience on earnings for
the ﬁrst nine years following school completion. Finding small, but signiﬁcant, positive
gains to high school work intensity on future earnings, Light concluded “high school
employment has a positive, skill-enhancing effect on wages for the ﬁrst six years after

graduation,” (p. 305).

87

Light’s conclusion, similarly suggested by Ruhm (1997) and DeSimone (2006), speaks to
the common belief regarding the source of the positive relationship between high school
employment and future economic outcomes: in—school work generates a net increase in
human capital. If this is indeed true, such a statement linking in-school work and human
capital development lacks speciﬁcity. In particular, while the phrase “human capital” is
typically used as a catch-all for skill and knowledge development, the construct itself is
quite complex (as discussed in Chapter 1). As a result, understanding how high school
employment affects human capital stocks is critical to explain why in-school work results
in future positive economic outcomes. Is the positive earnings impact of in-school work
attributable to increases in occupation-speciﬁc human capital under the rubric of
“marketable skills”? Or could it be that the positive relationship is due to increases in
general human capital, such as building character and socialization to the workplace?
These questions are vital to understand the impact of high school employment on future

economic outcomes.

This paper examines this ﬁrst question, namely whether high school employment
develops occupation-speciﬁc human capital that may be the underlying causal factor
relating in-school work and adult economic gains. Shaw (1984) deﬁned occupation-
speciﬁc human capital as “the accumulation of skills an individual acquires to perform
work within an ‘occupation’” Q). 320).50 Distinguishing occupation-speciﬁc from ﬁrm-

speciﬁc human capital, Shaw noted that not all of one’s occupation-speciﬁc human

 

Shaw (1984) deﬁnes an “occupation” as “a homogeneous skill classiﬁcation within which individuals

are perfect substitutes in demand and/or have inﬁnite cross elasticities of substitution in supply. Cabinet-
makers and engineers, for example, have occupation-speciﬁc skills,” (p. 320).

88

capital is lost upon a vocational switch. Instead, individuals are able to “transfer” certain
proportions of their accrued OSHC from one vocation to another based on the
commonality of skill sets required for the two occupations. For example, an accountant
would be able to transfer a large percentage of their accrued OSHC into their new role as
i an actuary; in contrast, an accountant would experience low transferability rates if they
became an electrician. The foundational work of a small but growing literature on
occupation-speciﬁc human capital, Shaw demonstrated that accrued occupation—speciﬁc
human capital dominated general experience in determining wages; later studies showed
OSHC to also dominate industry- and ﬁrm-speciﬁc human capital in terms of predicting
earnings (Kambourov and Manovskii 2004, 2005). Given the signiﬁcant inﬂuence of
accrued occupation-speciﬁc human capital on earnings, the development of OSHC
through high school employment could serve as a sizeable factor linking in-school work
and adult economic gains. While Stern and Nakata (1989) demonstrated that different in-
school jobs utilize varying types of skills, one would nonetheless expect that those who
were able to transfer more of their learned skills from their high school vocation to that of

their adult profession would be more productive, and thus earn more, all else equal.
DA TA AND MODEL

To address the question of whether the positive relationship between high school
employment and higher future earnings is attributable to the accumulation of occupation-
speciﬁc human capital, this paper utilizes data from the 1997 National Longitudinal

Survey of Youth. This data set tracks 8,984 individuals who were aged 12-16 in 1997,

89

following up with yearly questionnaires surrounding each respondent’s income, labor
market status, educational attainment, and a variety of other characteristics. The
successor to the still-active NLSY79, eight rounds of data from the NLSY97 were
publicly available at the time of this paper, spanning 1997-2004. While having only eight
years of data limits the time horizon to study adult outcomes compared to research
featuring the NLSY79, this is counterbalanced by the fact that (a) many prior studies
have used short time horizons to study post-schooling outcomes, and (b) the NLSY97

provides signiﬁcantly more timely data than the NLSY79.

In order to compute the effect of accrued occupation-speciﬁc human capital on adult
outcomes, it is necessary to ﬁrst obtain information about a person’s occupation while in
high school and their vocation as an adult, and then demonstrate how the transferability
of OSHC between professions relates to their post-schooling level of earnings. In other
words, the basic earnings model (below) applies, with “OSHC” representing an estimate
of the transferability of accrued human capital from one’s high school vocation into one’s
occupation two years following school Completion. The “ﬁrm” variable denotes

employment with the same company at both time periods and is included in order to tease

out potential tenure, or ﬁrm-speciﬁc, effects (Note that Xi represents a standard vector of

social and demographic variables):51

1n(wagei) = 130 + l31Xi + BzOSHCi + B3ﬁrrni + 8i

 

51

Wage is the CPI-inﬂated wage for the year representing the second following high school completion.

90

With a positive coefﬁcient on OSHC conﬁrming the development of marketable skills
through in-school work, the composition of this variable requires some explanation. First,
it is necessary that an individual divulge an occupation for the two time periods utilized
in this study: four months prior to, and two years following, high school completion.
Considering that most policy and research analyses of in-school employment focus on the
effects of non-seasonal work, the use of the former date captures an individual’s senior-
year occupation in February of most traditional academic calendars. Further, the use of
senior-year employment follows from Ruhm (1997), who demonstrated its dominance

compared to sophomore-year and junior-year job-holding in predicting future earnings.

This paper employs three measures of the transferability levels of occupation-speciﬁc
human capital from their in-school to post-school occupations. The ﬁrst method groups
each vocation into one of 14 categories aligned with the 2002 Standard Occupational

52 Thus, to gauge whether an individual can apply

Classiﬁcation (SOC) system.
occupation-speciﬁc human capital from their high school vocation to their adult
occupation, this method of calculating OSHC is an indicator variable equaling one if both
the adolescent and adult occupations fall into the same category. For example, this
indicator would equal one if an adult sales manager had worked as a retail sales clerk in

high school; it would similarly equal zero if that sales manager had been employed as a

carpenter’s assistant.

 

52

Most occupational categories are based upon the ﬁrst two digits of the relevant SOC codes. However,

given the lack of observations in the “white-collar” occupational categories, these have all been collapsed
into a singular “management and professional” classiﬁcation for use in this study.

91

While the above method is empirically simple, it is not altogether an accurate
representation of the transferability of occupation-speciﬁc human capital from one’s high
school vocation to their adult occupation. For example, due to the structure of the SOC,
the above method would afﬁx a value of zero to the transferability of OSHC between
retail sales clerks and, say, marketing managers despite the fact that acquired skills of
persuasion and customer service would be a beneﬁt across professions—and thus non-
zero. Therefore, to gauge more accurate predictions of cross-occupation transferability of
OSHC, this study turns to the transferability matrix advanced in Chapter 2. This matrix
estimates the transferability of occupation-speciﬁc human capital from occupation i to
occupation j using data from O*NET, the US. Department of Labor’s database that
provides rankings of 900 distinct occupations across 120 knowledge, skill, and ability

categories. Applied to the 508 detailed occupational codes employed by the NLSY97,53

the matrix provides two measures that will be utilized in this work: (a) Isa, the proportion

of OSHC transferred from one’s student vocation to their adult occupation, and (b) (Isa:

the proportion to which the acquired OSHC in one’s student vocation makes an
individual qualiﬁed to work in the adult profession. As described more thoroughly in

Onniston (2006), these proportions are estimated by calculating the ratio of the shared

 

In regards to the incongruence between the 900 O*NET occupations the 508 utilized by the NSLY97,

occupations are mapped one-to-one when available. Where more than one O*NET occupation is assigned
to a relevant NSLY97 occupation, mean values of the O*NET scores are used. While imperfect, this
represents the best available means of matching occupations, as demonstrated by Hirsch (2005) and
Ormiston (2006). .

92

knowledge, skill, and abilities used within both the student and adult vocations to the

total amount of KSAs utilized within one of the occupations alone.54

To understand these deﬁnitions, consider the case of a high school student who works as
a retail salesperson while enrolled, and later becomes a sales engineer. As constructed,
the transferability matrix suggests that 96.17 percent of the knowledge, skills, and

abilities utilized on the job as a retail salesperson could be applied to one’s role as a sales

engineer (i.e., tsa = 0.9617). On the ﬂip side, the matrix indicates that employment as a

retail salesperson would utilize only 41.45 percent of the knowledge, skills, and abilities

necessary to function as a sales engineer (i.e., qsa = 0.4145). While the former has

particular relevance to earnings (Shaw, 1984), the latter is of particular importance in

predicting occupational entry. For robustness, both measures are included in the analysis.

The sample of interest from the NLSY97 is restricted in three ways. First, respondents
must have been employed—and thus had an occupation—four months prior, and two
years following, their high school graduation. This'is central to examining occupational
comparability across time periods.55 Second, the sample is restricted to include only
males given that there are gender differences in both high school employment (Michael

and Tuma, 1984; Ruhm, 1997) and post-school job-holding characteristics (Oaxaca,

 

The denominator of Isa is the student occupation, while the denominator of (Isa is the adult occupation.

By limiting the sample to those who worked while in high school, this inherently eliminates the
endogeneity central to the works of Ruhm (1997) and Light (1999), and which plagued prior studies linking
high school employment to future economic outcomes. In other words, there are no concerns in this work
about the selection process of high school students into employment since all observations will have made
that selection.

93

1973; Macpherson and Hirsch, 1995). The ﬁnal restriction follows the lead of Light
(1999), as the sample is limited to only terminal high school graduates with no college
experience.56 This is done for two reasons. First, the restricted focus on this sub-sample
eliminates much of the underlying, unobserved heterogeneity (e.g., motivation, ability)
that is involved in occupational attainment and earnings, as well as eliminates the
conﬂating inﬂuence of college enrollment on job outcomes. Second, if high school
employment does signiﬁcantly build occupation-speciﬁc human capital responsible for
the positive relationship between in-school work and adult earnings, this outcome should
be most prevalent amongst non-college graduates given the differences in post-schooling
occupational outcomes across educational groups. For example, working as a retail
salesperson while in high school would likely have considerably more inﬂuence on the
occupational and wage outcomes for terminal high school graduates than for college
graduates given differences in the types of occupations available to each in the labor

market.

Table 16 provides summary statistics on the sample of 356 that were employed at both
points in the study and had an adult hourly wage between $2.00 and $50.00 (to eliminate
non-respondents and outliers).57 As might be expected, high school employment is
clustered into a limited number of occupational categories, especially food preparation
and serving, sales and related, and transportation and material moving vocations. The

distribution of adult vocations differs slightly, with the most notable difference coming in

 

This also excludes those who acquire a GED degree.

Note that while 356 is not an overly large sample, the availability of future rounds of data should

expand this sample considerably. It should be noted that Light (1999) worked with a sample of 685
individuals from the NLSY79.

94

the decrease in percentage of respondents employed in food preparation and serving
occupations (25.84 to 12.08 percent) compared with in-school employment. The overall
mean real hourly wage for terminal high school graduates two years following graduation
was $10.21. What is most interesting from Table 16 is that 37.9 percent of individuals
maintained a post-school occupation in the same category as the job they held four
months prior to school completion. Further, 20.2 percent of the sample was employed by
the same company for which they worked as a high school senior. Whether this is the
result of a cognizant effort by terminal high school students to ﬁnd the appropriate career
before graduation, or merely the post-school continuation of high school employment,

remains to be seen however represents a key issue to be examined in future research.

Table 17 summarizes the mean real hourly wage according to occupation and employer
status. The results demonstrate that there is no statistically signiﬁcant difference in mean
wages across any employment characteristics. For instance, those whose in-school and
post-school jobs fall in the same occupational category earn, on average, $10.26 per hour
while those with different occupations across the two time periods earn $10.19, a
difference that fails to be signiﬁcant (t = 0.02, p = 0.8952). Further, regardless of
occupation, those employed by the same ﬁrm at both periods cam, on average, $10.41; by
comparison, those who change employers cam, on average, $10.16. This difference also
fails all reasonable thresholds of statistical signiﬁcance (t = 0.15, p = 0.7006). An F-test
also fails to uncover any statistically signiﬁcant differences in the mean wages of the four
combinations of occupation and ﬁrm status (F = 0.08, p = 0.9717). While the results of

Table 17 are hindered by sample size limitations, it nevertheless fails to demonstrate any

95

TABLE 16. Summary Statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean / Percent
Real Hourly Wage, Two Years Following HS $10.21
Race/Ethnicity

White / Other 0.612

Black 0.171

Hispanic 0.216
Region (durirgg HS)

Northeast 0. 197

North Central 0.281

South 0.334

West 0.185
MSA Status (during HS)

C ity 0.247

Suburbs 0.497

Other 0.256
Other Social Variables

Married, Two Years FollowingHS 0.084

Union Member, Two Years Following HS 0.087

ASVAB Score (percentile) 29.72

Smoked Marijuana in High School 0.449

Second Language Spoken at Home 0.163
Occrpation HS: Senior Year Adult
Management and Professional 0.0197 0.0562
Entertainment and Media 0.0084 0.0112
Health Care and Syport 0.0084 0.0169
Protective Service 0.0056 0.0337
Food Prgparation and Serving 0.2584 0.1208
Cleaning and Building Service 0.0506 0.0478
Service, including Attendants and Personal Care 0.0365 0.0169
Sales and Related 0.1629 0.1124
Ofﬁce and Administrative Support 0.0758 0.0927
Farming, Fishing, and Forestry 0.0253 0.0140
Construction and Extraction 0.0815 0.1348
Installation, Maintenance, and Repair 0.0618 0.0983
Production, including Operators and Tenders 0.0337 0.0815
Transportation and Material Moving 0.1713 0.1629
OSHC

Same Occupational Category, Adult and High 0.379

School '

(Isa 0.681

0.735
tsa

Same Employer, Adult and High School 0.202
Observations 356

 

 

96

 

TABLE 17. Mean Hourly Wage by Occupational and Employer Status

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean Hourly
Observations Wage
Overall Sample 356 $10.21
Same Occupation While in High School 135 $10.26
Different Occupation While in High School 221 $10.19
t-test for Differences in Means t = 0.02
Same Employer While in High School 72 $10.41
Different Employer While in High School 284 $10.16
t-test for Differences in Means t = 0.15
Same Occupation, Same Emplger While in HS . 70 $10.43
Same Occupation, Different Employer While in HS 65 $10.07
Different Occupation, Same Employer While in HS 2 $9.63
Different Occupation, Different Employer While in HS 219 $10.19
F-test for Differences in Means F = 0.08

 

Note: “Occupation ” is deﬁned as the 14 categories listed in Table I.

prima facie evidence of the inﬂuence occupation-speciﬁc or ﬁrm-speciﬁc human capital

in dictating post-schooling wages.
RESULTS

While the summary approach taken in Table 17 failed to validate the inﬂuence of
occupation-speciﬁc or ﬁrm-speciﬁc human capital in predicting post-school wages, a
more careful analysis of the matter is undertaken through the estimation of the wage
equation presented earlier in this paper. Table 18 presents the results of a wage equation
that ﬁrst excludes the “ﬁrm” variable indicating employment with the same company at

both the in-school and post-school time periods. Contrary to the summary results of Table

97

 

TABLE 18. Regression Analysis on 1n(Hourly nge), Without Firm Controls

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Same
Occupational tsa qsa
Ln(wage) Category
Race/Ethnicity
Black -0.028 -0.023 -0.036
(0.060) (0.060) (0.060)
Hispanic -0.026 -0.025 -0.027
(0.070) (0.071) (0.070)
White / Other Base Base Base
Region
Northeast 0.094* 0.093“ 0.089
(0.055) (0.055) (0.056)
North Central 0.053 0.048 0.046
(0.051) (0.051) (0.051)
South Base Base Base
West 0.080 0.073 0.073
(0.062) (0.062) (0.062)
MSA Status
City -0.037 -0.041 -0.034
(0.049) (0.049) (0.049)
Suburbs Base Base Base
Other -0.100** -0.103** -0.097**
(0.047) (0.047) (0.047)
Married at 20 0.065 0.065 0.068
(0.070) (0.070) (0.071)
Union Member at 20 0.155" 0.156" 0.151"
(0.070) (0.070) (0.070)
ASVAB pct * 100 -0.0005 -0.001 -0.001
(0.001) (0.001) (0.001)
Smoked Marijuana in HS -0.034 ' -0.030 -0.033
(0.040) 40.040) (0.040)
Second Language at Home 0.012 0.011 0.014
(0.077) (0.077) (0.077)
OSHC 0.078* 0.170* 0.132
(0.041) (0.100) (0.086)
Occupation at age 20 (14 categories) Yes Yes Yes
R-squared 0.2669 0.2652 0.2640
Observations 356 356 356

 

 

Notes: Standard errors in parentheses. * - p<0. IO; ** - p<0.05; *** - p<0.0].

98

 

17, the regression estimates in the ﬁrst column of Table 18 demonstrate that, among
terminal high school, graduates, those who stay within the same occupational category
earn 7.8 percent more than those who have switched occupations, ceteris paribus, a result
signiﬁcant in a one-tailed, 5-percent test. Buttressed by reasonable coefﬁcients on the
other variables in the model, the positive coefﬁcient on OSHC is suggestive of a
signiﬁcant inﬂuence of high school employment in the generation of “marketable skills,”
considered by this study to be the development of occupation-speciﬁc human capital, on

post-school wages.

The ﬁnding that high school employment develops occupation-speciﬁc human capital is

further supported by the coefﬁcient on OSHC transferability in the second and third

columns of Table 18. The results using tsa suggest that a 10-percentage point increase in

the transferability of occupation-speciﬁc human capital from one’s high school vocation

to their adult profession would be expected to increase wages by 1.70 percent. The

coefﬁcient on tsa is signiﬁcant in a one-tailed, S-percent test, and the remaining

coefﬁcients are of reasonable sign and magnitude. The results in the third column of

Table 18 also support the hypothesis that in-school employment augments occupation-

speciﬁc human capital, as the coefﬁcient on Clsa indicates that a IO-percentage point

increase in the proportion to which in-school employment makes one qualiﬁed for the
post-school occupation would be expected to increase wages by 1.32 percent. This effect

is statistically signiﬁcant in a one-tailed, 10-percent test.

99

While the results of Table 18 are indicative of signiﬁcant positive effects of occupation-
speciﬁc human capital acquired through high-school employment on post-school
economic outcomes, these results disappear upon introducing a control variable
indicating whether the individual works for the same employer that they did while in high
school. The results of Table 19 demonstrate that upon accounting for ﬁrm-speciﬁc human
capital, as denoted by employment with the same ﬁrm both while enrolled and two years
following graduation, the coefﬁcients on all three OSHC variables are no longer
signiﬁcant at any reasonable statistical threshold. For instance, the coefﬁcient on OSHC
in the ﬁrst column suggests that post-school employment in the same vocational category
that one was working asa high school senior would be associated with a 6 percent wage
premium. However, such a result fails to be statistically signiﬁcant (t = 1.17, p = 0.241).

This trend of positive, yet not statistically signiﬁcant, coefﬁcients on OSHC is mirrored

in the second and third columns of Table 19, as noted by the coefﬁcients on tsa (B =

0.114, t = 0.88, p = 0.379) and qsa (B = 0.070, t = 0.59, p = 0.554). All other coefﬁcients

within these models are of reasonable sign and magnitude.

There are two possible explanations for the failure of the OSHC coefﬁcients to remain
statistically signiﬁcant upon the inclusion of the ﬁrm variable. First, the results of Table
18 may be the product of omitted variable bias, as the predicted OSHC effect might be
capturing both occupation-speciﬁc and ﬁrm-speciﬁc inﬂuences. Exacerbated by the
disproportionate number of individuals working in the same occupation that are also

employed by the same ﬁrm, one would expect that excluding employer controls would

100

TABLE 19.

tsa

Occupational

 

errors in

101

lead to an upward-biased coefﬁcient on OSHC. That said, an alternative explanation of
the decreased signiﬁcance of the OSHC coefﬁcients in Table 19 would be suggestive of
collinearity between occupation and ﬁrm controls, and there is some evidence from the
data supporting this hypothesis.58 Therefore, especially considering the sample size
limitations of this paper, the conclusions drawn from the results of Table 19 must be
tempered by the recognition of potential collinearity effects in the absence of more

sophisticated controls.59

 

Potential evidence of collinearity between the OSHC and ﬁrm variables is three-fold. First, the

coefﬁcients on the OSHC and ﬁrm variables are signiﬁcant when the other variable is excluded from the
model, but fail to be signiﬁcant upon the inclusion of the other as indicated in Table 19, including a lack of
joint signiﬁcance as denoted by the F-statistic. Second, there is considerable correlation between the three

OSHC variables and the employer control (e.g., r = 0.6424 for Isa and the ﬁrm indicator). Finally, the

computed VIFs for the occupation and ﬁrm variables from the models speciﬁed in Table 19 range from
1.74 to 2.15.

An alternative speciﬁcation of the results would be to include individuals who did not work four

months prior to high school graduation and setting all three of their OSHC variables equal to zero since
there would be no transferability of occupation-speciﬁc human capital from in-school employment to post-
school work. A concern with this speciﬁcation would be that, absent other controls, occupation and ﬁrm
variables would be proxying for the labor supply decision made in high school that was found to be

signiﬁcant in Ruhm (1997) and Light (1999), among others. In other words, the move from tsa=0.30 to

tsa=0.40 is drastically different than the move from tsa=0 to tsa=0.10; while the former speaks of

differences in the OSHC transferability of in-school employment, the latter represents the move from the
non-employment to employment while enrolled in secondary school. As a means of correcting this, the
speciﬁcation applied here includes an indicator variable denoting whether a person worked in high school.
The results of this strategy are included in Appendix C. While this model may suffer from even more
substantial collinearity than the speciﬁcation of Table 19, the results of this speciﬁcation suggest a similar
outcome: small, positive coefﬁcients on the occupation-speciﬁc and ﬁrm-speciﬁc human capital variables,
each of which fail to be statistically signiﬁcant in any reasonable test. Another potential concern about the
results of Table 19 are that they might be inﬂuenced by selectivity bias considering the restrictions of the
sample to only include terminal high school graduates. In other words, this analysis could be clouded by
endogeneity given the potential inﬂuence of in-school employment in dictating educational choice,
particularly the decision to pursue higher education. However, the application of a multinomial logit
sample selection correction outlined in Bourguignon, Foumier, and Gurgand (2007)—across a variety of
academic outcome structures—yielded no signiﬁcant evidence of endogeneity. Full results are available
from the author.

102

DISCUSSION

Recent, more sophisticated efforts by Ruhm (1997) and Light (1999) have only
conﬁrmed what decades of prior research illuminated: high school employment is
positively correlated with post-school economic outcomes, including earnings. To this
point, the standard causal inference has been that working while attending secondary
school allows an individual to accrue “marketable skills” that they can apply in the labor
market, an increase in human capital stock that generates higher wages. This paper has
examined this hypothesis using data from the NLSY97 on terminal high school
graduates, deﬁning the generation of marketable skills as the development of occupation-
speciﬁc human capital. To conﬁrm the role of high school employment as a signiﬁcant
“skill-enhancing” inﬂuence, one should ﬁnd higher rates of OSHC transferability

between high school and adult vocations associated with higher earnings.

On the surface, the results of this paper cast some doubt on the hypothesis that high
school employment provides a means of generating marketable skills. While simple
models exhibited positive, signiﬁcant effects of occupation-speciﬁc human capital
transferability, these inﬂuences disappeared once ﬁrm-speciﬁc effects were introduced.
However, potential collinearity between variables—exacerbated by a limited sample
size60—suggests that this paper by no means represents the ﬁnal word on the prospect of

high school employment generating marketable skills that are responsible for the post-

 

As future rounds of the NLSY97 become available, this paper will be updated to reﬂect the increase in

available observations, as well as explore the inﬂuence of occupation-speciﬁc human capital on a longer
time horizon of post-school economic outcomes given its potential obsolescence over time (Ruhm, 1997).

103

schooling economic gains of in-school job holding. A more careful analysis, including a
larger sample size, longer time horizon, and more sophisticated measures, may be
necessary to provide a more deﬁnitive conclusion regarding the hypothesis that
employment during secondary school instills marketable skills utilized in the labor

market during adulthood.

Regardless of the above results, it is hoped that the questions posed in this paper may
lead to a more careful analysis of the causal factors behind the positive relationship
between high school employment and post-school economic outcomes. In particular,
efforts to link in-school work with post-school gains routinely cite increases in human
capital, however such statements lack speciﬁcity and empirical backing. Thus, it is an
open question to determine how high school employment affects an individual’s complex
human capital stock as well as identifying potential institutional reasons behind the
positive relationship between in-school work and post-school outcomes. To the latter
point, a full examination of the topic must consider high school employment as a classic
“signal” in. the labor market; having prior work experience gives otherwise similar
individuals a distinct advantage in hiring and wage-determination outcomes. In addition,
prior employment may alter job-search strategies. Following exposure to entry-level
positions, more experienced workers may be more likely to actively initiate job search
strategies to locate higher-paid openings and have the self-efﬁcacy to obtain such a

position given their prior work history.

104

With decades of research demonstrating the positive effect of in-school employment on
post-school economic outcomes, it is imperative to begin to untangle the rationale behind
such a relationship, as well as how high school employment affects human capital, job
search strategies, occupational choice, and hiring and wage decisions on the part of
potential employers. Considering the potential negative academic outcomes attributed to
in-school employment, it is important to recognize how and why students beneﬁt from
holding a job in order to tailor both parental and policy decisions. While this study has
provided initial evidence casting some doubt on the relevance of high school jobs in
building marketable skills, the issue is far from resolved; there remains signiﬁcant work
to be done to fully understand the dynamics relating in-school employment to both

human capital formation and post-school economic outcomes.

105

APPENDIX A
EXAMINING OCCUPATION AS A MEDIATOR
BETWEEN EDUCATION AND WAGES
In examining the possibility that occupation serves as a mediator to the effect of
education on wages, one can turn to Baron and Kenny (1986) to examine the issue. The
authors describe a mediator relationship as follows: “In general, a given variable may be
said to function as a mediator to the extent that it accounts for the relation between the
predictor and the criterion... Whereas moderator variables specify when certain effects
will hold, mediators speak to how or why such effects occur,” (p. 1176). Full mediation,
in this scenario, would suggest that education operates solely through occupation; in
contrast, partial mediation recognizes the possibility of education effects wages two ways

— independently and through its role in occupational selection.

Given the above, a model of occupation as a mediator in the wage equation envelops
many of the relationships stressed in the better part of this paper. First and foremost, it
recognizes that the inﬂuence of education on wages can maintain both a direct and
indirect effect. As espoused by human capital theorists, this relationship allows for the
direct impact of schooling on wages, presumably through the development of one’s
human capital stock. However, the causal links inherent in the model acknowledge that
education also serves as a means of entry into particular occupations, through which

vocational-speciﬁc wage structures inﬂuence the outcome variable. As such, occupation

106

serves as a “how or why” individuals earn speciﬁc incomes based on their human capital
characteristics, thus mediating the relationship between education and earnings.
Therefore, the notion of occupation as a mediator explicitly captures the human capital
argument as well as the signaling/occupational contention advanced by Fogel (1979),

amongst others.

In turning to Baron and Kenny (1986), demonstrating a mediated relationship in this
situation requires three things. First, turning to Figure 2, a regression of education on
occupation must yield a statistically signiﬁcant coefﬁcient on the independent variable.
Likewise, a second condition mandates that a regression of occupation on wages results
in a statistically signiﬁcant coefﬁcient on the independent variable. Finally, a regression
of education and occupation on wages statistically reduces the coefﬁcient on education
when compared to a regression of education on wages. In other words, the three

conditions are as follows:

(a) occupation = 130 + B1education + ai
(b) wages = oro + a1occupation + 8]
(C) wages = 50 + 81 education + 620ccupation + ck

(d) wages = p0 + p1education + am

107

Condition (1): |B1| > 0

Condition (2): |a1| > 0

Condition (3): |p1| > |81|

While the above method of demonstrating mediation may seem straightforward,
difﬁculty arises in the present case since occupation is not a quantitative variable in
which one can apply OLS, as occupational choice models typically employ unordered,
logit estimation methods (e.g., Schmidt and Strauss 1975). This is problematic in this
situation, as Condition (1) becomes nonsensical in the face of such a logit estimation
procedure. To rectify this issue, the determination of mediation requires occupation, as
the dependent variable in (a), to be a continuous measure aligned with the causality of the
model espoused in Figure 2. In particular, the path from education to occupation to wages
represents the contentions of F ogel (1979) and that of the signaling literature, namely that

schooling serves to unlock doors to “good” jobs that feature elevated wages.

One’s ﬁrst inclination towards a quantitative proxy for occupation may be measures of
occupational prestige or socioeconomic indices, as each serves measures the social
“goodness” of a vocation. However, how these measures are constructed introduces a
reverse causality between wages and occupation, as the occupation variable would be
partially determined by the level of wages attributed to it. As such, this introduction of
reverse causality would invalidate occupation as a mediator, as deﬁned in Baron and

Kenny (1986).

108

Given the causal conditions required to test mediation, an alternative approach would be
to proxy occupation with the amount of knowledge required by the vocation, as it could
be argued that one of education’s primary goals is to equip students with the knowledge
necessary to obtain, and succeed in, a position within a particular vocation. As such, one
can exploit the information provided by ONET, a service offered by the US Department
of Labor that replaced the Dictionary of Occupational Titles (DOT) as the govemment’s
source of information on the knowledge, skills, abilities, and work activities for hundreds
occupations across the country. Within the ONET database, each occupation is scored on
the importance (rated from 1-5) and level (rated from 0-7) of 33 different knowledge
categories. Given that most post-secondary educational programs are designed to impart
speciﬁc knowledge, this paper captures each occupation’s knowledge requirement by
multiplying each knowledge importance (1M) score with its corresponding level (LV)
score for each vocation while utilizing a linear education variable. As such, “occupation”

is thus the calculation of:

33
(1) Occupationm = lip/31x ; (1M 1,- )(L Vi]. ) ,

where j represents each occupation and i denotes each of the 33 knowledge categories.
Any incongruities between the PUMS data and the more-speciﬁc occupational coding
structure of ONET are resolved by averaging separate ONET occupations found

underlying SOC codes employed by the PUMS data, a method employed by Hirsch

109

(2005). By utilizing the maximum knowledge score for each occupation, the vocational
score meets the criteria called for - a continuous variable ranging in possibility from 0 to
35 with strong, one-way causal inferences from education to occupation and occupation
to wages (and not in the reverse). In addition to its theoretical beneﬁts, the occupational

scores also make intuitive sense, with the top scoring occupations listed below: -

APPENDIX TABLE Al. Top-Scoring Occupations and Knowledge Categories
According to O*NET Rankings

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Occupation (SOC Code) Score Knowledge Category
Sociologists (19-3041) 33.45 Sociology
Mathematicians (15-202 1) 33.30 Mathematics
Interpreters and Translators (27-3090) 33.30 Foreign Language
Comuter Hardware Engineers Q7-2061) 33.30 Computers & Electronics
Aeroyspace Engineers (17-201 1) 32.63 Engineering & Technology
Actuaries (15-201 1) 32.44 Mathematics
Astronomers and Physicists (19-2010) 32.35 Physics
Physicians and Surgeons (29-1060) 32.19 Medicine and Dentistry
Computer Software Engineers (15-1030) 32.15 Computers & Electronics
Instructional Coordinators (25-9031) 31.70 Education & Training

 

While not perfect, the results give the highest scores of vocations typically requiring the
most schooling regardless of pay levels — physicians and surgeons, astronomers and

physicists, mathematicians, etc.

With the variables in hand, the computation of the four equations produces the following

estimations (standard errors in parentheses):

110

 

(a) occupation — 7.3886 + 0.9541 *education

(0.0029)

(b) 1n(wage) = 2.0969 + 0.0363*occupation
(0.0002)

(c) 1n(wage) = 1.4695 + 0.0691*education + 0.0210*occupation
(0.0003) (0.0002)

(d) 1n(wage) = 1.6244 + 0.0891*education

(0.003)

In examining the estimation results above, it is obvious that the ﬁrst two conditions
necessary to conﬁrm occupation’s place as a mediator hold, as the coefﬁcients in
equations (a) and (b) are indeed statistically signiﬁcant, indicating a positive impact of
education on occupation, and of occupation on wages. For the third condition, the results
in equations (0) and (d) indicate a direct effect of 0.0691 of education on the log wage
variable, with an indirect effect of 0.0200 (00891-00691) of education on log wage
operating through the occupation variable. To verify that occupation serves as a mediator,
it must be shown that the indirect effect is statistically signiﬁcant, thus fulﬁlling
Condition (3). From Baron and Kenny (1986), it can be shown that the standard error of

this indirect effect is:

 

_ 2 2 2 2 2 2
Sindirect —\/b Sa +61 Sb +SaSb

Sindirect : 00002

111

where a represents the coefﬁcient on education in equation (a) and b denotes the
coefﬁcient on occupation in equation (b). As a result, the corresponding t-statistic for the

indirect effect is:

100.0

 

 

t _ p, — 5, _ 0.0891 — 0.0691 _ 0.0200 _
’"d’m’ s 0.0002 0.0002

indirect

As the above ﬁnding fulﬁlls Condition (3), it veriﬁes the presence of an indirect effect of
education on wages acting through occupation. Therefore, the above examination
conﬁrms the relationship between education and occupation in the wage equation as

espoused in Figure 2, namely that occupation serves as a partial mediator in the inﬂuence

of schooling on wages.

112

APPENDIX B

APPENDIX TABLE B1. Nested Logit Results,
Manufacturing Workers, Females, 2000-2004 DWS

Displaced Blue-Collar

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Participation Choice Occupational Choice
(Branch) (Twig)
' Profession Blue-
Unemplo NILF al, Service Sales Clerical Collar,
yed Managem Non-Man.
ent
Age
20-24 0.1 13 1.518““‘l -0.115 -0.035 0.425 0.746 0.728
(0.652) (0.609) (0.956) (0.712) (0.762) (0.712) (0.730)
25-34 0.409 0.655 -0.073 0.397 0.072 1.162“ 0.365
(0.419) (0.440) (0.635) (0.445) (0.576) (0.464) @587)
35-44 Base Base Base Base Base Base Base
45-54 0.358 -0.355 -0.355 -0.037 -1.154* -0.314 0.279
(0.367) (0.410) (0.645) (0.447) (0.668) (0.520) (0.563)
55-64 0.712 0.530 “ 0.894 -0.064 -2.260** 0.617
(0.514) (0.522) (0.543) (0.714) (1.135) (0.716)
65 & up 0.492 1.799“ * * “ * *
(0.958) (0.869) 1
Job Tenure 0.007 0.042“ -0.012 -0.043* -0.002 0.022 -0.094**
(0.021) @020) (0.049) (0.026) (0.034) (0.029) (0.048)
Race
Black -0.236 -0.775* -1.012 0.056 -0.117 -2.1 l 1*“ -0.593
(0.394) (0.429) (0.926) (0.428) (0.613) (0.716) (0.667)
Other -0.432 -0.957* * -0.883 -1 .338 "‘ -0.748
(0.477) (0.522) 40.688) (1.100) (0.825
Married -0.495* 0.097 -0.842 -0.456 0.051 -0.019 -0.876**
(0.281) (0.296) (0.516) (0.323) (0)437) (0.372) (0.425)
Children <18 0.112 -0. 145 0.031 -0.l33 -0.282 -0.313* -0.218
_ (0.120) (0.137) (0.209) (0.147) (0.21 1) (0.178) (0.196)
Union, Old Job -0.142 -0.044 -l.934* 0.012 -0.524 -0.445 0.079
(0.416) (0.430) (1.140) (0.468) (0.733) (0.527) (0.631)
US Citizen 0.903“ 0.762* 0.461 0.199 0.554 0.681 0.122
(0.400) (0.420) (0.682) (0.435) (0.619) (0.487) (0.517)
Education
Less than HS -0.243 -0.485 -2.097* 0.015 -0.837 -2.632*** -0.l95
(0.308) (0.330) (1.089) (0.358) (0.562) (0.781) (0.459)
HS Diploma Base Base Base Base Base Base Base
Some 0.007 0.623 0.059 0.125 -0.061 0.087 -0.294
College (0.446L (0.434) (0.639) (0.480) (0.603) (0.467) 41.6449)
Vocational or 0.360 -0.417 -0.308 0.978 0.355 0.987 0.399
Assoc. lg (0.778) (0.977) (1.198) (0.735) (0.936) (0.702) (0.935)
College Deg. 1.486 1.268 2.418" * "' -0.627 *
(0.952) (0.988) (0.968) (1.285)

 

113

 

APPENDIX TABLE B1.
Manufacturing Workers, Females, 2000-2004 DWS (cont.)

Nested Logit

Results,

Displaced Blue-Collar

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Region
Northeast 0.103 -0.523 -0.044 0.132 -0.103 -0.988* -0.452
(0.437) (0.460) (0.773) (0.537) @715) (0.578) (0.632)
Midwest -0.877* -0.232 0.089 0.332 -0.175 -0.495 -0. 181
(0.478) (0.414) (0.715) (0.531) (0.713) (0.527) (0.608)
South -0.233 —0.394 -0.859 0.197 -0.082 -1.136** -0.632
(0.414) (0.422) (0.810) (0.495) (0.671) (0.543) (0.581)
West Base Base Base Base Base Base Base
C lg Status
In City 0.244 0.403 -1.156 -0.521 -0.793 0.708 -0.484
(0.382) (0.401) (0.871) (0.444) (0.659) (0.489) (0.631)
Balance of -0.138 -0.171 -0.936 -0.691* -1.139** -0.550 -0.032
City (0.336) (0.346) (0.583) (0.380) (0.563) (0.447) (0.482)
Non-Metro Base Base Base Base Base Base Base
Other -0.418 -0.484 -1 .295* -1.368** -0.737 -0.025 -0. 131
(0.428) ‘ (0.447) (0.760) (0.564) (0.594) (0.501) @585)
Notice of -0.240 -0.160 0.184 -0.354 -0.407 -0.032 0.358
Layoff (0.284) (0.296) (0.500) (0.329) (0.444) (0.369) (0.412)
UI Exhaustee -0.1 19 0.099 0.414 0.365 -0.572 0.495 -0.241
(0.3 15) (0.324) (0.568) (0.349) (0.540) (0.398) (0.498)
Inclusive Em- 1.223 Unem- 0.5000 NILF 0.5000
Values ployed ployed
Observations 633
L0 -likelihood -1029.0818
LR ch12(169) 574.4094
Prob > ch12 0.0000

 

 

Note: Standard error in parentheses.

114

 

APPENDIX C

APPENDIX TABLE C1. Regression Analysis on 1n(Hourly Wage), With Firm
Controls, SettingOSHC=0 for All Non-Employed High School Seniors

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Same
Occupational tsa qsa
Ln(wage) Category
Race/Ethnicity
Black -0.045 -0.043 -0.048
(0.040) (0.040) (0.040)
Hispanic -0.006 -0.006 -0.006
(0.050) (0.050) (0.050)
White / Other Base Base Base
Region
Northeast 0080* 0.081 * * 0.078*
(0.041) (0.041) (0.041)
North Central 0067* 0.065* 0064*
(0.037) (0.037) (0.037)
South Base Base Base
West 0122*” 0121*” 0.120***
(0.041) (0.041) (0.041)
MSA Status
City -0.014 -0.015 -0.013
(0.034) (0.034) (0.034)
Suburbs Base Base Base
Other -0.088" -0.089*** -0.086**
(0.034) (0.034) (0.034)
Married at 20 0.029 0.028 0.029
(0.051) (0.051) (0.051)
Union Member at 20 0169*“ 0.168"'** 0.167***
(0.046) (0.046) (0.046)
ASVAB pct "' 100 0.0001 0.0001 0.0001
(0.0005) (0.0005) (0.0005)

 

115

 

APPENDIX TABLE Cl. Regression Analysis on 1n(Hourly Wage), With Firm

Controls, Setting OSHC=0 for All Non-Employed High School Seniors (cont.)

 

 

 

 

 

 

 

 

 

 

 

 

Smoked Marijuana in HS -0.033 -0.032 -0.033
£028) (0.028) (0.028)
Second Language at Home -0.031 -0.033 -0.032
(0.053) (0.053) (0.053)
Worked in High School 0.062" 0.022 0.026
(0.031) (0.061) (0.057)
OSHC 0.062 0.085 0.085
(0.046) (0.085) (0.084)
Worked at Same Firm 0.020 0.037 0.032
(0.054) (0.051) (0.054)
Occupation at age 20 (14 categories) Yes Yes Yes
R-squared 0.2597 0.2586 0.2586
Observations 598 598 598

 

 

 

 

Notes: Standard errors in parentheses. * - p<0.10; "”" - p<0.05,' *** - p<0.01.

116

 

REFERENCES

Acemoglu, Daron, and Jom-Steffen Pischke. 1998. “Why Do Firms Train? Theory and
Evidence,” Quarterly Journal of Economics 113, 79-119.

Acemoglu, Daron and Jom-Steffen Pischke. 1999. “Beyond Becker: Training in
Imperfect Labor Markets,” Economic Journal 109, 1 12-42.

Addison, John T., and Pedro Portugal. 1989. “Job Displacement, Relative Wage Changes,
and Duration of Unemployment,” Journal of Labor Economics 7, 281-302.

Alsaker, Francoise D., and Flammer, August. 1999. Time Use by Adolescents in an
International Perspective: 11. The Case of Necessary Activities. Mahwah, N.J.: Erlbaum.

Altonji,'Joseph G. 1995. “The Effects of High School Curriculum on Education and
Labor Market Outcomes,” The Journal of Human Resources 30, 409-38.

Anderson, Deborah, and David Shapiro. 1996. “Racial Differences in Access to High-
Paying Jobs and the Wage Gap between Black and White Women,” Industrial and Labor
Relations Review 49, 273-86.

Arai, Mahmood. 2003. “Wages, Proﬁts, and Capital Intensity: Evidence from Matched
Worker-Finn Data,” Journal of Labor Economics 21, 593-618.

Bailey, T., and Roger Waldinger. 1991. “Primary, Secondary, and Enclave Labor
Markets: A Training Systems Approach,” American Sociological Review 56, 432-45.

Baron, Reuben M., and David A. Kenny. 1986. “The Moderator-Mediator Variable
Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical
Considerations,” Journal of Personality and Social Psychology 51 , 1 173-82.

Becker, Gary S. 1962. “Investment in Human Capital: A Theoretical Analysis.” The
Journal of Political Economy 70(2), 9-49.

Becker, Gary S. 1993. Human Capital. National Bureau of Economic Research: The
University of Chicago Press.

Belman, Dale and Heywood, John. 1991. “Sheepskin Effects in the Return to Education:

An Examination of Women and Minorities,” Review of Economics and Statistics 73, 720-
24. .

117

Belzil, Christian, and Hansen, Jorgen. 2002. “Unobserved Ability and the Return to
Schooling.” Econometrica 70, 2078-91.

Bishop, John. 1998. “Occupation-Speciﬁc Versus General Education and Training,” The
Annals of the American Academy of Political and Social Science 559, 24-38.

Bloch, Farrell E. and Sharon P. Smith. 1977. “Human Capital and Labor Market
Employment,” The Journal of Human Resources 12, 550-60.

Blaug, Mark. 1976. “The Empirical Status of Human Capital Theory: A Slightly
Jaundiced Survey.” Journal of Economic Literature 14, 827-55.

Borjas, George J. 1980. “The Relationship Between Wages and Weekly Hours of Work:
The Role of Division Bias,” The Journal of Human Resources 15, 409-23.

Bourguignon, Francois, Martin Foumier, and Marc Gurgand. 2007. “Selection Bias
Corrections Based on the Multinomial Logit Model: Monte Carlo Comparisons,” Journal
of Economic Surveys, 21, 174-205.

Bowman, Mary Jean. 1974. “Learning and Earning in the Postschool Years,” In Review
of Research in Education, ed. Frank N. Kerlinger and John B. Carroll, 202-44. Itasca,
Illinois: Peacock.

Bradbury, Katherine L. 2002. “Education and Wages in the 19805 and 19903: Are All
Groups Moving Up Together?” New England Economic Review, 19-46.

Bratsberg, Bemt, James F. Ragan, Jr., and Zafar M. Nasir. 2002. “The Effect of
Naturalization on Wage Growth: A Panel Study of Young Male Immigrants,” Journal of
Labor Economics 20, 568-97.

Budd, John W. 2005. “Ideas versus Ideology: The Origins of Modern Labor Economics —
Comment on Gallaway and Vedder,” Journal of Labor Research 26, 177-80.

Cain, Glen. 1976. “The Challenge of Segmented Labor Market Theories to Orthodox
Theory: A Survey.” Journal of Economic Literature 14, 1215-57.

Card, David. 1998. “The Causal Effect of Education on Earnings.” Working Paper No. 2,
Center for Labor Economics, University of California, Berkeley.

Carr, Rhoda V., James D. Wright, and Charles J. Brody. 1996. “Effects of High School

Work Experience a Decade Later: Evidence from the National Longitudinal Survey,”
Sociology of Education, 69, 66-81.

Carrington, William J. 1993. “Wage Losses for Displaced Workers: Is It Really the Firm
That Matters?” The Journal of Human Resources 28, 43 5-62.

118

Carrington, William J., Kristin McCue, and Brooks Pierce. 1996. “The Role of
Employer/Employee Interactions in Labor Market Cycles: Evidence from the Self-
Employed,” Journal of Labor Economics 14, 571-602.

D’Amico, Ronald. 1984. “Does Employment During High School Impair Academic
Progress?” Sociology of Education, 57, 152-164.

DeBeyer, J., and IE. Knight 1989. “The Role of Occupation in the Determination of
Wages.” Oxford Economic Papers 41, 595-618.

DeSimone, Jeff. 2006. “Academic Performance and Part-Time Employment among High
School Seniors,” Topics in Economic Analysis and Policy, 6, 1466.

Dickens, W., and K. Lang. 1985. “A test of dual market theory,” American Economic
Review 75, 792-805.

Doeringer, Peter, and Michael Piore. 1971. Internal Labor Markets and Manpower
Analysis. New York: Sharpe.

Dolton, Peter J ., and Michael P. Kidd. 1998. “Job Changes, Occupational Mobility and
Human Capital Acquisition: An Empirical Analysis,” Bulletin of Economic Research 50,
265-95.

Dorman, Peter, and Paul Hagstrom. 1998. “Wage Compensation for Dangerous Work
Revisited,” Industrial and Labor Relations Review 52, 116-35.

Duncan, Greg J., and Bertil Holmlund. 1983. “Was Adam Smith Right After All?
Another Test of the Theory on Compensating Wage Differentials,” Journal of Labor
Economics 1, 366-79.

Duncan, Otis Dudley. 1961. “Occupational Components of Educational Differences in
Income,” Journal of the American Statistical Association 56, 783-92.

Eckaus, Richard S. 1973. Estimating the Returns to Education: A Disaggregated
Approach. Carnegie Foundation.

Evans, Davis 8., and Linda S. Leighton. 1995. “Retrospective Bias in the Displaced
Worker Surveys,” Journal of Human Resources 30, 386-96.

Fairlie, Robert W., and Lori G. Kletzer. 1996. “Race and the Shifting Burden of Job
Displacement from 1982 to 1993,” Monthly Labor Review 119, 13-23.

Fallick, Bruce. 1996. “A Review of the Recent Empirical Literature on Displaced
Workers,” Industrial and Labor Relations Review 50, 5-16.

119

Farber, Henry. 2005. “What Do We Know About Job Loss in the United States?
Evidence from the Displaced Worker Survey, 1984-2004,” FRB Chicago Economic
Perspectives 29, 13-28.

Farber, Henry. 1997. “The Changing Face of Job Loss in the United States, 1981-1995,”
Brookings Papers on Economic Activity, Microeconomics, 55-128.

Farber, Henry. 1993. “The Incidence and Costs of Job Loss, 1982-1991,” Brookings
Papers on Economic Activity, Microeconomics, 73-119.

Fogel, Walter. 1979. “Occupational Earnings: Market and Institutional Inﬂuences,”
Industrial and Labor Relations Review 33, 24-35.

Fox, Douglas A. 1994. “Displacement and Occupational Mobility: Evidence from the
Displaced Worker Surveys,” Eastern Economic Journal 20, 31 1-24.

Frazis, Harley. 2002. “Human Capital, Signaling, and the Pattern of Returns to
Education,” Oxford Economic Papers 54, 298-320.

Freeman, James A. and Barry T. Hirsch. 2001. “Do Returns to Human Capital Equalize
Across Occupational Paths?” Research in Labor Economics 20, 217-42.

Fuchs, Victor R. 1982. “Self- Employment and Labor Force Participation of Older
Males,” The Journal of Human Resources 17, 339- 57.

Fuligni, Andrew J ., and Harold W. Stevenson. 1995. “Time Use and Mathematics
Achievement among American, Chinese, and Japanese High School Students,” Child
Development, 66, 830-842.

Gera, Surendra, and Gilles Greiner. 1994. “Interindustry Wage Differentials and
Efﬁciency Wages: Some Canadian Evidence,” The Canadian Journal of Economics 27,
81-100.

Gibbs, Mike, Kathy Ierulli, and Eva M. Meyersson Milgrom. 2002. “Careers in Firm and
Occupational Labor Markets,” Mimeo., CDDRL, Stanford University.

Gill, Andrew M. 1988. “Choice of Employment Status and the Wages of Employees and
the Self-Employed: Some Further Evidence,” Journal of Applied Econometrics 3, 229-
34.

Gill, Andrew M. 1994. “Incorporating the Causes of Occupational Differences in Studies
of Racial Wage Differentials,” The Journal of Human Resources 29, 20-41.

Goldstein, Jonathan P., Lloyd C. Irland, Jennifer A. Senick, and Eric W. Bassett. 2005.

“The Intergenerational Supply of Loggers Under Conditions of Declining Economic
Well-Being,” Industrial Relations 44, 331-40.

120

Gottschalk, Peter T. 1978. “A Comparison of Marginal Productivity and Earnings by
Occupation,” Industrial and Labor Relations Review 31, 368-78.

Greene, William. 2002. Econometric Analysis. Fifth Edition. Upper Saddle River,
N.J.:Prentice Hall.

Griliches, Zvi. 1977. “Estimating the Returns to Schooling: Some Econometric
Problems.” Econometrica 45, 1-22.

Groen, Jeffrey A. 2006. “Occupation-Speciﬁc Human Capital and Local Labor Markets,”
Oxford Economic Papers 58, 722-41.

Gyimah-Brempong, Kwabena, Rudy Fichtenbaum, and Gregory Willis. 1992. “The
Effects of College Education on the Male-Female Wage Differential,” Southern
Economic Journal 58, 790-804.

Hagner, David. 2000. “Primary and Secondary Labor Markets: Implications for
Vocational Rehabilitation,” Rehabilitation Counseling Bulletin 44, 22-29.

Hauser, Robert M., and John Robert Warren. 1997. “Socioeconomic Indexes for
Occupations: A Review, Update, and Critique,” Sociological Methodology 27, 177-298.

Heckman, James. 1979. “Sample Selection Bias as a Speciﬁcation Error,” Econometrica,
47, 153-161.

Hellerstein, Judith K., David Neumark, and Kenneth R. Troske. 1999. “Wages,
Productivity, and Worker Characteristics: Evidence from Plant-Level Production
Functions and Wage Equations,” Journal of Labor Economics 17, 409-46.

Hirsch, Barry T. 2005. “Why Do Part-Time Workers Earn Less? The Role of Worker and
Job Skills,” Industrial and Labor Relations Review 58, 525-51.

Hirsch, Barry T., and David A. Macpherson. 2004. “Wages, Sorting on Skill, and the
Racial Composition of Jobs,” Journal of Labor Economics 22, 189-210.

Hirsch, Barry T., and Edward J. Schumacher. 2004. “Match Bias in Wage Gap Estimates
Due to Earnings Imputation,” Journal of Labor Economics 22, 689-722.

Hotchkiss, Lawrence. 1993. “Effects of Training, Occupation, and Training-Occupation
Match on Wage,” The Journal of Human Resources 28, 482-96.

Hungerford, Thomas and Solon, Gary. 1987. “Sheepskin Effects in the Return to
Education.” Review of Economics and Statistics 69, 175-77.

121

Jacobson, Louis 8., Robert LaLonde, and Daniel Sullivan. 2005. “Is Retraining Displaced
Workers a Good Investment?” FRB Chicago Economic Perspectives, 47-66.

Jacobson, Louis 8., Robert LaLonde, and Daniel Sullivan. 1993. The Costs of Worker
Dislocation. Kalamazoo, Mich: W.E. Upjohn Institute for Employment Research.

Kambourov, Gueorgui, and Iourii Manovskii. 2004a. “Occupational Mobility and Wage
Inequality,” Working Paper;
www.econ.upenn.edu/~manovski/papers/occ_mob_and_wage_ineq.pdf.

Kambourov, Gueorgui, and Iourii Manovskii. 2004b. “A Cautionary Note on Using
(March) CPS Data to Study Worker Mobility,” Working Paper;
www.ccon.upenn.edu/~manovski/papers/CautionaryNote.pdf.

Kambourov, Gueorgui, and Iourii Manovskii. 2005. “Occupational Speciﬁcity of Human
Capital,” Working Paper;
www.ccon.upenn.edu/~manovski/papers/occupation__speciﬁc_HC.pdf.

Kerr, Clark. 1977. “The Balkanization of Labor Markets.” in Labor Markets and Wage
Determination, ed. Clark Kerr. London: University of California Press.

,9

Kleiner, Morris M. 2000. “Occupational Licensing, The Journal of Economic

Perspectives 14, 189-202.

Kletzer, Lori G. 2001. Job Loss from Imports: Measuring the Costs. Washington, DC:
Institute for International Economics.

Kletzer, Lori G. 1998. “Job Displacement,” The Journal of Economic Perspectives 12,
115-36.

Kletzer, Lori G., and Robert W. Fairlie. 2003. “The Long-Terrn Costs of Job
Displacement for Young Adult Workers,” Industrial and Labor Relations Review 56,
682-98.

Knight, J .B. 1979. “Job Competition, Occupational Production Functions, and Filtering
Down,” Oxford Economic Papers 31, 187-204.

Kostiuk, Peter F. 1990. “Compensating Differentials for Shift Work,” The Journal of
Political Economy 98, 1054-75.

Kruse, William J. 1977. “An Empirical Study of Labor Market Segmentation:
Comment.” Industrial and Labor Relations Review 30, 508-23.

Kumar, Pradeep, and Mary Lou Coates. 1982. “Occupational Earnings, Compensating
Differentials, and Human Capital: An Empirical Study,” The Canadian Journal of
Economics, 15, 442-57.

122

i \.

 

 

 

 

Langley, Paul C. 1978. “An Empirical Study of Labor Market Segmentation: Comment.”
Industrial and Labor Relations Review 32, 86-92.

Layard, Richard, and George Psacharopoulos. 1974. “The Screening Hypothesis and the
Returns to Education,” The Journal of Political Economy 82, 985-98.

Lazear, Edward P. 2003. “Finn-Speciﬁc Human Capital: A Skill-Weights Approach,”
NBER Working Paper No. W9679.

Lazear, Edward P. 1995. “A Jobs-Based Analysis of Labor Markets,” The American
Economic Review 85, in Papers and Proceedings of the Hundredth and Seventh Annual
Meeting of the American Economic Association, Washington, DC, January 6-8, 1995,
260-65.

Leonard, Thomas C. 2000. “The Very Idea of Applying Economics: The Modern
Minimum-Wage Controversy and Its Antecedents,” History of Political Economy 32,
117-44.

Light, Audrey. 1999. “High School Employment, High School Curriculum, and Post-
School Wages,” Economics of Education Review, 18, 291-309.

Lofstrom, Magnus. 2001. “A Comparison of the Human Capital and Signaling Models:
The Case of the Self-Employed and the Increase in the Schooling Premium in the 1980s,”
Research in Labor Economics 20, 191-215.

MacDonald, C., and C. Sirianni. 1996. Working in the Service Society. Philadelphia:
Temple University Press.

Macpherson, David A., and Barry T. Hirsch. 1995. “Wages and Gender Composition:
Why Do Women’s Jobs Pay Less?” Journal of Labor Economics 13, 426-71.

Marin, Alan, and George Psacharopoulos. 1982. “The Reward for Risk in the Labor
Market: Evidence from the United Kingdom and a Reconciliation with Other Studies,”
The Journal of Political Economy 90, 827-53.

Marsh, Herbert W. 1991. “Employment During High School: Character Building or a
Subversion of Academic Goals?” Sociology of Education, 64, 172-189.

Marsh, Herbert W., and Sabina Kleitman. 2005. “Consequences of Employment During
High School: Character Building, Subversion of Academic Goals, or a Threshold?”

American Educational Research Journal, 42, 33 1 -3 69.

Mayhew, Anne. 1971. “Education, Occupation, and Earnings.” Industrial and Labor
Relations Review 24, 216-25.

123

Mazerolle, Maurice J., and Gangaram Singh. 2004. “Economic and Social Correlates of
Re-Employment Following Job Displacement: Evidence from 21 Plant Closures in
Ontario,” The American Journal of Economics and Sociology 63, 717-30.

McNeal, Ralph B. 1997. “Are Students Being Pulled Out of High School? The Effect of
Adolescent Employment on Dropping Out,” Sociology of Education, 70, 206-220.

McNeal, Ralph B. 1995. “Extracurricular Activities and High School Dropouts,”
Sociology of Education, 68, 62-81.

Meyer, Robert H., and David A. Wise. 1982. “High School Preparation and Early Labor
Force Experience,” In The Youth Labor Market Problem: Its Nature, Causes, and
Consequences, ed. Richard B. Freeman and David A. Wise, 277-341. Chicago:
University of Chicago Press.

Michael, Robert T., and Nancy Brandon Tuma. 1984. “Youth Employment: Does Life
Begin at 16?” Journal of Labor Economics, 2, 464-476.

Milgrom, EM. 2004. “Boundaries of Internal Labor Markets: The Relative Importance of
Firms and Occupations,” Working Paper;
www.meyersson.com/boundary1 10804.pdf

Mincer, Jacob. 1958. “Investment in Human Capital and Personal Income Distribution,”
Journal of Political Economy 4, 281-302.

Mincer, Jacob. 1974. Schooling, Experience, and Earnings. National Bureau of
Economic Research: Columbia University Press.

Mortimer, Jeylan T., and Michael D. Finch. 1986. “The Effects of Part-time Work on
Adolescent Self-Concept and Achievement,” In Becoming a Worker, edited by K.
Borman and J. Reisman, 66-89. Norwood, N.J.: Ablex.

Mortimer, Jeylan T., Michael Finch, Michael Shanahan, and Seongryeol Ryu. 1992.
”Work Experience, Mental Health, and Behavioral Adjustment in Adolescence,” Journal
of Research on Adolescence, 2, 25-57.

Nakeo, Keiko, and Judith Treas. 1994. “Updating Occupational Prestige and
Socioeconomic Scores: How the New Measures Measure Up,” Sociological Methodology

24, 1-72.

Neal, Derek. 1995. “Industry-Speciﬁc Human Capital: Evidence from Displaced
Workers,” Journal of Labor Economics 13, 653-77.

Oaxaca, Ronald. 1973. “Male-Female Wage Differentials in Urban Labor Markets,”
International Economic Review 14, 693-709.

124

Orazem, Peter F. and J. Peter Mattila. 1991. “Human Capital, Uncertain Wage
Distributions, and Occupational and Educational Choices,” International Economic
Review 32, 103-22.

Osterman, Paul. 1975. “An Empirical Study of Labor Market Segmentation.” Industrial
and Labor Relations Review 28, 508-23.

Osterman, Paul. 1977. “An Empirical Study of Labor Market Segmentation: Reply.”
Industrial and Labor Relations Review 30, 221—24.

Parent, Daniel. 2000. “Industry-Speciﬁc Capital and the Wage Proﬁle: Evidence from the
National Longitudinal Survey of Youth and the Panel Study of Income Dynamics,”
Journal of Labor Economics 18, 306-23.

Picard, Pierre. 1993. Wages and Unemployment. Cambridge, UK: Cambridge University
Press.

Poletaev, Maxim, and Chris Robinson. 2004. “Human Capital and Skill Speciﬁcity,”
University of Western Ontario Working Paper #2003-6.

Prasch, Robert B. “How is Labor Distinct from Broccoli? Some Unique Characteristics of
Labor and Their Importance for Economic Analysis and Policy,” in The Institutionalist
Tradition in Labor Economics, ed. Dell P. Champlin and Janet T. Knoedler, 146-58.
Armonk, NY: Sharpe.

Quinn, Michael A. and Stephen Rubb. 2005. “The Importance of Education-Occupation
Matching in Migration Decisions,” Demography 42, 153-67.

Reder, Melvin. 1955. “The Theory of Occupational Wage Differentials,” American
Economic Review 45, 833-52.

Reiter, Ester. 1991. Making Fast Food: From the Frying Pan into the Fryer. Montreal:
McGill-Queen’s University Press.

Riley, John G. 1979. “Testing the Educational Screening Hypothesis,” The Journal of
Political Economy 87, Part 2: Education and Income Distribution, 8227-3252.

Robst, John, and Kathleen Cuson-Graham. 1999. “The Effect of Uncertain Educational
Requirements on Education and Wages,” Applied Economics 31, 53-63.

Rothstein, Donna S. 2007. “High School Employment and Youths’ Academic
Achievement,” Journal of Human Resources, 42, 194-213.

Ruhm, Christopher. 1997. “Is High School Employment Consumption or Investment?”
Journal of Labor Economics, 15, 735-776.

125

Ruhm, Christopher. 1991. “Are Workers Permanently Scarred by Job Displacements?”
American Economic Review 81, 319-23.

Rtunberger, Russell W. 1987. “The Impact of Surplus Schooling on Productivity and
Earnings,” The Journal of Human Resources 22, 24-50.

Sakamoto, A., and M. Chen. 1991. “Inequality and attainment in a dual labor market,”
American Sociological Review 56, 295-308.

Schill, William J ., Rosemarie McCartin, and Katrina Meyer. 1985. “Youth Employment:
Its Relationship to Academic and Family Variables,” Journal of Vocational Behavior, 26,
155-163.

Schmidt, Peter and Robert P. Strauss. 1975. “The Prediction of Occupation Using
Multiple Logit Models,” International Economic Review 16, 471-86.

Schoeni, Robert F., and Michael Dardia. 1996. “Wage Losses of Displaced Workers in
the 19905,” RAND.

Schumacher, Edward J., and Barry T. Hirsch. 1997. “Compensating Differentials and
Unmeasured Ability in the Labor Market for Nurses: Why Do Hospitals Pay More?”
Industrial and Labor Relations Review 50, 557-79.

Schumann, Paul L., Dennis A. Ahlburg, and Christine Brown Mahoney. 1994. “The
Effects of Human Capital and Job Characteristics on Pay,” The Journal of Human
Resources 29, Special Issue: Women ’s Work, Wages, and Well-Being, 481-503.

Schweitzer, Stuart 0. 1971. “Occupational Choice, High School Graduation, and
Investment in Human Capital,” The Journal of Human Resources 6, 321-32.

Seitchik, Adam. 1991. “Who are Displaced Workers?” In Job Displacement:
Consequences and Implications for Policy, ed. John T. Addison Detroit: Wayne State
University Press, 51-82.

Shaw, Kathryn L. 1984. “A Formulation of the Earnings Function Using the Concept of
Occupational Investment,” The Journal of Human Resources 19, 319-40.

Sicherman, Nachum, and Oded Galor. 1990. “A Theory of Career Mobility,” The Journal
of Political Economy 98, 169-92.

Siow, Aloysius. 1984. “Occupational Choice under Uncertainty,” Econometrica 52, 631-
46.

Smith, Adam. 1976. “An Inquiry into the Nature and Causes of the Wealth of Nations,”

in The Works and Correspondence of Adam Smith, ed. R.H. Campbell. London: Oxford
University Press. ’

126

Soopramanien, Didier, and Geraint Johnes. 2001. “A New Look at Gender Effects in
Participation and Occupation Choice,” Labour 15, 415-43.

Spence, Michael. 1973. “Job Market Signaling,” The Quarterly Journal of Economics 87,
355-74.

Stephenson, Stanley P. 1981. “In-School Labor Force Status and Post-School Wage Rates
of Young Men,” Applied Economics, 13, 279-302. '

Stern, David, and Yoshifumi Nakata. 1989. “Characteristics of High School Students
Paid Jobs and Employment Experience After Graduation,” In Adolescence and Work:
Inﬂuences of Social Structures, Labor Markets, and Culture, ed. D. Stern and D. Eichom,
189-234. Hillsdale, N.J.: Erlbaum.

Stevens, Ann Huff. 1997. “Persistent Effects of Job Displacement: The Importance of
Multiple Job Losses,” Journal of Labor Economics 15, 165-88.

Sullivan, Paul. 2005a. “Empirical Evidence on Occupation and Industry Speciﬁc Human
Capital,” Working Paper;
http://www-personal.umich.edu/~paulsull/Sullivan_wagepaper_Oct2006.pdf.

Sullivan, Paul. 2005b. “A Dynamic Analysis of Educational Attainment, Occupational
Choices, and Job Search,” Working Paper; ,
'http://www-personal.umich.edu/~paulsull/Sullivan_Occpaper_sept2006.pdf.

Taubman, Paul, and Michael L. Wachter. 1986.“Segmented Labor Markets,” Handbook
of Labor Economics, Vol. 2, ed. Orley Ashenfelter and Richard Layard. Amsterdam:
North Holland.

Taubman, Paul J., and Terence J. Wales. 1973. “Higher Education, Mental Ability, and
Screening,” The Journal of Political Economy 81, 28-55.

Thomas, Ward F ., and Paul Ong. 2002. “Barriers to Rehiring of Displaced Workers: A

Study of Aerospace Engineers in California,” Economic Development Quarterly 16, 167-
78.

Topel, Robert. 1991. “Speciﬁc Capital, Mobility, and Wages: Wages Rise With Job
Seniority,” Journal of Political Economy 99, 145-75.

Vedder, Richard, and Lowell Gallaway. 2005. “Ideas versus Idedlogy: The Origins of
Modern Labor Economics — Reply to Budd,” Journal of Labor Research 26, 181-82.

Weiss, Andrew. 1995. “Human Capital vs. Signaling Explanations of Wages,” The
Journal of Economic Perspectives 9, 133-54.

127

Weiss, Yoram. 1971. “Learning by Doing and Occupational Selection,” Journal of
Economic Theory 3, 189-98.

Wheeler, Christopher H. 2001. “Search, Sorting, and Urban Agglomeration,” Journal of
Labor Economics 19, 879-99.

Wooldridge, Jeffrey M. 2001. Econometric Analysis of Cross Section and Panel Data.
Cambridge, Mass: The MIT Press.

128

lllllllllllllllllIllllllllllllll

 

..-. u. .............