41149478

THE FIRM'S CHOICE OF HRM PRACTICES: ECONOMICS MEETS STRATEGIC HUMAN
RESOURCE MANAGEMENT
Author(s): BRUCE E. KAUFMAN and BENJAMIN I. MILLER
Source: ILR Review, Vol. 64, No. 3 (April 2011), pp. 526-557
Published by: Sage Publications, Inc.
Stable URL: http://www.jstor.org/stable/41149478
Accessed: 14-11-2016 19:57 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to ILR Review
This content downloaded from 14.142.16.3 on Mon, 14 Nov 2016 19:57:23 UTC
All use subject to http://about.jstor.org/terms
THE FIRM'S CHOICE OF HRM PRACTICES: ECONOMICS MEETS

STRATEGIC HUMAN RESOURCE MANAGEMENT
BRUCE E. KAUFMAN AND BENJAMIN I. MILLER*
The authors compare and contrast two theoretical approaches to explain

a firm's choice of human resource management (HRM) practices - one fr
strategic human resource management (SHRM) and the other from econ
ics. They present HRM frequency distributions depicting key empirical p
terns that both theories must explain and then review and apply SHRM the
to explain these patterns. Since no economic model has thus far explic
considered firms' choice of HRM practices, the authors develop one based
standard microeconomic production theory. The model yields a new theor
cal construct, the HRM demand curve, and a new empirical estimating tool,
HRM demand function. Together, these provide an alternative explanatio
HRM frequency distributions, new insights on the limitations of SHRM theo
and the first alternative to the standard "Huselid-type" regression mo

Using recent survey data on HRM practices at several hundred Americ
firms, the authors estimate representative HRM demand functions to il
trate the empirical implementation of the model. They find that althou
both theoretical approaches have value, the economic model seems super
in terms of generality, logical coherence, predictive ability, and congrue
with empirical data.
interaction between the two academic realms
in both economics and
remains limited
despitehuman
the fact that potenmanagement departments
study
gains from
tradethey
are large
(Mitchell
resource management tial
(HRM),
but
use
200 1).1
Economists pride
themselves
on havquite different theories,
methods,
and
tools.
ing strong theory
and typically and
regard the
As a result, the intellectual
exchange
management HRM literature as light on sub-
stance and heavy on description and pre-
scription; management researchers, however,
* Bruce E. Kaufman is Professor of Economics and Se-
view economists' models as far too simplistic
nior Associate of the W.T. Beebe Institute of Personnel
and Employment Relations at Georgia State University

in Atlanta, Georgia; Senior Research Fellow of the Cen1 For example, in Lazear and Shaw's (2007) review artitre for Work, Organization and Wellbeing, Business
cle on Personnel Economics, only 2% (2 out of 81) of
School, at Griffith University in Brisbane, AU; and Printheir citations are to management journals (California
cipal Research Fellow of the Work and Employment
Research Unit, Business School, at the University Management
of
Review, Management Science); the figure is
Hertsfordshire in Hatfield, U.K. Benjamin Miller is 4%
a if one includes Lazear's (1986) article in the fournal
manager in the Global Transfer Pricing Group of
of Business. Conversely, in Lepak and Shaw's (2008) re-
view article on Strategic Human Resource ManageDeloitte Tax LLR An economist specializing in labor
economics, he earned his PhD from the Andrew Young

ment, only 3% (2 out of 72) of the citations are to
economics journals (the American Economic Review) . The
School of Policy Studies at Georgia State University. The
point of intersection for economics and management is
authors express appreciation to the Bureau of National
principally industrial relations journals, albeit modestly
Affairs for generously providing the data used in this
so (six citations in the former case, five in the latter).
paper.
Industrial and Labor Relations Review, Vol. 64, No. 3 (April 2011). by Cornell University.
0019-7939/00/6403 $05.00
526
THE FIRM'S CHOICE OF HRM PRACTICES 527
and abstract to capture much

productivity
of reality slowdown,
and
heightened g
competition,
and the success of Jap
pride themselves on developing
theories
that, even if less formalistic and

mathematimanagement
methods. As a result, sch
economics, industrial
relations, and mancally expressed, nonetheless insucceed
at bet-
numerous books and arter explaining how and why agement

HRM published
performs
in real-world organizations.2 ticles from the mid-1980s onward concerning
Industrial relations (IR) wasHPWS
born
the
and in
the potential
of an HPWS-type
human resource
configuration
to boost pro1920s in the United States partly
as an
attempt to bridge and where possible
ductivityintegrate
and provide firms with sustained
the economics and management
competitive
fields
advantage
since (Kochan, Ratz, and
McKersie
both are clearly central to a study
of1986;
theKleiner
em- et al. 1987; Huselid
ployment relationship - the 1995;
IR field's
Black and core
Lynch 2001; Boxali and
subjectarea (Kaufman 1993, 2010a)
Macky2009).
. Through
the 1950s, the IR field in the United States
On the management side, these develop(much less so elsewhere) was successful in
ments played a central role in defining and
performing this bridging and integrating energizing the new subfield of strategic
role, evidenced by the roster of well-known human resource management (SHRM).
labor economists/IR scholars who were also
Within SHRM /^subject of research quickly
leading researchers in personnel/HRM. Ex- became the linkage between HRM practices
amples include E. Wight Bakke, Herbert and firm performance and, most particuHeneman, Charles Myers, Sumner Slichter, larly, the oft-hypothesized positive relation-
George Strauss, and Dale Yoder (Kaufman ship between the use of "advanced" HRM
2002). In the 1960s, however, labor econom- practices and the achievement of higher
ics and management began to drift apart - profitability and other related organizational
and industrial relations to hollow out - and
performance outcomes (Combs et al. 2006;
Boxall and Purcell 2008).
by the 1980s contact and interchange between the two became minimal.
By wide agreement, a pioneering contriLabor economics and human resource
bution to this research stream is Huselid 's
management continue in most respects

to Academy of Management Journal paper,
(1995)
proceed as the two proverbial ships in
theImpact of Human Resource Manage"The
night, with one particular exception:
the
ment
Practices on Turnover, Productivity,
study of high performance work systems
and Corporate Financial Performance." The
(HPWS) and, in particular, the relationship
centerpiece of the article is a regression
between HRM practices and firm performodel that explains variation in companymance. New ideas and innovations in
level productivity, turnover, and financial
management and the behavioral
sciences as a function of two HRM comperformance
associated with high involvement
positework
variables - "employee skills and organizational
structures" built out of nine
practices, such as self-managed teams,
crosstraining, employee participation,separate
and gainHRM factors, and "employee moti
built out of four HRM factors. The
sharing forms of pay, spurredvation,"
research
interest in these topics, as did contemporaconclusion Huselid reaches is that "the mag-
neous events in the economy, such

asthe
a returns for investments in High
nitude of
Performance Work Practices [HPWP] is substantial" (p. 667) and that this positive "main
2 This bifurcation of approach and opinion goes back

persisted
many decades. Frank Stockton, dean of effect"
the business
(if perhaps attenuated)
even
different
school at the University of Kansas, observed
in when
the early
factors
are introduced.
1930s, "Labor economics men . . . disdain
personnel
control and contingent
further because of its apparent lack of theory.

The perHuselid's
article solidified the basic propsonnel instructor, on the other hand, thinks that at least
osition
that
has anchored subsequent
he is working in terms of reality, and may be inclined to
SHRM research - that an identifiable subset
dislike the fault-finding tone of labor economics and to
of HRMquespractices, typically associated

belitde the socioeconomic approach to industrial
tions" (Stockton 1932: 224).
with a high involvement and human capital
528 INDUSTRIAL AND LABOR RELATIONS REVIEW
In particular,
we re-examine the central
employment model, contributes,
on averquestion that
underlies
SHRM research age, to higher firm performance.
Thus,
the
whatthe
is thecenter
firm's optimal
testable hypothesis that lies at
of (performance
choice of HRM practices? SHRM empirical research is, maximizing)
"more HPWP
andThis
endeavor
to seriously engage the SHRM
- higher firm performance."3
proposiliterature on this
At the same time, we
tion, however, is often generalized
to topic.
apply
this matter(e.g.,
a theoretical and empirito all types of investment bring
in toHRM
cal approach
solidly grounded in economWright et al. 2005) . This expansive
statement
ics. field
Relative to
research, the
of the central hypothesis of the
inmanagement
part
model is of
indeed
simple and abstract; noneemanates from the very definition
"HRM."
we believe
That is, a number of writerstheless,
(e.g.,
Beerit deserves
and attention because it and
yields a Stodd
number of important insights
Spector 1984; Dulebohn, Ferris,
1995) have claimed that the HRM
field Inthat
and implications.
particular, it suggests
that the from
standard HRM-firm
performance
emerged in the 1980s is distinct
the
regression
model
used in management is
earlier personnel administration
and
industrial relations fields because it is based on
most likely mis-specified; the predicted effect of "more HRM" on firm performance is
both an alternative view of employees (capinot reliably positive and, indeed, should be
tal assets rather than expenses) and a differzero in a situation of long-run competitive
ent approach to management (involvement
equilibrium; and that a better-specified emrather than control) . Hence, this view posits,
pirical tool for SHRM research is what we
at least as a first order approximation, HRM
an HRM demand function. We then proHPWP and one may therefore writecall
the
ceed to estimate the parameters of an HRM
central proposition of the field as "more
demand function using cross-section data
HRM - > higher firm performance." Finding
for several hundred American firms. Aldefinitive empirical and theoretical evidence
the theoretical model, data set, and
for this proposition has been called though
the
equations can be validly criticized
"Holy Grail" of the field (Boselie, Dietz,regression
and
Boon 2005: 67) because it demonstrates that
on various grounds, the end-product is dis"HRM matters" (Gerhart 2005: 175).
tinctly new and charts a considerably differSome labor economists have also used a
ent direction for research on firms' choice
of HRM practices. At the same time, we abHuselid-type HRM-firm performance regres-
sion model in their empirical work butjure

in "economic imperialism" and in the spirit
of industrial relations attempt to inform the
general they have chosen not to cross disciand regression models with inplinary boundaries and directly engage theoretical
and
interact with SHRM researchers and the
sights and findings from management.
SHRM literature (e.g., see Black and Lynch
2001; Bartel 2004; a partial exception is Framing the Research Question

Cappelli and Neumark 2001). We seek to do
otherwise in this paper, partly because we be-The personnel/HRM field up to the early
1990s was heavily descriptive and techniquelieve there are indeed large gains to be realoriented; in the last one to two decades,
ized from the exchange between economics
however, management scholars have made a
and management and partly to help
substantial effort to put the field on a firmer
strengthen the IR field as the intellectual
theoretical foundation. Boxall, Purcell, and
bridge and integrating link between the
Wright (2007: 4) called the new approach
two.4
3 Combs et al. (2006) reviewed the literature since

Huselid (1995) and performed a meta-analysis of 92
empirical HRM-firm performance studies. They framed
the central hypothesis as, "the use of HPWPs is positively
related to organizational performance" (p. 504).

4 Other recent IR contributions include Appelbaum
et al. (2000); Delaney and Godard (2001); Gunderson
"analytical HRM" and explained that its central mission is "not to propagate perceptions
of 'best practice' in 'excellent companies'
but primarily to identify and explain what
(2001); Batt (2002); Godard (2004); Rochan (2007);
and Grimshaw and Rubery (2007).
Figure 1. Frequency Distribution of HRM Practices and H
<D
'
3
/
'
O"
/
'
'
/
I
/
'
Firms with Few Firms with Many 0 2000 4000 6000 8000 10000
Advanced HRM Advanced HRM Per Capita Expenditures on HRM
Practices Practices
Panel
(a)
Panel
(b)
Source: Freeman and Rogers (1999: 124) Source: Bureau of National Affairs (2006)
happens in practice." These authors identi- tices, some of which were simple whereas
fied "what happens in practice" as the ob- others were of increasingly greater complexserved outcomes and behaviors associated
ity or sophistication. Example practices inwith the practice of HRM; the purpose
of a personnel/HR department, an
cluded
HRM research, in turn, is to develop explanopen-door dispute resolution policy, and an
atory theory of these empirical practices.employee involvement program. The au-
thors combined the ten items into a scale

The practice of HRM covers a very wide
range of topics, methods, techniques, they
and called "advanced human resource practices," which is the variable measured on the
situations, so many outcomes are available
for study. Is there, however, one outcome,
or
horizontal
axis; it is also the type of explanaset of outcomes, that arguably represent
thevariable typically used as the measure of
tory
major research issue in the field? One answer
HRM practices in a Huselid-type regression
is provided by Boselie, Dietz and Boon (2005:
model. The data reveal a bell-shaped curve
67) who explained that "the study of HRM
but with a significantly skewed right-hand
is, in its broadest sense, concerned withtail.
theThe minority of firms in the left-hand
selections that organizations make from
the
tail
employed few if any formal or tangible
myriad policies, practices, and structures HRM
for practices (e.g., 30% of the respondents
managing employees." There may be other
said their firm had no personnel/HR departand perhaps superior ways to representment)
this ; conversely, the minority of firms in
statement analytically, but one illuminating
the right-hand tail employed many. The bulk
of firms are located somewhere in the
approach is represented in Figure 1 .
middle.
Figure 1 shows two representative HRM
frequency distributions. An HRM frequency disThis pattern of HRM adoption is appartribution shows firms ranked from low to
ently broadly representative, since an alterhigh based on the breadth and depth ofnative data set, plotted in Panel (b), yields
their HRM programs, as measured by somemuch the same picture.5 Rather than a count
metric such as a count of practices or dol-of specific practices, these data show the an-
lars of expenditure. The HRM frequencynual expenditure per employee on HRM.
distribution in Panel (a) comes from data These data were collected for the years
collected in a 1994 national survey of sev-
eral thousand employees and managers

5 Using WERS data, Bryson, Gomez and Kretschmer
(Freeman and Rogers 1999). These people(2005) found a broadly similar bell-shaped distribution
were asked to indicate whether their organiof HRM practices among British firms, although
zation had each of ten different HRM prac-considerably less so in the right-hand tail.
skewed
its central research

question: what level and
2004-2005 by the Bureau of National
Affairs
(BNA, 2006) from several hundred
Amerimix of HRM
practices maximizes firm performance,
why? In this section we procan companies. They show the
same and
inverse
vide a brief
review
of the dominant stream of
U-shaped distribution, again with
the
righttheorizing
and empirical
modeling in manhand tail distinctly skewed to
the right.
In
agement
on these matters as they are princithis sample, the annual HRM
expenditure
pally developed
in the SHRM literature. This
ranged from a low of $36 to a high
of $7,392
also helps
per employee; approximately section
one-half
of puts
thein context our alternative
theory and
empirical model.
firms spent between $475 and
$1,842.
The
Researchers from economics are conmedian was $932.
fronted
certain challenges when engag
These frequency distributions
helpwith
frame
ing
the theoretical
the mission and objectives of
HRM
theory,literature in managemen
on HRM
(and vice versa). According to
particularly in the case of SHRM.
Consistent
with Boselie, Dietz, and Boon's
Lazear
(2005)
anddefiShaw (2007: 110), personnel
economics
draws largely from microeco
nition quoted above, the goal of
HRM theory
and
is to explain why individual nomic
firmstheory
choose
a models in the financ
In management,
particular expenditure level field.
and package
of however, the most important disciplinary
HRM practices. It follows as matter
of logic, influence is psychology
and its business
school offshoot organiza
then, that the firm's choice regarding
HRM
tional
behavior
(OB)
practice and expenditure becomes the cen-(Weber and Kabst
2004; models
Gerhart 2007)
- terra incognita to most
tral explanandum in theoretical
and
the task of these models is (or
economists.
should Accompanying
be) to
psychology is
also
a focus on individual
differences in psyidentify the set of independent
variables
that
influence the firm's choice of HRM. Looked
chological variables (e.g., motivation, cogniat in terms of the frequency distributions intion) that are abstracted (or ignored) in the
Figure 1, the core mission of HRM theorystandard economic model of the rational
actor {homo economicus). Further, most theocan be interpreted as predicting and explainretical work in personnel economics employs
ing where individual firms are located in
these distributions - that is, at a low, meconsiderable mathematics; in management
dium, or high HRM level - along with corol- research on HRM, however, mathematics is
conspicuously absent - even discouraged.6

quality, and implementation of these HRM Management research is also considerably
practices. Other relevant questions include more trans-disciplinary than economic rethe shape of the HRM distribution at a pointsearch and contains numerous diverse theoin time; changes over time in this distribu- retical perspectives. Schler and Jackson
tion; variation in distributions across indus- (2001), for example, listed thirteen sources
tries, nations, and other groupings; and the of theoretical contribution in HRM research
composition of the firm's HRM bundle at (e.g., cybernetics, population ecology, behavioral, organizational, business strategy)
any particular point in the distribution.
and Subramony (2006) discussed four different theoretical perspectives regarding choice
Review of Management Theory and
among HRM practices. Adding further comEmpirical Modeling
lary but more complicated issues of coverage,
plexity is the fact that even the definition of
The HRM frequency distributions help

frame the problem of choice that firms con- 6 We originally submitted this article to a leading Amerifront when they consider investing in people can managementjournal. The associate editor in charge
management practices; in doing so, theseof
the paper declined to send it out for review. Among
distributions also handily summarize the re-other reasons, the editor wrote, "the majority of your
arguments are based on equations. . . . While it is cersearch issue facing HRM scholars at both thetainly acceptable to model one's ideas via equations, it is
theory and empirical level. Although these not an approach that fits well with Journal X." In the
distributions have not been utilized in the
realm of economics, the equations used in this paper
SHRM literature, they nevertheless highlight are considered both few and elementary.
HRM is diverse and unsettled; Pauuwe and
Boselie (2005: 69) observed, for example,
that "there appears to be no consensus on
the nature of HRM."
practices are often associated in economics

and industrial relations with a well-
developed internal labor market (ILM)-
employment system. Boselie, Dietz,
Boon (2005: 73) described this selecti

With due regard given to these hurdles
HRM variables as a "highly managem
and pitfalls, we believe the following highly
standpoint." Other SHRM aut
condensed and synthetic account providescentric
a
reasonable summary statement of the mainhave proposed different lists; in gen
however, the universalistic perspective sh
line of management theorizing on firms'
two propositions. The first is that the p
choice of HRM practices.7 For considerably
more detailed reviews, see American authorspay-off to these HRM practices is grea
such as Becker and Huselid (2006); Allen
they are implemented as a package; the
and Wright (2007); and Lepak and Shaw ond is that on average, more investmen
(2008); and international authors such as
careful selection, training, and empo
Boselie, Dietz, and Boon (2005); Wall and
ment of employees (core functions
Wood (2005); Grimshaw and Rubery (2007); "advanced" HRM practices) increases
and Boxali and Purcell (2008).
panies' performance.
The contingency perspective argues
Theorizing on firms' choice of HRM practices usually proceeds in the context of the best choice among HRM practices is co
three-way typology developed by Delery tional on important contextual fac

(Lepak and Snell 1999). The optimal s
and Doty (1996). They distinguished three

groups of theoretical perspectives: universalistic, contingency, and configurational.8 The
universalistic perspective posits that there
are certain "best practices" that are always a
profitable choice because they universally
improve organizational performance. Pfeffer
(1998), for example, listed seven: employment security, selective hiring of new personnel, self-managed teams/decentralized
decision-making, pay-for-performance, ex-
HRM practices for one firm may be quite

ferent for another, thus implying a firm
timal choice of HRM should follow a "best
fit" rule in which HRM practices are chosen

so they best align with variables such as orga-
nizational size, industry, production technology, workforce/job skills, and state of the
labor market (Datta, Guthrie, and Wright
2005). Several contingent factors receive
particular emphasis. An "external" contintensive training, reduced status differentials,

gency factor is the firm's business strategy,
and extensive information sharing. These meaning that different strategies for competitive advantage in the product market
(e.g., low cost versus superior service) re7 Two caveats are required. First, the SHRM literature
quire different HRM strategies and sets of
we summarize has largely American roots and is mostly
developed in American management journals; some
practices (an idea often labeled "vertical
non-USA researchers have also embraced it, but many fit"). An "internal" contingency factor is the
human capital requirements of the firm's

production process, meaning that optimal
2009). Second, mainline HRM research largely omits
HRM practices will differ considerably bereference to more radical (Edwards 2009) and heterodox perspectives, including critical management stud- tween companies employing, respectively,
ies (Fleetwood and Hesketh 2008) and the labor process workers with low and generic skills versus
paradigm (see Edwards 1979; Thompson and Harley high and specialized skills. In line with a con2007). Labor economics does much the same.
tingency perspective, Lepak and Snell (1999)
8 Some studies identify a fourth perspective, the contextual, in which HRM strategies and practices adapt to fit used permutations of human capital "value"
others have argued that the literature does not generalize to other countries (Brewster 2004; Boxali and Macky
different national-level cultural, social, and political and "uniqueness" to derive four different tycontexts. These can be thought of as additional contin- pologies of best performing HRM practices
gencies. In practice, most SHRM writers make the con-
figurational perspective a component part of the
(called "HRM architectures"); similarly,
Arthur (1994) argued that firms pursuing a

ing the framework to a two-way typology (Proctor low cost strategy are more likely to adopt
universalistic and contingency perspectives, thus reduc2008).
a command and control system of HRM
Theamainline
of management theorizing
practices whereas those pursuing
differenin SHRM
movedatoward a synthesis
tiation strategy are more likely
tohas
adopt
commitment system.9
and integration of these three perspectives
(Boselie, emphaDietz, and Boon 2005; Becker and
The configurational perspective
Huselid 2006)congru, stimulated in part by growing
sizes the roles of complementarity,
skepticism
of a "universalistic-only" position
ence, and synergy as important
influences
(Boxali Although
and Macky 2009). Following Huselid
on a firm's optimal HRM choice.
(1995), this in
integrative
approach maintains
to varying degrees implicit or explicit
the
that therehypotheis a core of "best practice" HRM
other two perspectives, the central
techniques
and policies,
sis here is that the performance
effect
of typically associated
with anon
HPWS-type
employment system,
HRM practices depends critically
assembling the right combination or which
system
in most
of to
pracnearly all organizations rein improved
performance. The exact
tices such that all the separate sults
HRM
elements
(e.g., selection, compensation,specification,
training, delineation,
benand organizaof both the performance and
efits, and employee relations)tional
fit level
together,
HRM variables
remain unsettled, however.
support each other, and develop
the maximum attainable synergy (an idea
For example,
often the
called
specific set of HRM practices variesperforconsiderably across studies,
"horizontal fit"). Here, the potential
in general they are intended
mance effects of HRM choice although
are multiplica-
tive rather than additive, to

implying
include "advanced,"
low "sophisticated," or
returns if all but one or two of the HRM ele-
"high-performance" practices. In some cases
they are measured at the establishment level,

ments fit together, but high returns if all are
but more often they are measured at the

successfully implemented as complete package. Part of the competitive advantage of an
company level. In the large majority of studHPWS is widely attributed to the fact it isies,
a these practices are included as a count
measure (yes if present, no if not) ; in some
complete package that captures the full
range of complementarities and synergies - acases, a practice coverage or practice intensubject that has attracted researchers notsity measure is used (e.g., see Guest et al.
2003) . With regard to performance, both fionly in management but also in economics
nancial outcome variables, such as rate of
(Ichniowski, Shaw, and Prennushi 1997;
Black and Lynch 2001; Laursen and Foss return on assets and Tobin 's q, and interme2003) and industrial relations (McDuffie
diate organizational outcomes, such as turn1995; Appelbaum et al. 2000) . Another close over, productivity, and quit rates, are used.
affiliate of the configurational idea is the Some authors have argued, however, that
large literature on employment systems broader and sometimes subjective measures
(Osterman 1987; Begin 1991; Marsden 1999; of organizational effectiveness should also
Barton, Burton, and Hannan 1999).
be included, such as stakeholders' perceptions of the firm's performance or factors
that measure satisfaction of employees' in9 Practically every one of these ideas (e.g., strategy, com- terests (Godard 2004; Purcell and Kinnie
mitment, competitive advantage, high road versus low
2007).
road) was recognized and discussed in the early literature of industrial relations (Kaufman 2001, 2008). Fur-
ther, Commons (1919) laid out the basic theoretical
rationale for the high performance employment model,
stating that winning employee commitment through
high involvement practices "is valuable because it
The presence and magnitude of these
variables are hypothesized to have a positive

effect on firm performance. Huselid (1995:
668) called this channel of the HRM-firm
performance relationship the "main effect."
A major area of SHRM research, in turn, has
been to identify the causal linkages in the
brings larger profits and lifts the employer somewhat

above the level of competing employers by giving him a
more productive labor force than theirs in proportion
to the wages paid" (p. 26). These antecedents and con- transmission mechanism that runs from adtributions have been widely ignored in the SHRM litervanced HRM practices to higher firm perforature, in part because researchers often inaccurately
characterize IR as dealing only with unions and collec- mance (often called in the SHRM literature
the "black box"). The majority of studies
tive bargaining (e.g., Scarpello 2008).
draw on an amalgam of enhancing

arguments
in manufacturing
from
relative to serthe resource-based view of
firm
(RBV)
,
vicethe
industries
and HPWPs
implemented
as
"abilities, motivation, opportunities"
a system have twice the
theory
performance effect
(AMO), and human capital
as individual
theory
practices.
to support the hypothesized positive
main
effect.10
In terms
of the
HRM frequency distributions in Figure
1, SHRM
researchers who
The contention is that advanced
HRM
practices are performance enhancing
because
emphasize the universalistic
perspective prethey (1) increase employees'
knowledge,
dict that most
firms should be in the righthand tail of the
distribution or should be
skills, and abilities; (2) empower
employees
to act; and (3) motivate moving
employees
in that direction.
to act
The minority who
(Combs et al. 2006:503).
take a "strong contingency" perspective pre
that firms the
will locate
across the HRM
In the integrative SHRMdict
model,
positive main effect is then amplified
or attenufrequency distribution
from "low" to "high,"
ated by contingent anddepending

configurational
on whether their external and
factors. Contingent factors

include
external
internal
"fit" variables
make employees a
and internal variables, such commodity-like
as business strat"hired hand" or a valuable
egy, industry, production
technology,
and
"human
resource" capital
asset (Lewin 2001;
workforce characteristics.Orlitzky
Theseand
contingent
Frenkel 2005). The middle and
variables can have either an additive or mul-
arguably dominant position in SHRM is a
tiplicative effect but, typically, this effect blend

is
of the universalistic and contingency
presumed not so great as to completely offperspectives ("weak contingency"), with the
set the positive main effect (Kaufman 2010b)

.
configurational
perspective common to
The positive main effect is also influenced by
both.11 That is, the middle ground holds
the degree to which the various HRM practhat theory indicates most firms have under-
tices are implemented in a synergistic packinvested in HRM practices and should there-
age. This description of the integrative model

fore move further toward the right-hand
is highly consistent with the results found tail
in of the HRM frequency distribution, cou-
a recent meta-analysis of 92 empirical HRMpled with recognition that a full-blown

firm performance studies by Combs et HPWS
al.
is not well suited for every organiza(2006). The authors found a statistically sigtion. Hence, some modest-to-significant difnificant positive main effect between HRM
ferentiation in HRM systems is required and
practices and firm performance; calculated
a portion of firms can therefore be expected
that a one-standard-deviation increase in the
to locate closer to the middle or, possibly,
use of HPWPs translates, on average, to a even
4.6
the left-hand part of the distribution
(Huselid and Becker 2006). 12
percentage-point increase in gross returnon-assets (ROA)- from 5.1% to 9.7%; and
found evidence of significant contingency
and configurational effects. For example,

HPWPs are nearly twice as performance
11 The distinction between strong and weak contingency
comes from Kaufman (2010b).
12 The hypothesized positive effect of (advanced)
HRM practices on firm performance also comes from

10 RBV argues that HRM practices add value by helping
transform the employees into "valuable, rare, inimitable, and non-substitutable" resources, thus giving firms
a sustainable source of competitive advantage (Barney

1991: 105); Allen and Wright 2007: 89). AMO theory
argues that HRM practices add value by expanding employees' skills and knowledge, incenting employees to
use these for valued organizational outcomes, and giv-
normative concerns among SHRM researchers. It is
widely acknowledged that both the HRM field in universities and the HRM function in industry have long
been regarded as marginal, low status players; a major
motive in SHRM research, therefore, is to change this
perception by demonstrating that HRM is a large and
under-appreciated contributor to firm success. Wright
et al. (2005: 409-10) remarked, for example, "In response to these longstanding and repeated criticisms
ing employees greater opportunities to participate
that HR does not add value to organizations, the past 10

[sic] years has seen a burgeoning of research attempting to demonstrate that progressive HR practices result
ally all strategic HRM research is based."
worry, therefore, that SHRM theory and empirical
(Appelbaum et al. 2000; Boxall and Purcell 2008). Allen

and Wright (2007: 90) claimed that "the resource-based
view has become the guiding paradigm on which virtu-
in higher organizational performance." One must
These predictions are tested

in
single
seem
to aoffer
a sobering but tantalizing
equation regression model verdict
pioneered
by
on firms'
optimal choice of HRM
Huselid (1995), typically usingpractices.

cross-sectional
data. The regression model takes
form:
The the
sobering
part of the message is that it
appears many American firms are seriously
(1) Perf, = 0 ^HRM, + jAj
+ e, in HRM; the tantalizing part
under-investing
is that firms could substantially improve their
where Perfi is a measure of financial or orgafinancial and operational performance by
nizational performance at the ith firm, HRMi
upgrading and expanding their people manis a vector or composite index of HRM pracagement systems. In terms of economic thetices at this firm, X is a vector of contingent
ory and the HRM frequency distributions in
and control variables, and is a randomly
Figure 1, many, and perhaps most, firms are
distributed error term. The maintained hy(apparently) substantially "out of equilibpothesis in most of the SHRM literature is
rium" (away from the optimal level of HRM)
j > 0 (more HRM -> higher performance),
or, alternatively, in an equilibrium that for
although moderated in a positive or negasome market failure reason is significantly
tive direction by one or more of the continless than socially optimal. In either case,
gent variables in the vector X13 Some studies
these firms are leaving a huge amount of
also use equation (1) to test for the configumoney "on the table" and the economy is oprational perspective, typically by including
erating considerably inside its efficiency

interaction terms between separate HRM
frontier.15 The implication is that these firms

practices in the vector HRMif with a hypoth-
should spend more on HRM and thereby
esized positive interactive effect between

migrate further toward the right-hand tail of
them (e.g., see MacDuffie 1995; Delery and
the distribution, leading to a gradual shrinkDoty 1996).
age of the left-hand side and to a parallel inSHRM researchers have discussed at
crease in national productivity and economic
length numerous problematic or difficult-toperformance.
solve theoretical and empirical issues with
These findings, it must be noted, pose a
this regression model (Wright et al. 2005;
potentially significant challenge to economic
Wall and Wood 2005; Gerhart 2007; Purcell
theory, or at least to its competitive core.
and Kinnie 2007) . Alternative specifications
Economists (e.g., Lazear 2000) are predisof the performance and HRM variables have
posed to think that market failures are
already been mentioned; other problems intypically small-to-modest and firms and marclude direction of causality (simultaneous
kets typically adjust fairly quickly to incenor, alternatively, higher performance - >
tives, certainly so in the case of extra-large
more HRM), omitted variables, and conincentives such as those suggested in the
struct validity of certain contingent and
control variables (e.g., measurement and
HRM literature (e.g., a potential near-
doubling in ROA) . Yet extreme dispersion in

specification of alternative business strateHRM practices is evident not only for congies). Nonetheless, the meta-analysis results
temporary firms, as illustrated in Figure 1,
of Combs et al. (2006) - finding a statistibut also for firms going back fifty and even
cally significant positive main effect - one hundred years (Baron, Jennings, and
provides seemingly powerful evidence of Dobbin 1988; Kaufman 2008; 2010c).
the importance of HRM methods in firm

performance.14 In particular, these results
with data on more than 3,200 firms that "the effect of a
one-standard-deviation change in the HR system is
results are systematically biased by the career, identity, 10-20% of a firm's market value."
and status concerns of the researchers.
15 In this spirit, Huselid (1995: 668, emphasis added)
remarked, "the substantial variation in HRM practices

10 Equation ( 1 ) depicts the contingent variables A, in an
additive specification; a multiplicative specificationadopted by domestic firms [relative to an HPWS con-
figuration] . . . suggests that, at least in the near term,

such [above competitive] returns are available for the
14 Further evidence is provided by Becker and Huselid
would include an interactive variable HRMi ~.
taking."
(2006: 907), who concluded after an empirical study
One possibility is that economists

in HRM practices. Though
are the analytical apwrong and a large market failure
proach utilized
or persishere is entirely standard in
tent disequilibrium in facteconomics,
blocksitoptimal
has surprisingly not been earHRM investment. If such is the case, the
lier developed by researchers in the economskepticism that management writers express
ics of personnel (for reviews, see Gunderson
2001; Lazear and Shaw 2007).
toward economic models appears well
grounded. A different and largely opposite
For purposes of modeling, we assume the
possibility is that the large number of "low firm has choice over a range of possible
HRM" firms in the left-hand side of the disHRM practices; the portion that is mandated
tribution is an approximate efficient equilibby government - larger in some countries
than
rium outcome and these managers have
in others - enters our model later as an
correctly gauged the optimal level of HRM exogenous

"shift factor." We further assume
even if it is lamentably "not much." If this isthat the firm's short-run objective is maxithe case, it suggests SHRM theory and emmum financial return, which for simplicity's
pirical methods considerably over-estimate
sake we treat as maximum profit. In economthe profit pay-off to more HRM; it also sug- ics, this assumption is completely standard
gests that the slow and partial uptake of adto the point that it is considered a "given"; in
vanced HRM practices by many companies is
management, however, objections immedinot the fault of the managers who do not ap-ately surface. For example, what about other
preciate and act on the evidence provided
firm goals and other performance outcomes
(Guest et al. 2003)? Likewise, researchers
in modern SHRM research (Rynes, Giluk
and Brown 2007); rather, it reflects the misoutside of economics point to the interests
specified models and normative biases of the
of other stakeholders, such as employees
HRM academics.
and communities, and issues concerning
In the tradition of industrial relations, the

we social legitimacy of business and profit-
hazard the prediction that both sides probamaking (Godard 2004; Boxall 2007).
bly capture important elements of truth andEconomists' responses to these objections
thus reality is probably somewhere in are

thetwo-fold. First, the variable "profit" can
middle. A long-standing IR principle, after
be broadly interpreted to subsume the dolall, is that markets, and particularly labor
lar value of all those intangible or subjective
factors that influence the bottom-line worth
markets, are prone to numerous frictions,
failures, and imperfections (Dunlop 1994);
of an enterprise, at least as long as they are
indeed, from a Coasean/institutional ecoin principle fungible into a dollars and cents
nomics perspective the very existence of equivalent (e.g., the stock price should re-
multi-person firms with an employment relationship represents a form of market failure
flect how much shareholders value the extra
"social legitimacy" created by improved
(Kaufman 2010a). Nevertheless, for the pur- HRM). Where this is not possible, or the exposes of this paper we put on our "econoercise unduly strains credulity, then economists' hats" and critically examine SHRM mists will argue that abstracting from these
theory and empirical methods from the per- other goals is usually a good trade-off since
spective of economic theory. Rather than endoing so promotes more analytical modelgage purely in criticism, however, in what ing (which HRM scholars say they want more
follows we also present an alternative model, of, as quoted above) and most likely does
an alternative estimating equation, and new not alter major predictions and hypotheses.
empirical evidence on the factors that shape Regarding other intermediate organizational
observed HRM frequency distributions.
outcome variables, a virtue of the theory developed here is that it makes intermediate
goals such as turnover reduction and proModeling the Firm's Demand
for HRM Practices
ductivity increase an explicit function of the
quest for maximum profit; a corollary beneWe seek to explain the individual firm'sfit is that it also demonstrates that some
decision regarding the extent of investment
SHRM studies have indeed mis-specified the
performance relationship between

HRM and
to "advanced"
and "sweatshop" to "HP
these variables (explained below).
and make
Finally,
no a priori categorization, as
SHRM studies
do, as to what is likely
we also note that the profit-maximization
assumption formally limits our
performing
model to versus
prihigh-performing.
verge
from it
in the second instance b
vate sector for-profit firms,
although
we
maintain
logical consistency in the
would not be surprised if manytoof
the implithe HRM
practices
must be converted
cations and predictions of the
model
also
common unit of measurement for the
apply to non-profit organizations.
duction
function.
Mirroring the stan
The key analytical innovation
is that
we
the capital (K) input variab

treat HRM practices as a factortreatment
input in of
proheterogeneous
duction. That is, the firm's output
is assumedcollection of buildings
chines,
and
tools typically aggregat
produced by capital, labor, and
HRM
practices. HRM is used, therefore, because it
converting them into a dollar amoun
boosts productivity; however, it also carries a similarly aggregate the diverse HRM
cost since HRM must itself be internally pro- tices into dollars of expenditure. I
duced (e.g., by an HR department) or
bought in external markets (e.g., HR consultants or vendors). Thus, the firm's optimal
expenditure on HRM is determined by the
same marginal-type decision rule found

throughout microeconomics: to maximize
model, therefore, explicit/ tangible
practices and activities, often but not
performed by a personnel/ HRM d
ment, are represented by the HRM ex

ture variable. This variable, in turn, be
the dependent variable in our regr
profit, keep investing in more HRM as long model. Other means of informal or unstrucas the marginal revenue gained exceeds the tured labor coordination, typically performed as a general part of management by
marginal cost incurred and stop when the
two become equal. This basic insight is not employers and line managers, is subsumed
new. In the HRM literature, Jones and Wright as part of total work hours devoted to pro(1992) discussed an economics-based ap- duction (L).16
The model indicates that firms adopt
proach along this line and Kaufman (2004;
2010d) has further developed it. Our contri- HRM because it helps produce more output
bution is to take the model to the next level.
and profit. But how does HRM do this? This
To do so, we begin with the firm's ex- subject is the much debated "black box"
issue in the literature (Becker and Huselid
panded production function:
2006; Boxali and Purcell 2008). We suggest
the following approach, which is not only
(2) Q = f(K, L, HRM),
analytically tractable and insightful but also
of the three links in the causal
where Qis output, Jîs capital, L is inclusive
labor, and
identified
by SHRM writers (knowlHRM is HRM practices. Since thechain
model
is
edge, skills,
intended to explain firms' investment
in abilities; empowered to act;
motivated
to act) .
HRM practices, we must be clear on
how the
HRM construct is defined in equation (2).
The definition of "HRM practice" given by
Wright, Dunford and Snell (2001:

703)
is
16 A large
portion
of the SHRM literature equates formal
and procedures with the activity of HRM,

"those HRM tools used to manage HRM
thepractices
human
giving the subject a distinctly functional perspective
capital pool." We follow this definition
in
(mirrored in HRM textbooks). In reality, HRM is a geone respect but diverge from it in
two
othneric
"people
management" function (Boxali and
ers. We follow it in that we equate Purcell

HRM 2008)
practhat can be conducted with zero formal
practices and no HRM department, as takes place today

tices with formal, tangible, and measureable
when specifian employer personally and informally conducts
activities, policies, methods, or tools
labor management in a small firm (Marlow 2006) or
cally created and used to managewhen
people
in in a large firm a century ago handled
a foreman
organizations. We diverge from it in

the
first
most
aspects
of personnel on an entirely informal basis
(Kaufman
2008, 2010c). Our two-part specification of
instance by including in the variable
HRM
the"simple"
HRM variable captures both aspects.
the entire range of practices - from
on workplace
may increase Q by reThe first revision of equation
(2) is to safety
exaccidents and production
pand the labor term from L ducing
to Le.workplace
This efdowntime.
The direct effect is independent
fectively transforms labor from
a commodity
of of
the "human"
of labor.
input (like a machine or lump
coal) toaspect
a
Thenumber
second channel
human input. The L term is the
of by which additional
influence production is the
persons/ hours of labor; the inputs
termofe HRM
repreindirect referred
HRM effect (the middle term). The
sents what Appelbaum et al. (2000)
to as "effective labor." It is broadly
indirect
defined
effect captures
to
the influence that
include the combined effect of more motivamore HRM practices have on output as they
indirectly change the effective amount of labor,
tion, effort, empowerment, and skill/knowl-
through factors such as improved motivaedge acquisition. If e = 0 (e.g., workers sleep

all day on the job or have zero skills for the
tion, greater work effort, better citizenship
job), then Le = 0 and no output is forthcombehavior, and skills upgrading.17 This causal
ing from the production function; the higher
link occurs when labor becomes a unique
e is, conversely, the more effective labor the
"human" factor input; it is also where this
organization gets from each worker and the

model incorporates the "value creation"
more output is produced.
insights from RBV, AMO and human capital
The second revision makes the amount of
theories.18
effective labor, Le, a function of the level of
HRM expenditure; that is, Le(HRM). The
idea is that more HRM practices may con-
17 In the early years of the personnel/IR field, employertribute to increased effective labor either by
employee cooperation and unity of interest were viewed as
boosting motivation, effort, and opportuniuniversal contributors to more "effective labor"

ties for action (e.g., through an employee
(Kaufman 2003a, 2008), albeit attainable through alt
native HRM configurations rather than one "best pr
involvement program or gain-sharing comtice"
pensation plan) or by increasing workers'
set as in some versions of modern SHRM theo
The early personnel/IR view held, in turn, that co

skills, knowledge and abilities (e.g., through
eration and unity of interest depend critically on usi
more training or an information sharing

HRM to achieve high morale and fair treatment (the f
program).
mer being impossible without the latter). Regard
These revisions lead to the expanded production function in equation (3).
(3) Q = f[K,e(HRM)^L, HRM] .
morale, Hall (1925: 35) remarked, "morale is to the

dividual what temper is to steel." The modern HR
firm performance literature has repackaged mor
into a more expansive and amorphous concept of mo
vation/commitment whereas fairness in SHRM models
and the scholarly conversation has slipped to a duly-
noted but modest-to-peripheral role. SHRM thus mirHere arises an important insight regarding the causal transmission mechanism rors
be-its opposite - neoclassical labor economics - in
that both slight equity and fairness as determinants of
tween HRM and firm performance. Our

efficient production. Also similar is their preoccupation
model suggests that HRM influences firm

with maximizing firm performance and returns to
shareholders (capital) while giving little to no attenperformance through two distinct channels.
tion, importance, and counter-balancing consideration

The first, which we label the direct HRM effect,
to the independent and sometimes conflicting interrepresents the independent contribution

ests of employees (labor). Byway of contrast, early IR -
that more units of an HRM practice (repreincluding its prominent management wing - strongly
sented by the right-hand term in the producemphasized that fairness is crucial to economic performance; fairness, in turn, requires that all competing intion function) have on output, holding
terests get voice, due process, and reasonable outcomes.
constant the amount of labor and capital
This is a stakeholder (not shareholder) model of the
services. For example, more expenditurefirm

on and employment relationship.
employee selection, such as greater invest18 If there is indeed systematic underinvestment
in HRM
SHRM theory implies, a principal reason is probably

ment in hiring tests, personal interviews, as
and
that the value created by HRM is often largely intangipsychological assessment, increase output
ble (e.g., morale), difficult to measure, and only real-
independent of any change in the quantity of
ized in future years, thus causing it to be under-capitalized
labor (by better matching of people to jobs,

in the firm's profit maximization calculation (Slichter
1919; Kaufman 2008: 240-41).
and so on). Alternatively, extra expenditure
captures
the is
direct
effect (the effect of
The choice problem for the
firm
toHRM
se-
more HRMthat
on Q holding
lect the level of HRM practices
best constant L) and
first termthis,
captures the
the indirect HRM effect
achieves its profit objective.theGiven
(the effect of (4):
more HRM on Q as it creates
firm's challenge is to solve equation
more effective labor) . If labor were a com-
(4) n = P^flK, e(HRM)*L, HRM]

- V^HRM - W*L
modity (i.e., inanimate factor input), the

term d(e)/d(HRM) becomes a constant and
falls out of the first order condition, leaving

only
Equation 4 demonstrates that profit
(II)the direct effect. If only the direct effect
were present, human resource management

would not be substantively different from
Revenue is PQ with the production funcoperations management or from an early
tion in equation (3) substituted for Q As-
is the difference between revenue and cost.
version of scientific management.

suming capital is fixed in the short-run, there
are two elements of variable cost: labor cost
The right-hand side of equation (5) is the
unit price of HRM services, V. In other words,

and the cost of HRM practices. Assuming
equation (5) shows the marginal decisionthe cost of labor per unit is the wage W (for
rule
earlier cited: the firm should keep inall employees, including managers, and investing
additional money in HRM practices
cluding benefits and other such costs), total
as long as the extra revenue created exceeds
labor cost is WL. The HRM practices also
have an explicit cost, denoted by V, sincethe extra cost; when the two become equal
the optimal level of HRM practices has been

they are themselves produced with capital
and labor. The total cost of HRM is, there-
reached.19 Importantly, when profit is maxi-
cally viewed as a component of a firm's HRM
leads to lower performance - not higher.
fore, VHRM. Although the wage Wis generi- mized further HRM investment necessarily
Our contention is that this model is not
package, the choice problem considered

here is the optimal level of management simply an exercise in empty formalization
"manufactured" HRM, so Wand Vare sepa- but, rather, is a vehicle that yields a number
rately distinguished. Just as the price of labor of new and useful insights.20 Consider the
(W) is assumed a "given" for this exercise
(set by the market), so too is the cost of pur- 19 This model is extended in several ways in Kaufman
chasing/producing extra HRM practices
(2010d); for example, the composite HRM variable is
( V) . This simplification makes the model far disaggregated into i individual HRM practices (i =
more tractable, does not materially affect staffing, training, compensation, etc.), the profitthe results, and broadly accords with reality maximizing configuration of HRM practices is derived
(horizontal fit), complementarities among HRM prac(e.g., a firm can obtain additional trainers, tices are introduced, and the horizontal fit dimension
job evaluations, payroll processing, and so of HRM strategy is modeled.
on, at a going market price) .
*For example, an ott-cited dehmtion ot SHKM is "tne
The optimal level of HRM is determined pattern of human resource deployments and activities
intended to enable an organization to achieve its goals"
by differentiating equation (4) with respect (Wright and McMahan 1992: 298). Our model
to HRM and solving for the first order condi- reveals this is simply a verbal re-statement of the stan-
tion. This is done in equation (5):
3Q de ] L
an =p{de dHRM) =y
dHRM dQ
+ dHRM
The left-hand side of the first-order con-
dard profit maximization equation in economics, illustrated by equations (4) and (5) . This is a surface insight;
a deeper insight emerges by then asking what

separates the two fields (i.e., labor economics and
SHRM) if they share the same objective function and

choice problem? One answer (Kaufman 2010a) is that
the fields are two branches of a larger umbrella field
called industrial/employment relations (aka: institutional labor economics) in which the former branch
theorizes allocation and coordination of labor resources
via markets and prices and the latter theorizes the same
dition (the bracketed term) is the marginal thing but through organizations and command/
administration with Commons/ Coase transaction cost
revenue product (MRP) of HRM practices.
It is composed of two parts: the second term theory determining the boundary. (Modern personnel
Onesome
implication
of this model, accordfollowing. Figure 1 illustrates that
firms
ingly, is the
following: each firm's place in
invest little in HRM practices whereas
others
invest in an intermediate level and others in
the HRM frequency distribution is determined by a comparison of benefits versus

a high level. Our model explains this variacosts of additional investment in HRM praction thus: the management of each firm,
tices. For some firms, this calculation yields a
using equation (5), compares the extra prozero level of investment in HRM practices
ductivity and revenue generated by using an
whereas for others it yields an HPWS. Aladditional unit of HRM practice in production with the extra cost incurred. Some firms,
though the mainline of the HRM-firm performance literature predicts that "more
given their size, technology of production,
HRM is better" for profitability, our model
skill, demographic characteristics of the
suggests that this is unlikely to be true outworkforce, and other such factors (spelled
side an unrealistic scenario where all (or
out in more detail below), find that profits
most) firms realize continuous marginal
are maximized with near-zero HRM expenditure. This might be an "externalized" orgains in profitability from further investment
in HRM (Kaufman 2010b).
"market"-type employment system, as described by Delery and Doty (1996, Table 1), A second implication concerns the
definition of "best practice" HRM. In this
in which demand and supply set pay rates,
framework, one cannot make a universalistic
motivate employees (through threat of unstatement that best practice HRM is comemployment) , and provide new recruits and
posed of some particular set of HRM practraining opportunities. Others find that
tices, or that best practice is represented by
profits are maximized with an intermediate
an HRM-intensive employment system lolevel, and yet others find, given their size,
technology of production, and other intercated toward the right-hand tail of the frenal and external characteristics, that a high
quency distribution, such as an HPWS.
level of HRM practices maximizes profit. ExRather, in this framework "best practice" has
amples include an HPWS, high involvement,
only one meaning and metric - that is, the
and internal employment system (again see
HRM practice (or set of practices) that leads
Delery and Doty 1996, Table 1).
to the most profit (highest financial performance) and greatest probability of long-run
economics does not theorize a dual coordination econ-
survival for the company. Thus, in some situ-
ations an HPWS may be best practice whereas

omy in a substantive sense, nor does it give management
in of
others a low-road sweatshop employment
coordination autonomous status, since market forces
competition and the Invisible Hand still determine

system may be best practice (Lewin 2001).
management choices and HRM practices/structures
One hundred years ago in America, "best
(Lazear 2000).) Illustrative of this duality, through
practice"
meant next-to-zero formal HRM
the 1950s personnel/ HRM was widely regarded as
practices, just as it does in many parts of less"applied labor economics" and subsumed with human
countries today (Kaufman 2008;
relations as the management wing of industrial developed
relations (Kaufman 2000). Our model implies that manage2010c; Tessema and Soeters 2006).
ment is one side of the economics coin; history does as
A third revisionist implication concerns
well. Bossard and Dewhurst (1931) state that for busithe
predicted effect of more HRM on firm
ness subjects like management, "economics ... is
the
performance.
Taking profitability as the perfoundation or basic subject to be studied" (p. 325)
and
many management historians (e.g., Holbert 1976:formance
31)
measure, most SHRM studies preconsider Henry Towne's article, "The Engineer as dict
Econa positive effect. If it is positive, however,
omist" (1886), to have founded the field of manage-
this means that firms can increase profit by
ment (the first professional managers were often
investing in more HRM, which is to say they

engineers by training) . As the IR and personnel/HRM
fields began to separate after the 1960s, behavioral

have not yet reached the optimal equilibscientists replaced economists and SHRM replaced
rium level of HRM predicted by equation
IR regarding the external/strategic dimension of em-
(5). Economic theory suggests, therefore,
ployment relations. Not surprisingly, modern HRM

that
education is mostly "applied psychology/ OB" and
no-unless market failure is large and persistent, the positive sign hypothesized in
ticeably slights the role and contribution of economics
(Scarpello 2008).
SHRM - if it empirically exists - is most likely
a short-run relationship that tends

tofailure
diminmarket
or obstacle to equilibrium.
ish and even disappear over time

as competiThe "one-eighth"
rule advanced by Pfeffe
tion and share-holder pressure
leadisfirms
to
(1998)
one approach
to this challenge;23
capture the unexploited profit
theopportunity
most popular explanation, however, rest
contained in HPWS practices.21
with
Competition
RBV theory. The idea here is that HR
thus implies that extra economic
helps
profit
turn from
employees into rare, inimitabl
HRM investment should decline and ultiand valuable human capital assets (akin to
mately go to zero. Marginal reasoning
and
non-contestable
form of differentiated prodthe law of diminishing returns implies
uct)
the
, the benefits of which cannot be easil
same. In other words, firms invest in duplicated
more
and competed away (Allen an
HRM as long as the marginal revenue
Wright
gain
2007). This explanation does ratiooutweighs the marginal cost, but duenalize
to di-a positive sign on the HRM variable
minishing returns the marginal gainHowever,
falls
it is not obvious why in practic
(e.g., consider the marginal returnother
fromfirms cannot easily and in short orde
sending employees to more and moreduplicate
train("contest") any particular system
ing classes) and the marginal cost rises
ofuntil
HRM practices of the formal or tangibl
kindis(most HRM practices are neither pro
a point of zero marginal profit gain
reached.22 A persistent positive HRM prietary
effect nor particularly complicated; other
wise,
HRM would be more of a skilled and
is not per se ruled out but then has to be
exhigh-status
profession) as used in HRM-firm
plained, as noted earlier, by some form
of
performance models (Priem and Butler
2001). The positive HRM effect, therefore,

21 Huselid (1995) noted this implication of economic
may
well come from other unobserved or
theory in a sentence (p. 668) but then dismissed its
omitted
practical significance on the grounds that the large
esti- factors, such as a rent to superior
mated positive effect of advanced HRM practices
on
management
ability (e.g., differential suc-
firm performance, coupled with the low level

of adcess
at implementing
vanced HRM adoption at many firms, suggests that
HRM and creating/
maintaining positive employment relations) .
large profit gains remain available for capture. Becker

and Huselid (2006: 905) reiterated this theme Or,
but as
ex-Cappelli and Neumark (2001) found
inmarket
a particularly thorough study, it may not
plicitly tied the existence of large quasi-rents to
exist at all.
failure, citing lack of knowledge, lack of managerial
competence, and failure to implement. However, A
SHRM
fourth insight of this model is to point
theory seems caught in a contradiction since appeal to
out
what economic relationship SHRM
substantial market failure is at odds with the emphasis
researchers
given in study after study to increased competition
in the
are estimating with a Huselid-
economy (and thus the asserted need to adopt

high
type
regression model. The standard HRM-
performance HRM practices) .

firm performance regression model is, in
22 The numerical value of the coefficient j depends on
effect, an attempt to estimate the first-order
the metric of the performance measure used as the decondition given by equation (5). That is, the
pendent variable and the specification of the regression
regression
coefficient j in equation (1)
equation. If the metric is percentage rate of return
on
capital, in long-run competitive equilibrium with
homo- AYl/AHRM, which is exactly what
measures
geneous firms all firms earn the same "break-even"

is solved for in the first order condition. We
(normal) rate of return, say 10%. In this case, the
HRM
know
further that to find the maximum
variable varies across firms but the dependent variable

is constant, causing t to take a value of zero. Ifpoint
the de-of the profit function II = f(HRM'
which
is a slightly different way to look at
pendent variable is instead (say) dollars of profit,
the
regression equation should be specified in non-linear
the first-order condition in equation (5) and
terms, such as Perft = 0 + ^HRM, + faHRM2 + 3X, +
e, in order to capture the (predicted) parabolic shape
in the functional relationship between profit and HRM
The "one-eighth" rule (Pfeffer 1998: 29) means that
intensity (profit rises, reaches a peak, and then23
declines
as a function of HRM intensity) . In this case, one-half
the pre-of managers do not know that an HPWS inperformance, inertia keeps one-half of these
dicted sign on the coefficient is x > 0 but 2creases
< 0. One
criticism of the SRHM literature, therefore, is
thatimplementing
it
from
a HPWS, and one-half of these do
not successfully
implement it. Thus, 100% of firms benwidely neglects the idea of (eventual) diminishing
reefit from an HPWS but only 12.5% implement it. Also
turns and thus omits the term faHRM2 (Kaufman
see Pfeffer (2007).
2010b).
essentially what a Huselid-type

Figure
regression
2. The HRM Demand Curve
model represents, requires that one differ-
entiate the function and set the first deriva-
Price (V)
tive equal to zero. But the first derivative,

ATI/AHRM, is j and in long-run competitive
equilibrium is necessarily zero, as noted
above. For j to be an unrestricted positive

number, as explicitly or implicitly assumed
in many SHRM studies, one must argue that
(1) firms do not maximize profits (to a great
extent); (2) for some reason, firms are pre-
Vi .--'E^'^/C'^C
vented from getting anywhere close to maximum profit (per the large estimated shortfall
in ROA); and (3) the marginal revenue gain

from more HRM always exceeds the mar- 0 HRM3 HRM1 HRM2 HRM
ginal cost.
A fifth insight concerns potential misspecification in SHRM studies of intermediate outcome variables, such as productivity,
sloping portion (not shown here for simp
quits, and turnover. These studies (e.g.,
ity of exposition), but eventually will sl
Huselid 1995) typically hypothesize that if
downward, given the operation of the law
more HRM leads to lower quits and turnover
diminishing returns. The common sens
or higher productivity, then this too reprethe downward slope is that beyond s
sents an improvement in firm performance.
point, additional investment in HRM pr
Such is not necessarily the case, however, at
tices, such as additional sophistication in
least if profit is the index of performance.
lection tests, have a successively sma
For example, more HRM expenditure may
positive effect on productivity and revenue
improve productivity, but the cost increase
Assuming the price of HRM practices
of the additional HRM may far outweigh
constant Vv the profit-maximizing leve
any resulting efficiency gains, thus reducing
HRM practices is HRMj (point A) . It i
profit and performance (Cappelli and
this point that the equilibrium condition
Neumark 2001).
equation (5) is satisfied. If this falls anywh
to the left, the MRP of HRM exceeds the
The HRM Demand Curve and
HRM Demand Function
marginal cost and the firm adds to profit by
expanding expenditure on HRM practices;
if it falls anywhere to the right, the opposite
holds
Extensions of the model yield further
in-true.
sights and implications, as well as a new tool
Figure 2 shows that a firm's use of HRM
for empirical HRM analysis. The place
to
practices
follows the law of demand, just as
start is a derivation of the HRM demand
its use of other factor inputs does. Thus, a
curve.
rise in the price of an HRM activity from V1 to
This curve depicts the relationship

V2 causes abemovement up the HRM demand
tween the price of HRM (V) andcurve

the Dj
firm's
and a decline in quantity demanded
quantity demanded of HRM practices
from HRMj(exto HRM3 (point A to point B).
pressed in dollars of expenditure),
holding
A firm's
demand for HRM practices is also
all other factors constant. Such a curve
is de-by all those variables that shift
influenced
HRM demand curve. These variables

picted in Figure 2 as Dr This curve the
is derived
affect one
by plotting the marginal revenuemust
product
ofof the two determinants of
HRM. In other words, the MRP signifies

the marginal revenue product:
the HRM input's
extra dollars of revenue gained from
investthe marginal
physical product (the extra
ing in one more unit of HRM practices.
The or the marginal revenue
output produced)
MRP schedule could initially havefrom
an this
upward
extra production (or both) . Theory
suggests a number of these frequency

shift variables;
distribution of HRM practice
others are more a matter of common sense
Thus, the left-hand tail of the HRM fre-
observation or empirical determination quency

(dedistribution is described by the firms
that have a zero to small demand for HRM
scribed below) .
Before proceeding further, it is useful(e.g.,
to demand curves to the left of D3); the
repackage equation (5) into a more tractacenter of the distribution is given by the mable format. This is done in equation (6).jority of firms that have intermediate HRM
(6) HRMl=f(Ql,Wl,Vl,Xl).
demand curves (in a band around Dj); and
the skewed part of the right-hand tail is given
by the relatively small number firms that

have a very high demand for HRM (demand
Equation (6), in effect, inverts the profit
maximization equation in equation (5) and

curves scattered far to the right of D2) .
expresses the demand for HRM in the ith This model also explains changes in the
firm as a function of the level of the firm's
HRM frequency distribution across time and
countries. At the turn of the twentieth cenoutput, the prices of factor inputs, and a host
the HRM frequency distribution was

of other independent variables capturedtury,
in
highly compressed and centered very close
the vector X. Equation (6) is called the HRM
to the vertical axis (Kaufman 2008, 2010c).
demand function. It parallels the labor deFor example, in 1902 the world's largest
mand function, which is a staple of labor
economics (Hamermesh 1993). Holdingcompany,
all
the United States Steel Corporation, employed 160,000 people but used
other variables constant, changing the level
practically zero formal HRM practices (e.g.,
of V in equation (6) causes a movement
no2.hiring office, employment application
along the HRM demand curve Dj in Figure
or job/wage schedule). The reason is
Holding V constant and changing one of form,
the
that nearly all firms were using a highly exother variables in the demand function (e.g.,
ternalized labor management system and
larger scale of output) shifts the HRM dethus had near-zero HRM demand curves.24
mand curve to the right (D2) or to the left
Over the ensuing decades, however, the
(D3). At a constant price of Vp a rightward
HRM demand curves of many firms shifted
shift of the firm's demand for HRM practices
successively to the right due to changes in
leads to an increase in expenditure on HRM
technology, unionization, legal
practices from HRMj to HRM2 (point A production
to
regulation of employment, and other such
point C); a leftward shift reduces expendifactors, causing the mean and variance of
ture on HRM practices from HRMj to HRM3
the HRM frequency distribution to increase
(point A to point E) .
The HRM demand curve and demand
as well. Variation in HRM demand curves
alsoexexplains different HRM frequency

function model provides an interesting
distributions
among countries, such as beplanation for the shape of the HRM
fretween
quency distribution at a point in time
and the United States, France and India.
for changes in it over time. At a point in
time, each firm has particular valuesShift
of the
Factors of HRM Demand
variables V, W, and X, and, inserting these
Our model implies that variation in HRM
into the demand function yields its optimal
demand
level of HRM expenditures on specific
prac- curves explains the variation in
tices. Plotting these equilibrium values traces
out the HRM frequency distribution,

24 Thisas
does
not mean these firms practiced zero human

management, since in every multi-person organization then and now a "boss" must in some way coplot the position of firms' HRM demand
ordinate
curves in Figure 2 and, for a given price
(e.g., and manage the acquisition and utilization of
the labor input. Rather, it means HRM is in this case an
Vj), determine the same distribution of equiundifferentiated part of line management (subsumed
librium values of the HRM practice expendiin the variable L) and involves few if any formal and
ture variable. In effect, the distribution
of
systematized/scientific
personnel methods (the HRM
variable).
HRM demand curves maps out an identical
shown in Figure 1. Alternatively, one

can
resource
relationship
is uniformly found in
firm-level adoption rates This
of HRM
practices.
To give the model greaterpirical
explanatory
studies (Boselie,
con- Dietz, and Boo
A theoretical
rationale for this rel
tent, it is next necessary to2005).
identify
the spetionship
is factors)
that larger in
sized organizations
cific independent variables
(shift
the demand function that cause this variatail both more people to manage and grea
distance
tion. The first two we discuss (Q and W)
are between executives and workers,
use of more staff and resources
explicitly identified in equation (6)necessitating
and
come directly out of microeconomic producto coordinate production efficiently. Altion theory; the remainder are subsumed
though
in totalHKM rises with firm size, due to
the vector X and are suggested by theeconomies
ecoof scale (arising from spreading
nomics, IR, and HRM literatures. The list
fixed investment and set-up cost over a larger
that follows is suggestive, not definitive; isemployee base) expenditure on HRM per
tailored to the American context; and is inemployee probably declines on average for
tended to help motivate the specification of many firms (Brewster et al. 2006).
the empirical HRM input demand function
Wage Rate. The second variable in the
estimated in the next section of the paper. HRM demand function in equation (6) is
Certain important determinants of HRM de- the wage rate W. The wage may be either a
mand, such as employment laws and social/ substitute or complement for HRM practices
political institutions, are omitted because (Ichniowski, Shaw and Prennushi 1997). In
they do not significantly vary among firms in
a national cross-section and thus do not
the former case, firms may use a higher Win
lieu of formal HRM practices. An example of

explain variation in demand curves.25
this is efficiency wage theory, in which an
Firm Size. The demand for HRM practices
employer pays a higher-than-market wage to
should increase with firm size, measured
by who are thus motivated to selfemployees,
level of output ( Q) or level of employment
enforce higher work effort. Firms can then
reduce direct HRM control devices, such as
(jointly determined by Qand Win equation
(6) and therefore not explicitly shown).26
supervision and time clocks. In this case, a
higher wage would shift the HRM demand
curve to the left. The opposite would occur
25 Another potentially interesting omitted variable
is the
in a situation
in which W and HRM practices
firm's profit level. A reasonable hypothesis is that firms

are complements.
with above-normal profit invest part of it in more HRM
In high performance
work systems,
for example, a high wage and
in pursuit of other sub-goals, such as enhanced
em-
high
level
ployee satisfaction and loyalty. We have no data
on firm-
of HRM go together. One reason
level profit, however, and so we cannot testisthis.

that The
an HPWS requires a unitarist employissues of simultaneity and reverse causality ment
are examrelationship and paying a high wage
ined in Wright et al. (2005). On one hand, our model is
creates higher employee commitment and
not free of simultaneity concerns in empirical estimaloyalty
and removes a source of potentially
tion (e.g., between the unionism independent
variable
and HRM expenditure dependent variable);

on the distributive bargaining (tacit or
disruptive
other hand, it probably is less widespread or severe

than
formal).
in the standard SHRM regression model (e.g., the deProduction Technology. Internalization of
pendent performance variable could well be related to
is encouraged by production
most or all of the HRM independent variables employment
as well as
that are more capital intensive;

some of the exogenous control variables, technologies
such as
unionism).
complex; feature greater worker interdepen26 The neoclassical production theory utilized
for this
dences
(e.g., team forms of production);
model assumes as a "given" that firms exist and largely
allow greater
takes market structure and the size distribution and
of firms
room for discretionary ef-
More capital intensive and complex

as a datum in the analysis, yielding in turn fort.
an a primi
fixed HRM frequency distribution and pattern
of em- make employee selection more
technologies
ployment systems. Institutional economics, utilizing
difficultthe
and important, and turnover more
transaction cost concept of Commons and Coase, endoexpensive.
In addition, more extensive intergenizes the number, size, and structure of firms and
dependencies
in production increase the
thereby endogenizes the underlying shape of the HRM
need to maintain
and promote effective emfrequency distribution and the pattern and structure
of
employment systems (Kaufman 2010b).
ployee coordination and cooperation, and
greater room for discretionary

greaterwork
demandeffort
for HRM practices, per the
heightens the importanceimplications
of maintaining
of the resource-based view of
and promoting employee commitment
the firm (Lepak and
and
Snell 1999). This conmorale (Lepak and Snell 1999;
sideration
Appelbaum
may link certain workforce characteristics to HRM demand. Past research
et. al. 2000).
Industry and Organizationalhas

Characteristics.
shown, for example, that white men have
A variety of industry and

greater
organizational
specific OJT, so one may predict that
organizations
with more women and minor
characteristics potentially affect
the demand
ity employees
for HRM, although the predicted
signshave
area smaller demand fo
HRM
practices. Another
relevant variable i
not always clear-cut (Datta,
Guthrie,
and
educational
attainment
of A
the firm's workWright 2005; Sun, Aryee, and
Law
2007).
force. Education
is part of company's stock
case in point concerns differences
in HRM
of human capital
and SHRM theory predicts
demand between goods-producing
(e.g.,
that firms with a more
educated workforce
manufacturing) and service-producing
orgain broad
greater HRM
to get higher producnizations. Heterogeneity in invest
these
catthis valuable
asset (Becker and
egories is possibly so large tivity
as tofrom
preclude
a
Huselid2006).
meaningful prediction; alternatively,
one
Economic/
Market Conditions. Firms operatcould reasonably argue that
on average,
ing in more
stablepracproduct markets and ecomanufacturing firms use more
HRM
nomic environments
have a greater incentive
tices to boost productivity indirectly
through
to adopt ILMs and disformal HRM practices

devices such as employee involvement,
(Orlitzky
and Frenkel
pute resolution, and training.
Likewise,
it is 2005; Ordiz and
Ferndez 2005).
ILMs involve greater emreasonable to expect that for-profit
organizaployee investment
expense, transform labor
tions may have a different demand
for HRM
into a quasi-fixed
cost, and introduce greater
than not-for-profit organizations;
governorganizational
These conditions bement organizations are possibly
alsorigidity.
a discomeBegin
progressively
less economic in the face
tinct entity. According to
(1991),
of greater
volatility
of sales and employment
not-for-profit organizations
make
greater
use of cultural and social norms as control
and shorter product life-cycles. ILMs and extensive
HRM practices are also promoted
and motivational devices, thus suggesting
a
when labor markets remain at or close to full

lower demand for formal HRM practices
employment. Not only does full employment
(other things being equal). Government
increase the pressure to carefully select, demay be hypothesized to have a greater develop, and retain employees (due to scarcity
of qualified labor in the external market) ,
greater bureaucracy and formal procedures;
but it also reduces the ability of firms to use
however, formal procedures are not necessarily expensive per se and expenditure levels the threat of unemployment as an effective
and less costly motivation or discipline deon HRM practices (e.g., training, employee
vice (thus enhancing the indirect HRM efrelations) may actually be more intensive
in some for-profit firms (Brewster et al.fect). HRM demand, therefore, should be
2006).
positively associated with the employee turnover rate (e.g., necessitating more expendiWorkforce/Training Characteristics. Internalmand for HRM to the extent that it entails
ture on recruitment, selection, and training)

ization of employment and demand for
and negatively associated with the rate of

HRM practices will also be greater in firms
unemployment.
whose production involves greater specific
on-the-job training (OJT) (Grimshaw and Unionization. The presence of a union in a
Rubery 2007) . Specific OJT creates a form offirm, or the threat of unionization, can have
asset specificity, thus raising market transac-contradictory effects on HRM demand simi-
larly, the effect may well differ across nation cost. Work systems that provide more
tional industrial relations systems. A union
opportunity for workers to develop and
apply new knowledge for improvements inin the American context endeavors to
processes and products will also have anegotiate more formalized, structured, and
standardized employment
management
variation
in HRM practices but also a tool for
practices, thus suggesting aempirically
positiveanalyzing
relationthis issue. Using a
ship (Verma 2007). Similarly,

the
threat
ofto estimate
unique data
source,
we proceed
union organization may motivate
firms
an HRM demand
function to
for a sample of
several hundred
American
firms.
upgrade their HRM function.
But
unions
also perform certain HRM functions (recruitment through a hiring hall, dispute
resolution through a grievance system), thus

Data Set, Estimating Equation,
relieving the employer ofand
making
these exVariable Definitions
penditures; they also resist certain HRM practices, such as incentive pay The
and
performance
majority
of the information we use
appraisal. A negative relationship,
was collected therefore,
by the Bureau of National Afis also possible (Brewster et
2006).
fairsal.
(BNA)
for its 2005/2006 HR Department
Strategic Role of the HRM
Function.
Firms
Benchmarks
and Analysis
reports. Earlier years
differ on the degree to which
were notlabor
availableand
to us the
and parts of the surmanagement of labor is a vey
strategic
determiinstrument
were also different. The
nant of profitability. Whenbreadth
laborand
has
a small
depth
of information on firms'
HRM activities
in the BNA dataset
are, to our
effect on firm performance,
management
is
likely to invest little in HRM;
conversely,
knowledge,
the most extensive available in a
when labor has a large effect
the
bottom
public on
source
in America.
Nonetheless, the
line, management is likely to
considerdatainvest
source also
has limitations, particularly
able resources in HRM. One

way
HRM
compared
to data
sets inavailable in other
vestment takes place is creation

staffing
countries and
(e.g., the
Workplace Employment
of a personnel/HRM function.
The (WERS)
more
Relations Survey
in the United
strategically important labor

andthat
labor
manKingdom)
are more
detailed and longitudinalthat
(see Guest
et HRM
al. 2003).
agement are, the more likely
the
function is included in the formulation and
These data are useful because they come
execution of both HRM strategy and the
from a comprehensive survey of HR departoverall business strategy. Firms with strategiments from a diverse group of firms. In addi-
cally involved HRM departments, therefore,

tion to providing information on the
should be linked to higher levels of HRM exdependent variable in our model (HRM ex-
penditures (Lepak, Bartol, and Erhardtpenditures), the BNA survey data also in-
2005; Brewster et al. 2006).

clude information on many of the
Employee Relations Philosophy. Company
independent (shift) variables. Other indeowners and top executives differ in their phipendent variables were developed from alternative data sources (Bureau of Labor
losophies and attitudes toward employees
and labor management practices. This fac-Statistics, Equal Opportunity Commission),
tor most closely corresponds to the "taste"
matched to the BNA firms through their
variable in the traditional microeconomic
three-digit (sometimes two-digit) North
theory of demand. Quite apart from profit

American Industry Classification System
considerations, some owners and executives(NAICS) code. The combined data set comprefer an employee-oriented approach and
prised 614 observations; after deleting
invest more in HRM in order to achieve betobservations with missing data the sample
ter workforce treatment and esprit de corps was reduced to 381. The data on HRM
whereas others view employees as expend- expenditure reported by BNA come from
able commodities and hence give short-shrift
mix of organizational levels - sometimes a
to HRM (Boxali 2007).
entire firm and at other times an autonomous sub-division or individual establish-
Estimating an HRM Demand Function
ment (plant). The single largest industry
concentration is Services.
Our theoretical model offers not only inThe estimating equation, given in equasights and implications regarding inter-firm tion (7), is a linear version of the HRM
Table 1. Definition and Predicted Signs of Independen

(Dependent Variable: HRM Expenditure per Emplo
Theoretical Predicted
Independent Variables Shift Factor Sign Source

Full-time employees (log) Firm Size - BNA
Annual worker earnings (log) Wage ? BNA
vi
/i
Production
_...
vi Non-labor i i operating costs per worker i (log) /i x + BNA
Non-profit private sector organization (dummy variable = 1 ) - BNA

[for-profit sector = omitted group]
Government/public sector organization (dummy variable = 1) Industry/ ? BNA

[for-profit sector = omitted group] Organizational
Service/non-manufacturing industry (dummy variable = 1) - BNA
[goods producing = omitted group]
HR Department has partial influence on strategy
(dummy variable = 1) [Low or Zero influence

on strategy = omitted group]
HR Department has substantial influence on strategy
(dummy variable = 1) [Low or Zero influence S t " R 1 / + BNA
on strategy = omitted group] S Imploy^ t R 1
HR Department has high influence on strategy , ., ,
/j u, i?rT ry . strategy n Philosophy , ., ,
(dummy /j variable u, = 1) [Low or Zero ry influence . n
on strategy = omitted group]
HR department in central corporate office (dummy variable = 1 ) + BNA

Firm's strategic focus is on employee satisfaction and morale + BNA
(dummy variable = 1 )
Unionized percent of workforce Unionization ? BNA

Female percent of workforce, by industry and state - EEOC
Non-white percent of workforce, by industry and state Workforce/ - EEOC
Bachelors degree percent of workforce, by industry Training + BLS
White collar percent of workforce, by industry + EEOC
Unemployment rate by state/industry, by state and industry - BLS
Cyclical Instability of Employment (log of coefficient of variation Economic/ _
employment,1991-2005), by industry and state Market
Separation rate per 100 workers, by industry + BLS

with
HR function.
demand function in equation
(6)the
but
with Deflating total HRM
expenditures
by the number of employees
capital and labor inputs explicitly
included:
gives a metric that is far more easily compared across firms and is not so greatly influ^ = a + lQi+2Li+3tfi
+4^+5X,+ei.
The dependent variable is the log of HRM
expenditures per employee (or "per capita")
in firm i. HRM expenditures include "costs
of labor, materials and equipment, overhead,
and administration incurred by HR in performing its core function and duties" (BNA
2006: 114), where "HR" is defined broadly to
include expenditures attributed to both the
human resource department per se and to
other managers and programs associated
enced by differences in absolute firm size.

Similarly, expressing the dependent variable
in log form transforms the coefficients on
the independent variables into estimates of
marginal percentage change (elasticities) .
The use of an expenditure measure of
HRM, it should be noted, has an important

advantage over the "practice count" specification used in many HRM-firm performance
regressions. A measure of HRM usage
formed by the addition of discrete practices

may contain substantial measurement error;
for example, two firms both report usage of
analysis,
course,
also require addiemployee involvement methods
butofone
is would
a
tional variables.
low-level suggestion box system
while the
The HRM
demand function is estimated
other is a high-level team system.
A "count"
with
ordinary
least squares (OLS). Alternaspecification may well weight
both
equally
tive better
estimation
methods were performed
whereas an expenditure measure
captures the underling difference
but, in
as we
scale
report
and
below, the results did not
materially change. We also disaggregate

scope of usage across the two firms.27
expenditure
We were able to develop datathe
for
nineteendata into nine HRM subindependent variables in the
functions,
regression
as specified in the BNA survey,
and re-estimate
the demand function for
model. The variable definitions,
hypothe-
each. The
sub-functions
are recruitsized sign, and source are given
in nine
Table
1
(the hypothesized signs come ment,
fromtraining,
the discompensation, benefits a
cussion of the previous section);
summary
ministration,
employee relations, externa
statistics are given in Tablerelations,
2. We discuss
personnel management, heal
these variables in more detail in the next secand safety activities, and strategic plannin
tion on results. Note here, however, that sev- This disaggregation provides insight on th
eral variables contained in the theoretical
extent to which the independent variable
model are not included in the regression
have consistent, stable relationships acros
HRM sub-functions.
equation, such as the cost of capital (r)different
and
cost of HRM practices (V) . Data on each are
unavailable. Omission of the latter is not
Empirical Results
critical on the reasonable assumption that
the per capita cost of producing/buying
Table 3 reports the coefficient estimates
and standard errors for all of the model
HRM practices (e.g., an employee training
class, a job evaluation) is most likelyspecifications.28
relaThe first column shows the
estimated coefficients for the "total" or
tively uniform across firms, particularly
within an industry (and thus captured in
the
"aggregate"
HRM demand function (using
industry dummies). Data for several potentotal firm level HRM expenditures); coltially important control variables are
not 2-10 show the results for disaggreumns
available in the BNA data set; hence, wegated
con- HRM demand functions. We initially
structed measures from other sources, but
they are measured at a higher level of aggre-
gation (e.g., by three-digit industry). 28

The
We explored several alternative specifications and
positive aspect is that these control variables
conducted a number of robustness checks (reported in
Miller
are less likely to be endogenously related
to2008) . We re-estimated the equations using quantile regression, for example, to determine whether the
the firm's HRM expenditures; the negative
size of the coefficients varies over the interval range of
aspect is greater measurement error. the
Eviindependent variables (for example, if the "threat
dently, there are also other independent

effect"
variables that one could well think should
of unions on HRM expenditure is not linear
across different levels of union density) . We found little
variation. We also tested for division bias (potentially

also go in the estimated HRM demand func-
introduced when HRM expenditure is deflated by emtion; our problem is that they are notployment
given that employment is also an indepen-
attainable from the BNA data set and are dif-
ficult to acquire elsewhere. A longitudinal
dent variable) by re-estimating the equations with HRM

expenditure as the dependent variable and employment
size (ES) and ES-squared as independent variables. The

number of statistically significant independent vari-
ables is identical both ways. We also tested for simulta-
27 Another large problem in the standard HRM-firm neity between employment and HRM expenditure,
performance regression model is that effectiveness of since in equation (4) the firm is assumed to solve for
HRM implementation is a significant but largely unob- both together. To do so, we re-estimated the equations
servable variable intervening between practices and using two-stage least squares and instrumented the
performance. Since our model explains HRM expendi- firm's employment level using the state/industry level
ture, not firm performance, the issue of effectiveness of employment. The results did not change. Finally, we
(or management quality) does not affect the estimated also included regional dummies, but these had no dis-
cause-effect relationships.
cernible effect.
Table 2. Summary Statistics

Dependent Variable Mean Std. Dev. Minimum Maximum
Total Per Capita HRM Expenditure 1321.37 1217.45 36.00 7392.00

Per Capita HRM Expenditure on Recruitment 204.19 264.04 3.93 1840.00
Per Capita HRM Expenditure on Training 157.73 210.64 1.20 1629.34
Per Capita HRM Expenditure on Compensation 340.97 512.94 1.80 4435.20
Per Capita HRM Expenditure on Benefits Admin. 205.30 277.79 2.00 2179.20
Per
Per
Per
Per
Per
Capita HRM Expenditure on Employee Reis. 1 17.04 155.64 0.40 1272.25

Capita HRM Expenditure on External Reis. 48.93 75.23 0.36 693.00
Capita HRM Expenditure on Personnel Mgt. 68.05 99.01 0.36 1017.80
Capita HRM Expenditure on Health & Safety 86.89 134.59 0.40 1055.00
Capita HRM Expenditure on Strategic Plan. 72.33 118.46 1.59 819.65
Independent Variable Mean Std. Dev. Minimum Maximum
Full-time employees (log) 6.74 1.58 3.78 13.82

Annual earnings or "wage" (log) 10.49 1.35 4.47 15.20
Non-labor operating costs per worker (log) 9.41 2.45 1.35 15.49
Non-profit private sector organization
(dummy variable = 1 ) [for-profit sector = omitted group] 0.04 0.19 0.00 1.00
Government/public sector organization
(dummy variable = 1) [for-profit sector = omitted group] 0.35 0.48 0.00 1.00
Service/ non-manufacturing industry
(dummy variable = 1) [goods producing = omitted group] 0.39 0.49 0.00 1.00
HR Department has partial influence on strategy

(dummy variable = 1 ) [Low or Zero influence on
strategy = omitted group] 0.22 0.42 0.00 1.00
HR Department has substantial influence on strategy

HR Department has high influence on strategy

HR department in central corporate office
(dummy variable = 1) 0.31 0.14 0.00 1.00
Firm's strategic focus is on employee satisfaction
and morale (dummy variable = 1) 0.35 0.48 0.00 1.00
Unionized percent of workforce, by industry 10.02 18.95 0.00 65.00

Female percent of workforce, by industry and state 51.25 17.67 0.54 82.20
Non-white percent of workforce, by industry and state 33.35 9.56 14.50 99.32
Bachelors degree percent of workforce, by industry 19.01 9.19 8.00 33.00
White collar percent of workforce, by industry 62.73 21.46 10.92 99.27
Unemployment rate by state/industry, by industry and state 5.06 1.03 2.70 10.40
Cyclical Instability of Employment (log of coefficient of
variation employment, 1991-2005), by industry and state -2.16 0.50 -4.47 -1.62
Separation rate per 100, by industry 37.35 17.65 15.20 75.50
p-values reported in the last row of Table

included a dummy variable in the equations
implies that the estimated deman
to separate the observations obtained This
from,
have explanatory power. In add
respectively, the years 2004 and 2005,functions
but it
tion,
a
standard
Chow test reveals that th
was statistically insignificant and we therecoefficient values on the independent var
fore dropped it.
ables are statistically different for seven
All ten regression equations are statistinine HRM sub-function regressions. Th
cally significant from zero, indicated the
by the
* Ef * ?* # CO "X* 00 00 ^* '^s- O^ '""^ ^^ '""*v j'*""v "v ^s 'îs> ^ ^ *^
^ ^ * '5 aim f-t o co 01 x co Ttô co o r^^H *- co x co t^ -1 2 .
^ ^> TS S ifi tJ< l> CO O CO fi X ^ CD i- i ^00 CO t^ (N ( O1 l>
^1S ^ P ^ P P. ^ ^ P ^ P ^ ""! ^ c^cvi co w "~! ^ ^ *""! 5S
c/5 o ^^ p
1
; ^ t _ ! ^
^ ^ -}f lO * <O 00
H ^v ^-v l> ^v lfi ^^ ^-v 00 ^^ ^-v ^-v
o^-tfv^ CD <<fJ> COiO COCD ( W i- ii- i X OCO 00 <> fi i- i CO
^j, c^^1 coto jico cd i- " ^ con xx x cdx% oi oo rîfi
qc^ ^P ^P P. ^^ coco c^ . cî '"I5^ P0^ ^""I '"!'"!
> ^^
"**
31
JS X- ^ ifr ^ O CO ^-v <*"> ^~v ^- s s- s /- v ^-v 4f /-- v
-Ngf h N 00 >O X (N h i- i (D ^fX 1>O1 fi r^ i- (X CD ^f fi
O O>P TfTf1 tO fi fiCO CO fi CJ5C31 XCJ1 OiCO fiO O l> CMCO
^^^^ ^.O -1 P ^^ '"^ P0^ i-i Ol <-i Ol OIOI -< <>I^h

|
| w
-- w
! w

^ d
^ ! w
w-w
w
2^ S^*cd"i-hXV rr O)^ (N-^ ^<^ IM^v ^ ^l>^^
j-vS.S Ttî-H "fCD <MI> Oil- ( fiCO Tt<l> TH CMf X fiCO rH IO

^ ^ S "p fi fi^ t CO I>X ^X CO fi X !> i-hI> i- i 31 CMCD
^ - "51 -S co oi ^^ P0^ P0^ P0^ P5^ I~!csf ^ ""! P1^
c*c ' ' '
C/5
g-f-^^-X-^v*^^.
gSo ^CO4f<Ji^Tt^ ^v ^vCD^-*^^*^-^*^-v .^*^-^
G^-2'S OO Ol> CD^H r-HCO CO COOI>fi OiCO OCO ^ XO
nZ,-,C3 -^f "^ fi XCO CO CM i- i O"^1 XCO CJ1CO ^<O OCO
g"S ^P ^P PP P^. OCO P0^. ^^ COOi ^CM i-HrH W H
t*3 a; o ' * ' ' ' '

c"2S *V*XVOS ^-v ^-v*>-vCD^-Ôi^-v ^v^^^X^
^ ^> 'S "^CD CO t^ I>I> i- i fi X OiX CMX TfOi fi CO cOc
tr> ^g i- tCO CMTf -^CM CDO1 ( t 60 fi fi ih Oi CO * fi i- i
^-^"0 ^P ^P PP ^^ cocM >;O pcî pf--1 ocn en h P1"1
Q^ ' d
O
o
Q Ôi-X-CDifi-^1^^^- ^^*^-vC31^-vO1^v ^ ^^^^v
^^ Q <^ fi^H-- ii- itôi(M01^cDC0XC0tÎ>I>C0X

^- S cD^cDcDt^CO^CCMCOcDCO^HtDiîfiifiifiX^
v- ^P '"'P PP . ^ "~i ^ P^ P0^ P0^ ^^ 0^ -1 t *" '
g" d< ' ' ' dd d dd d
fcyQ
^_ -~ #cS *P WO tD/*^ W^ ^-vCO/^vCM^^ >^^ ^^*^-v
r^ ^_ S ^^r-HCDX^-iCO^cDOiXr^CMCO^CinCr-iOicD
O- "2 CO CNTfi I>O i- i fi fi^f OtF i- i C31 ^X I>X COCO i- i
? co o peo o cooi -; P1"1 f~l'~! I~!'~l ^t1"1
^ _ d dd d d dd dd o h1 d
"S
*
r;
#
-^
*
*
^-v
co
^-v
i-
*
t
&
#rf #W COCO Ôz-sO/îfirNCOrv ^ W /-v * ^
- v Jg X Oi fi CO I>t^ i- i Oit^X OcO h O) ^ O 00
C'j v -5 i- i Tft ififi fiCO Ôl X>>Oi cDX 00 W i- ii- i ^ h fi OIOI
^^ ^^^TtHpOfpOfpOICOIr-HOlpi- ITJH^H
S ,/d>^c^c!>cô^d>o ^^ ^ ^^ vw d ^^ < W d> W
C^l^^ ^^ ^ ^^ >W|V_^ vw ^^ W l W >W
^2" i- ico xco j>oj oi> fi oT Tt< oT co ? - i ^ ^^ co" 0100s
OrOg ^P 'IP PP OO CO-- I "^ "- ' i- IrH Oi- I r-Hr-H <- " C^
a, _^^d d d o d o o o
fe o a t -a - I
l U { 1 ! !jl
<oôooi-<oooigmoo com .-. io cm

& . oo ^ o co co x ^-.T-Hj><Mi>a5^în^-v
rN^2 COtJh cOi^ CO Ol h CO CM m I%iO CO CO O Ol CO "*f ,- i
S " S OO OO OO CO 1% OO CM CD O CO O ifl CM CM ,- rf i
^J, v I"H ^ ^f ^f ^ io CM
cog do do do d cm do do do do d
I
'
Eôoî-iGsTt<n> cm oT ^v
^ O CO O iO O l> iO x-v CM O !> O Q5 /- v CD ^^ ,- i ^
q^ iS <^^ ^ ^ O 00 CO Qi tO ^ CO CO Oi CO CO Qi O^ O CO co ^h
^_ i^J S*1 O O O O O O c75 Qi O O Ô CO Ô iO O^ CO iO : <vr

Q C/5 O O O O O O i""H CD O O O O O CM O CM CO Ô UN
[g
pS
OO4<OO
^ v^ v_^ ' " '
pS
^ v^
i v_^ I ' I " I I '
^j
g
OOJ>l>CT5l>
^hI>CDO^X
O500
^^
.-vOO^T^CMO^
x-v
-^8 1? H Ot^ CO00 Tt^CM Hifl O 00 O> 00 XJ> r-i M cO

0 Ot^P OO OO OO lOOO OO COm ^<J5 rt^CM CM ^ Q5
'w|2^ P. PP PP h co pp ^ ^ P*"! '"I5^ ^ co <M
oocôo
^2 comai^ cm ^.
SSi- iiCCDCM'- iCM /-s 00 Ol fl Ji ^ ^-v __

r-vF-S O r^ OO O) CO^ mCM CM1> 00 CO ^m îO t- i ^ __ 1>
^ S"S OO OO O*-i 1>CO i- ( O iCI> O Ol O Ol CO C7i|>
T3
Wl$|"2 ^> ^ î ^ ^ ^ Pi ^ ^ ^ ^t &* &* COCM Tt^ T - '

toc do od do TfCM do do do do d
3 SS O"' m m 00 CD ^ O "^ CO ^-v
iSo cocDâicoai /^hoo co W/^W/-> o_

G^-^'-S H CO COCO GM 00 i- ir-H iOn CMO Orf ^ O ^ o !j
sS.-,e oo oo oo O) oo oco --ho h <n co v| ^
S"^ ^l ^ ^t î ^l ^ CMCO O O PP P5^ P0^ ^
i<i OS do od do cmcm_ d '
O
u
'u
V
Oh
O
C+H
j- Ol P * ^ O f # O i- ^v *
^2 j- g OlOCOOiOt^'X-^Ôl^t^CD^^v*^^ 00
^ <ii.'N rttcO i- i CO CO ^ I>r-i iT)^ Ol H OCO 00 00 IO io 00 00
^ ^ s^ oo i- i-ho oo oo o oo coin i>xo i> oi r^ ^
^^-e <î <î & & &. &. ^P ^ ^ î ^ *. i"1. P1"! ^ 2 ^ ^
^^ do do, * ^ ^ " ' ^^ ^
MOtOOW^*
CMT^
_J
- NI* O> Old Ho7r iCDOlCDCOOOCOCOiOOOcO >
^* sjS O O O O O O CM 00 O O CM CO iO O iO ^P CO -H ^h "^ 1)
^ ^ OOOOOOCMiOpOpOCMCMl>CMCOQQCO "^
1/3
P Ol o O 00 O W CO^^-^ ,_ *
Sj>OlCO^XCM ^vt^COÎ> ^*^v ,_ ^ ja ^fr
^
SS OCMCMiDCMr^r-HOO^rft^tÔ^XCDX1^^ ja
O "2 OOOOOOmtÔOCO^r^Ôi^cDOco T>
^ pqqqpppoqppppr-j-jCMf-jiricM^1 g
^
dododocMi-HOodododod
s CMCO^COr- lOOOOli-^ÔÔOOlCOCMOOCM- i-^ w
cj
-'S
oooooo^cor-HOi-Hio^tînoocM^
&
cj
^ S oooooo^cor-HOi-Hio^tînooc
ci dooodocMCModdoooodo *
$ I - w ! w , w - w , w w ,u a;
fi
i-
- v * ^ oo'Ho'o^cocovoovcoî>ôo^t-co^ 3
^ - v O l OOOOpp^ThppOp-H-HCM^^ÔO 3 g^^J
^dd do do cm^ do do do ooor-H ~ ~^
j , - ~ ~ ~ ~ ^ ^ ~ c_
,
-o
25
g g s s s| s ^H i -b
I I I 1 PI jf*! j II
indicates that the shift factors of HRM de-
capital and other non-labor inputs find it admand have a different size of quantitative vantageous
efto "leverage" or "safeguard" this
fect across sub-functional areas.
investment by also investing extra in HRM in
Column ( 1 ) of Table 3 shows that nineorder

of
to have a high-productivity and highmorale work force (the direct and indirect
the nineteen independent variables have staHRM effects).
tistically significant (p < .10 or lower) regression coefficients. Two of these variables - the
A particular virtue of the BNA survey is
log of employment and the log of average
that it contains three questions that bear
annual worker earnings (a proxy for the
on the strategic orientation and involvement
wage) - are found to be statistically signifiof the company's HR function. The first
cant in all ten of the model specifications.
strategy-related question in the survey con-
Employment, as measured by the average

cerns the level of strategic involvement of the
number of fulltime employees, has a negaHRM function, with five answers ranging
tive relationship with the firms' expenditures

from "no involvement" to "high involve-
on HRM practices. The estimated coefficient

ment." Following SHRM theory, we hypoth-
indicates that a 1 % increase in the size of a
esize that a firm with an HR function that is
firm's work force is associated with an ap- more strategically involved makes greater
proximate 0.3% decrease (on average) in per per capita HRM expenditures, particularly
capita HRM expenditures. The marginal ef- in those firms that cite a high level of involve-
fect is similar across all nine sub-functions.
ment. This idea is captured by the three
We interpret this finding to indicate thatdummy variables that represent, respectively,

firms realize scale economies (decreasing
"partial," "substantial," and "high" involvement of the HRM function. However, conunit costs) in provision of aggregate and in-
dividual HRM services.
trary to expectations, but congruent with the

The second variable that is statistically sig-empirical findings in a number of other
nificant in all ten equations is annualstudies (discussed in Becker and Huselid

2006) , this variable at all levels of involveemployee earnings, that is, annual compensation per employee (the firm's total annual
ment is statistically insignificant across nine
payroll divided by employment). The esti-of the ten equations (the exception is emmated coefficients are positive in all equaployee relations) . One inference is that HRM
tions, indicating that firms paying higher
strategy has no discernible effect on total per
annual earnings also provide more HRM
capita HRM expenditure, counter to the
services, other things being equal. We interpredictions of most SHRM contingency
pret this finding to indicate that the level ofmodels. A second possibility is that HRM
compensation is a complement with other asstrategy affects HRM expenditure but that
pects of HRM expenditure, suggesting that

variation in HRM strategy is itself largely capfirms with an HRM-intensive employment
tured by variation in the other independent
system (such as an HPWS) also provide
variables, suggesting that strategy is not a
higher pay to create synergy and other protrue independent variable but more of an
ductivity advantages.
intervening variable, itself explained by vari-
The next variable represents the capitalous variables external and internal to firms
intensity of the firms' production process,(Kaufman 2010d). The fact that the emmeasured by the non-labor operating costs
ployee relations sub-function equation is the
of the firm per employee (i.e., the capital/
only one with a significant positive sign
labor ratio). The coefficients are positive in
raises the interesting possibility that emall equations and statistically significant in
ployee relations, among all HRM functional
all but two. This finding indicates that firmspractice areas, is the one most at the center
with more capital-intensive production proof SHRM (suggesting, in turn, the strategic
cesses utilize, other things being equal, importance

a
of fairness and the indirect
effect).
greater amount of HRM services per capita.
This result may be interpreted as revealing The second strategy-related question conthat those firms spending large amounts on
cerns the reporting level of the chief HRM
executive. We hypothesize that

per
capita variation with respect to
show
discernible
HRM expenditures are higher in

firms
whose
union
status.30
One could well expect that
HRM executive reports directly

for to
the the
Employee
CEO,Relations and Safety and
Health
rather than, say, a vice-president
ofsub-functions
finance, this variable would be
on the presumption that thispositive,
indicates
whichthat
it is in both cases, but not significantly
so.
employees are a larger strategic
concern.
This variable is also not statistically
The different
survey also categorized firms into
different
broad industries/sectors: manufrom zero, with the exception of
the benefits
facturing,
non-manufacturing or service,
administration equation. Again,
therefore,
strategy - or at least these measures
government,
- seems
and non-profit. We treated
manufacturing
as the excluded category and
not related to HRM expenditures
across
firms.
created dummy variables for the other three.
The third strategy-related question concerns the major performance criterion on
In the aggregate HRM equation, per capita

HRM expenditures are lower in services (rel-
which the HRM function is evaluated. This
ative to manufacturing) and, using a ten-
significance test, also lower in the

variable may also reflect the employee percent
rela-
sector but not in the non-profit

tions philosophy of the firm and itsgovernment
top
sector.
management. If the performance goal
is At the sub-function level, per capita
expenditure on Compensation and Benefits

cost-containment, for example, we hypotheAdministration
size that less is spent on per capita HRM
is lower in Service sector
firms and, for Compensation, also lower in

whereas if it is employee morale and satisfacfirms (p < .10). Recruitment extion, then a larger amount is spent. To non-profit
test
penditure per employee is higher in nonthis, we enter a dummy variable for those
respondents who chose the criterionprofit
goal firms (p < .10), however.
We also included a number of control
"employee morale and satisfaction."29 Our
variables
hypothesis is broadly supported. The perfor- assembled from other sources and
matched
them to the BNA data set through
mance variable has statistical significance
of
either detailed industry or industry/state
varying degrees in six regressions, including
designations. These results must be viewed
the Total Expenditure equation. Interestas suggestive, given the lack of firm-level
ingly, the coefficient is negative and signifidata. Of these variables, the one that showed
cant in the Compensation sub-function
the most frequent and statistically significant
regression. This result may be a statistical
to HRM expenditures was the
quirk; alternatively, it is congruent with relationship
the
employee turnover rate (separations per 100

prediction of satisfaction/hygiene theory
by industry). It is positive and sigthat firms find non-wage measures toworkers,
be a
more effective means to boost morale and
nificant in the aggregate regression and in
satisfaction.
The next variable drawn from the BNA
four of the sub-function regressions. This re-
sult is consistent with the hypothesis that
firms with a higher turnover rate spend more

survey is the percentage of employees in the
onisHRM per capita, other things being

reporting unit represented by a union. It
not statistically different from zero in allequal.

ten We would particularly expect this rela-
for the Recruitment sub-function,

regressions, indicating that in this datationship
set
per capita expenditure on HRM does which
not is the case (for p < .10).
A related variable included to capture cyclical volatility in production and sales over
time is the log of the coefficient of variation
29 The question in the BNA survey allows respondents to

circle two of the five possible answers; thus using a in
set industry
of
employment (from 1991-2005).
(say) four categorical dummy variables is precluded
since some observations take more than one value. The

30 We
results for all three strategy variables are problematic
to reran the regressions but omitted the industry
dummies and also interacted the union and industry
the extent that the firm's HRM expenditure level influvariables,
but in neither case found a statistically signifienced the survey respondent's choice in answering
the
cant effect.
firm's strategic involvement level.
tion
as a center of research
A plausible hypothesis is that
industries
with attention and
then develops
a formal
model capable of exunstable sales and production
would
have
less developed internal labor
plaining
markets
this distribution.
and The model treats
therefore less formal and extensive HRM
HRM as a factor input into production and
programs and expenditures. These suggests

indus- that firm-level differences in the
marginal
revenues and costs of HRM practries might, however, need to spend more
on
tices of
lead to systematic differences in HRM
HRM for recruitment and training. None
the estimated coefficients are statistically

adoption
sigand expenditure. The transmis-
sionAdmechanism (black box) between HRM

nificant, with the exception of Benefits
ministration (p < .10). The unemployment
and firm performance is also modeled, with
rate, measured at the state/industrydirect
level,and indirect effects distinguished. We
also has no discernible effect.
developed these insights formally in terms of
The aggregate regression reveals that

an HRM demand curve and HRM input de-
firms in industries with a greater proportion

mand function. None of these concepts and
of female employees and employees withideas

a
has heretofore been formally precollege education also have a higher per
sented by other researchers in the managecapita HRM expenditure, although the coefment or economics of personnel literatures.
ficient for college education is only weakly On the empirical side, this paper offers a
significant and the result for proportion fenew HRM estimating equation - the HRM
male is counter to our hypothesis. The latter
demand function - that expressly makes
may reflect measurement error. Few significhoice of HRM practices (and expenditures)
cant effects, however, are found in the sub-the dependent variable. It is the first major
function regressions. The proportion of the
alternative to the Huselid-type regression
workforce composed of white-collar employmodel that has for the past fifteen years
ees also has no detectable influence, except
dominated HRM empirical research. Furfor the Recruitment function.
ther, we offer substantive reasons why the
Conclusion
conventional HRM-firm performance regression model, and the hypothesized
"more HRM - > higher performance" predic-
In a survey of HRM research written more

tion, are subject to potentially serious mis-
than a decade ago, Guest (1997: 263) argued

specification. Finally, we also obtain a unique
that the field still required "a theory about
data source on HRM practice expenditures
HRM, a theory about performance and
a
at several
hundred American companies and
theory about how they are linked." Weuse
do it to estimate an HRM input demand
not claim to have provided complete answers
function. The regression results demonstrate
to these three theoretical challenges; we
do firms' demand for HRM is indeed systhat
claim, however, to have revealed serious
tematically linked to a variety of economic,
weaknesses in the answers provided by technological, organizational, and manageSHRM researchers and to have advanced an
ment characteristics. Examples include firm
innovative economics-based model that
size, level of wages, female proportion of the
yields new tools and insights for advancing

workforce, industrial sector, and HRM perthis research program. The theory andformance
emgoal. These results, in turn, help
pirical strategy have admitted limitations
explain the position of firms in the HRM freand the model may seem too simplistic
or
quency
distribution and the shape of this
distribution.
rationally calculative to many SHRM
re-
searchers; nonetheless, we believe these

Our hope is that both the theoretical
shortcomings are more than outweighed byand empirical innovations open a new door
the greater generality of the approach, widefor fruitful work in modern HRM and exrange of behavior explained, and numerouspand the dialogue between HRM researchhypotheses generated.
ers in economics and management. We
With regard to theory, the paper presents hope the economics portion of the paper
the notion of an HRM frequency distribu-brings to management researchers greater
and causes
of potentially larg
appreciation for the fruitfulness
of formal
market
failure
economic models, important
economic
con-and disequilibriu
pose
of industrial relations is to
cepts such as equilibrium and
competition,
integrate
these disparate viewpo
and the general economic way
of thinking.
the
process, meld theories of m
Similarly, we hope our in-depth
presentation
theories
of
organizations into a w
of SHRM theory and empirical
work
motigreater
than the sum of the par
vates economists to give more
consideration
to this extensive literature, a to
richer
model
of
have
modestly
pushed forwar
and,and
in so
doing, leave the
firms and management, the ect
human
disstronger
than when we arrived.
cretionary aspects of labor, and
the existence
REFERENCES
Technology on Productivity." Review of Economics and

Allen, Matthew, and Patrick Wright. 2007. "Strategic
Statistics,
Management and HRM." In Peter Boxali,
John Vol. 83, No. 3, pp. 434-45.
Bosehe, Paul,
Purcell, and Patrick Wright, eds., Oxford Handbook
ofGraham Dietz, and Corme boon, U. WVb.
"Commonalities
and Contradictions in HRM and PerHuman Resource Management, pp. 88-107. Oxford,
UK:
formance Research." Human Resource Management
Oxford University Press.
Journal,
Vol. 15, No. 3, pp. 67-94.
Appelbaum, Eileen, Thomas Bailey, Peter Berg,
and
Bossard, James, andj. Frederick Dewhurst. 1931. UniverArne L. Kalleberg. 2000. Manufacturing Advantage:
sity
Education for Business. Philadelphia: University of
Why High-Performance Work Systems Pay Off
Ithaca:
Pennsylvania Press.
Cornell University Press.
Boxali,
Peter. 2007. "The Goals of HRM." In Peter
Barney, Jay. 1991. "Firm Resources and Sustained
ComBoxali,
John Purcell, and Patrick Wright, eds., Oxford
petitive Advantage." Journal of Management,
Vol.17,
No. 1, pp. 99-120.
International Handbook of Human Resource Management,
pp. 48-67.
Oxford: Oxford University Press.
Baron, James, Jennings, Paul Devereaux and
Frank
Dobbin. 1988. "Mission Control? The Development of
"Human
Personnel Systems in U.S. Industry." American Sociologi-
Resource Managemen
and Significance." In Peter Boxal

Patrick
Wright, eds., Oxford Inter
Bartel, Ann 2004. "Human Resource Management
and
Human
Resource Management, p
Organizational Performance: Evidence from
Retail
cal Review, Vol. 53, pp. 497-514.
University Press.
Banking." Industrial and Labor Relations Review, Oxford
Vol. 57,
No. 2, pp. 181-203.

Barton, James, Burton, Diane, and Michael source
Hannan. Management, 2nd ed. New
Macmillan.
1999. "The Road Not Taken: Origins of Employment
Systems in Emerging Companies." In Glenn Carroll
and David Teece, eds., Firms, Markets, and Hierarchies,

on High-Performance Work Syste
High-Involvement Stream." Hum

pp. 428-64. New York: Oxford University Press.
mentjournal, Vol. 19, No. 1, pp. 3
Brewster, Chris. 2004. European rerspectives on
Human Resource Practices, Quit Rates, and Sales
batt, Kosemary. zvvz. Managing customer services:
Human Resource Management." Human Resource

Growth." Academy of Management Journal, Vol. 45,
Management Review, Vol. 14, No. 4, pp. 365-82.
No. 3, pp. 587-97.
Becker, Brian, and Mark Huselid. 2006. "Strategic
Ommeren. 2006. "What Determines the Size of The
Human Resources Management: Where Do We Go
HR Function? A Cross-National Analysis." Human
From Here?" Journal of Management, Vol. 32, No. 6,
Resource Management, Vol. 45, No. 1, pp. 3-21.
Bryson, Alex, Raphael Gomez, and Tobias Kretchmer.
2005. "Catching a Wave: The Adoption of Voice and
High-Commitment Workplace Practices in Britain,
1984-1998." CEP Discussion Paper No. 676. London
in Personnel and Human Resource Management, Vol. 2,
Centre for Economic Performance, London School
Kenneth Rowland and Gerald Ferris, eds., pp. 261-97.
pp. 898-925.
Beer, Michael, and Burt Spector. 1984. Human Resources Management: The Integration of Industrial
Relations and Organizational Behavior." In Research
Stamford: JAI Press.

Begin, James. 1991. Strategic Employment Policy: An Orga-
nizational Systems Perspective. Englewood Cliffs, NJ:
Prentice-Hall.
of Economics.
Bureau of National Affairs. 2006. HR Department Bench-
marks and Analysis 2005-2006. Washington, D.C.:
Bureau of National Affairs.
Cappelli, Peter, and David Neumark. 2001. "Do 'HighBlack, Susan, and Lisa Lynch. 2001. "How to Compete:
The Impact of Workplace Practices and Information Performance' Work Practices Improve Establishment-
Level Outcomes? Industrial and Labor

Relations
Review,
Godard,
John.
2004. "A Critical Asse
Vol. 54, No. 4, pp. 737-75.

Performance Paradigm." British J
Combs, James, Youngmei Liu, Angela
Hall, and
Relations,
Vol.David
42, No. 2, pp. 349-78
Ketchen. 2006. "Do High Performance
Grimshaw,
Work Practices
Damin, and Jill Ruber
Matter? A Meta-analysis of Their Effects
and HRM."
on OrganizaIn P. Boxali, J. Purcell,
tional Performance." Personnel Psychology,

Oxford Handbook
Vol. 59, of Human Reso
pp. 88-107. Oxford, UK: Oxford U
Commons, John. 1919. Industrial Goodwill New York:
Guest, David. 1997. "Human Res
McGraw-Hill.
and Performance: A Review and
No. 3. pp. 501-28.
Datta, Deepak. James Guthrie, and Patrick Wright.

International Journal of Human Re
2005. "Human Resource Management and Labor Pro-Vol.8, No. 3, pp. 263-76.
ductivity: Does Industry Matter?" Academy of Managementjournal, Vol. 48, No. 1, pp. 135-145.
Sheehan. 2003. "Human Reso
Delaney, John, and John Godard. 2001. "An IndustrialCorporate Performance in the
Relations Perspective on the High-Performance Para-Industrial Relations, Vol. 41 (Ju
digm." Human Resource Management Review, Vol. 11
Gunderson, Morley. 2001. "E
(Winter), pp. 395-430.
and Human Resource Managem
Management Review, Vol. 11 (W
Delery, John, and Doty, D. Harold. 1996. "Modes of
Theorizing in Strategic Human Resource
Hall, E.K. 1925. "Executive Initiative and the Formula-
Management: Tests of Universalistic, Contingency,

tion of Sound Personnel Policies." In Henry Metcalf,
and Configurational Performance Predictions."ed., Linking Science and Industry. Baltimore: Williams
and Wilkins.
Academy of Management Journal, Vol. 39, No. 4,
Hamermesh, Daniel. 1993. Labor Demand. Princeton:
pp. 802-35.
Dulebohn, James, Gerald Ferris, and James Stodd.
Princeton University Press.
1995. "The History and Evolution of Human
Holbert, Hayward. 1976. A History of Professional Management in American Industry. New York: Arno.
Resource Management." In Gerald Ferris, Sherman
Rosen, and Donald Barnum, eds., Handbook of Human Huselid, Mark. 1995. "The Impact of Human Resource
Resource Management, pp. 19-41. Cambridge, MA:
Management Practices on Turnover, Productivity, and
Blackwell.
Corporate Financial Performance." Academy of ManDunlop, John. 1994. "Organizations and Human Re-agement Journal, Vol. 38, No. 3, pp. 635-672.
sources: Internal and External Markets." In Clark Kerr
Ichniowski, Casey, Kathryn Shaw, and Giovanna
and Paul Staudohar, eds., Labor Economics and Indus- Prennushi. 1997. "The Effects of Human Resource
trial Relations: Markets and Institutions, pp. 375-400.Management Practices on Productivity: A Study of
Cambridge: Harvard University Press.
Steel Finishing Lines." American Economic Review, Vol.
Edwards, Richard. 1979. Contested Terrain: The Transfor-87, No. 3, pp. 291-313.
mation of Work in the Twentieth Century. New York: Basic
Jones, Gareth, and Patrick Wright. 1992. "An Economic
Books.
Approach to Conceptualizing the Utility of Human
Edwards, Paul. 2009. "The Employment Perspective
Resource Management Practices." In Gerald Ferris
and Kenneth Rowland, eds., Research in Personnel and
in HRM." In John Story, Patrick Wright, and Paul
Edwards, eds., The Routledge Companion to Strategic

Human Resource Management, Vol. 10, pp. 271-300.
Human Resource Management, pp. 40-51. London:
Greenwich: JAI Press.
Routledge.
Kaufman, Bruce. 1993. The Origins and Evolution of the
Ferris, Gerald, Angela Hall, M. Todd Royle, and Joseph
Field of Industrial Relations in the United States. Ithaca,
Martocchio. 2004. "Theoretical Development in the

Field of Human Resources Management: Issues and
Challenges for the Future." Organizational Analysis,
Vol. 12, No. 3, pp. 231-54.
Fleetwood, Steve, and Anthony Hesketh. 2008. "Theorizing Under-Theorization in Research on the HRMPerformance Link." Personnel Review, Vol. 37, No. 2,
pp. 126-44.
Freeman, Richard and Joel Rogers 1999. What Workers

Want. Ithaca, NY, ILR Press (an imprint of Cornell
University Press).
Gerhart, Barry. 2005. "Human Resources and Business
Performance: Findings, Unanswered Questions, and
an Alternative Approach," Management Revue, Vol. 16,
No.2, pp. 174-85.
NY: ILR Press (an imprint of Cornell University
Press).
ment: Its Roots as Applied Ec
Backhouse and Jeff Biddle, eds., To
plied Economics, pp. 229-56. Du
versity Press.
HRM and Participative Managem

Early Industrial Relations." Hum
ment Review, Vol. 11 (Winter), pp
dustrial Relations to the F
Management Decision, Vol. 4
Interest in Industry." In Bruce K

ages." In Peter Boxali, JohnBeaumont
Purcell, and
and
Patrick
Roy
Helfgott, eds., Ind
Wright, eds., Oxford International

Handbook
of and
Human
Human
Resources
Beyond: The Evol
Resource Management, pp. 552-80.
Oxford,
UK: Management,
Oxford
ployee
Relations
pp. 11
University Press.
M.E. Sharpe.
Resource Management Review, Vol. 16, No. 4, pp.

Resource Management." In 467-77.
Bruce Kaufman ed.,
Marsden,
David.
1999. A Theory of
Employment Systems:
Theoretical Perspectives on Work and
the
Employment
Rela-
MicroCornell
Foundations of
Societal Diversity. Oxford, UK:
tionship, pp. 321-66. Ithaca, NY:
University
Oxford University Press.
Press.
MacDuffie, John. 1995. "Human Resource Bundles and

Manufacturing
of Human Resource Management in
AmericanPerformance:
Industry.Organizational Logic
and Flexible Production Systems in the World Auto
Ithaca, NY: Cornell University Press.
Industry." Industrial and Labor Relations Review, Vol. 48,
No. 2, pp. 197-221. for Labor Economtrial Relations and its Implications
Miller,
Ben. 2008. "Estimating the Firm's
Demand for
ics and Human Resource
Management."
Industrial
and Labor Relations Review,

Vol.
64,
No. 2,Practices."
pp. 74-108.
Human
Resource
Management
Diss.,
Georgia State University.
Mitchell, Daniel. 2001.
"IR Journal Reand Conference
Why It is Fundamentally Mis-Specified."
Industrial
Literature from the 1960s to the 1990s: What Can
lations, Vol. 49 (Spring), pp. 286-313.
HR Learn From It? Where is It Headed?" Human
Resource
Management
Vol.
11 (Winter), pp
Studies of HRM Programs and
Practices
in Review,
Early
American
375-94. University Press.
Industry. Ithaca, NY: Cornell
Ordiz, Monica, and Esteban Fernndez. 2005. "Influ-
ence of the
Sector and the Environment
on Human
Human Resource Management
Practices."
Interna-
Resource
Practices' Effectiveness."
Jourtionaljournal of Human Resource
Management,
21, International
Nos.
nal of Human Resource Management, Vol. 16, No. 8,

4-6, pp. 615-36.
pp. 1349-73.
Kleiner, Morris, Richard Block,
Myron Roomkin, and
Marc, and Stephen
Frenkel.
2005.
Alternative
Sidney Salsburg. 1987. HumanOrlitzky,
Resources
and
the
PerforLow-Involvement
Work Practices
and Business RePerformance of the Firm. Madison:
Industrial
Relations
search Association.
mance." International Journal of Human Resource Man-
agement, Vol. 16, No. 8, pp. 1325-48.

Kochan, Thomas 2007. "Social Legitimacy of the HRM
Osterman, Paul. 1987. "Choice of Employment Systems
Profession: A US Perspective." In Peter Boxali, John
Internal Labor Markets." Industrial Relations, Vol.
Purcell, and Patrick Wright, eds., Oxford Handbookinof
26, No. 1, pp. 46-67.

Human Resource Management, pp. 599-620. Oxford,
Paauwe, Jaap, and Paul Boselie. 2005. "HRM and Performance: What Next? Human Resource Management
UK: Oxford University Press.
Journal, Vol. 15, No.

4, pp. 68-83.
Transformation of American Industrial
Relations.
New
Pfeffer, Jeffrey. 1998. The Human Equation. Boston:

Laursen, Keld, and Nicholai Foss. 2003. "New Human Harvard University Business School Press.
Resource Management Practices, Complementarities,
and the Impact of on Innovation Performance." Cam-tional Behavior Perspective
York: Basic Books.
bridge Journal of Economics, Vol. 27, No. 2, pp. 243-63.
plained." Journal of Economic Pe
Lazear, Edward. 1986. "Salaries and Piece Rates, Jour-pp. 115-34.

nal of Business, Vol. 59, No. 3, pp. 405-31.
Priem, Richard, and John Butler. 2001. "Is the Resource-
based 'View' a Useful Perspective for Strategic Man-
agement
Academy of Management Review,
nal of Economics, Vol. 115, No. 1,
pp.Research?"
99-145.
Vol. 26, No. 1, pp. 22-40.
Proctor, Stephen.
2008. "New "JourForms of Work and the
ics: The Economist's View of Human
Resources.
Paradigm."
In Paul Blyton,
nal of Economic Literature, Vol. High
21, Performance
No. 4, pp.
91-114.
JackHuman
Fiorito, and Edmund
Lepak, David, and Scott Snell. Nicholas
1999.Bacon,
"The
Re- Heery,
Handbook of
Industrial
Relations, pp. 149-69.
source Architecture: Towardeds.,
a Sage
Theory
of
Human
London: Sage, pp.Academy
149-69.
Capital Allocation and Development."
of
Purcell,
Nick Kinnie, 2007. "HRM and busiManagement Review, Vol. 24, No.
1, John,
pp. and
31-48.
ness Performance." In Peter Boxali, John Purcell, and
Wright, eds.,
Contingency Framework for thePatrick
Delivery
ofOxford
HR Handbook
Prac-of Human Resource
Management,
pp. 533-51.
Oxford, UK: Oxford
tices." Human Resource Management
Review,
Vol.
15, No.
University Press.
2, pp. 139-59.
Rynes, Sara, Tamara Giluk, and Kenneth brown. Wo/.
"The VeryInternational
Separate Worlds of Academic
and
America: Looking to the Future."
Journal
Practitioner Periodicals
Human
Resource
of Human Resource Management"
Vol. in
19,
No.
8,Managepp.
1486-99.
ment: Implications for Evidence-Based Management."
Academy of Management Journal, Vol. 50, No. 5, pp.
Lewin, David. 2001. "Low-involvement Work Practices
and Business Performance." Proceedings of the 53rd 987-1008.
AnScarpello, Vida. 2008. "Parallel Approaches to Developnual Meeting, Industrial Relations Research Association,
ment of the HRM Field and HRM Education." In Vida
pp. 275-92. Champaign, IL: IRRA.
Marlow, Susan. 2006. Human Resource Management
in Smaller Firms: A Contradiction in Terms?" Human
Scarpello, ed., The Handbook of Human Resource Management Education, pp. 3-20. New York: Sage.
Schler, Randall, and Susan Jackson.

2001.
"Human
Towne,
Henry.
1886. "The ReEngineer and Economist,"
source Management: Past, Present

and of
Future."
Transactions
the AmericanIn
Society of Mechanical Engi-
Roger Blanpain and Chris Engles,neers,

eds.,
Vol.Comparative
7, pp. 425.
Labour Law and Industrial Relations in Industrialized
Verma, Anil. 2007. "What Do Unions Do in the WorkMarket Economies, pp. 181-205. The Hague: Kluwer
place: Union Effects on Management and HRM Poli-
Law International.
cies." In James Bennett and Bruce Kaufman, eds.,
Slichter, Sumner. 1919. "The Management of Labor."
What Do Unions Do? A Twenty Year Perspective, pp.
New Brunswick, NT: Transactions Press.
Journal of Political Economy, Vol. 27 (December), 275-312.
pp.
813-39.
Wall, Toby, and Stephen Wood. 2005. "The Romance of
Stockton, Frank. 1932. "Personnel Management in Human
the
Resource Management and Business PerforCollegiate School of Management." Personnel Journal,
mance, and the Case for Big Science." Human RelaVol. 12, pp. 220-27.
tions, Vol. 58, No. 4, pp. 429-62.
Subrahony, Mahesh. 2006. "Why Organizations Adopt
Weber, Wolfgang, and Rudiger Kabst. 2004. "Human
Some Human Resource Management PracticesResource Management: The Need for Theory and Diand Reject Others: An Exploration of Rationales."
versity." Management Revue, Vol. 14, No. 2, pp. 171-77.
Human Resource Management, Vol. 45, No. 2,Wright,
pp.
Patrick, Benjamin Dunford, and Scott Snell.
195-210.
2001. "Human Resources and the Resource Based
Sun, Li-Yun, Samuel Aryee, and Kenneth Law. 2007.

View of the Firm ." Journal of Management, Vol. 27, pp
"High-Performance Human Resource Practices, Citi-701-21.
zenship Behavior, and Organizational Performance: A

Relational Perspective." Academy of Management Jour-Allen. 2005. "The Relationship
nal, Vol. 50 No. 3, pp. 558-77.
and Firm Performance: Exam
Thompson, Paul, and Bill Hartley. 2007. "HRM and thePersonnel Psychology, Vol. 58, pp
Worker: Labor Process Perspectives." In Peter Boxali,
Youndt, Mark, Scott Snell, J
John Purcell, and Patrick Wright, eds., Oxford Interna-Lepak. 1996. "Human Resource
tional Handbook of Human Resource Management.facturing Strategy, and Firm Per
pp. 147-65. Oxford, UK: Oxford University Press.
Management Journal, Vol. 39, N

41149478

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

41149478

Diunggah oleh

Hak Cipta:

Format Tersedia

THE FIRM'S CHOICE OF HRM PRACTICES: ECONOMICS MEETS STRATEGIC HUMAN

THE FIRM'S CHOICE OF HRM PRACTICES: ECONOMICS MEETS

The authors compare and contrast two theoretical approaches to explain

HRM frequency distributions, new insights on the limitations of SHRM theo

and the first alternative to the standard "Huselid-type" regression mo

interaction between the two academic realms

in both economics and

stance and heavy on description and pre-

scription; management researchers, however,

* Bruce E. Kaufman is Professor of Economics and Se-

view economists' models as far too simplistic

nior Associate of the W.T. Beebe Institute of Personnel

and Employment Relations at Georgia State University

economics, he earned his PhD from the Andrew Young

THE FIRM'S CHOICE OF HRM PRACTICES 527

and abstract to capture much

that, even if less formalistic and

numerous books and arter explaining how and why agement

bution to this research stream is Huselid 's

management continue in most respects

neous events in the economy, such

2 This bifurcation of approach and opinion goes back

(if perhaps attenuated)

control and contingent

further because of its apparent lack of theory.

of HRMquespractices, typically associated

528 INDUSTRIAL AND LABOR RELATIONS REVIEW

ent direction for research on firms' choice

of HRM practices. At the same time, we abHuselid-type HRM-firm performance regres-

sion model in their empirical work butjure

2001; Bartel 2004; a partial exception is Framing the Research Question

3 Combs et al. (2006) reviewed the literature since

related to organizational performance" (p. 504).

of 'best practice' in 'excellent companies'

but primarily to identify and explain what

(2001); Batt (2002); Godard (2004); Rochan (2007);

and Grimshaw and Rubery (2007).

THE FIRM'S CHOICE OF HRM PRACTICES 529

Figure 1. Frequency Distribution of HRM Practices and H

thors combined the ten items into a scale

lars of expenditure. The HRM frequencynual expenditure per employee on HRM.

eral thousand employees and managers

530 INDUSTRIAL AND LABOR RELATIONS REVIEW

its central research

conspicuously absent - even discouraged.6

lary but more complicated issues of coverage,

plexity is the fact that even the definition of

The HRM frequency distributions help

management practices; in doing so, theseof

the paper declined to send it out for review. Among

THE FIRM'S CHOICE OF HRM PRACTICES 531

HRM is diverse and unsettled; Pauuwe and

Boselie (2005: 69) observed, for example,

that "there appears to be no consensus on

the nature of HRM."

practices are often associated in economics

developed internal labor market (ILM)-

employment system. Boselie, Dietz,

Boon (2005: 73) described this selecti

three-way typology developed by Delery tional on important contextual fac

and Doty (1996). They distinguished three

HRM practices for one firm may be quite

timal choice of HRM should follow a "best

fit" rule in which HRM practices are chosen