International Journal of Human Resource Management

ISSN: 0958-5192 (Print) 1466-4399 (Online) Journal homepage: http://www.tandfonline.com/loi/rijh20

Human resource strategies and firm performance:

what do we know and where do we need to go?

Lee Dyer & Todd Reeves

To cite this article: Lee Dyer & Todd Reeves (1995) Human resource strategies and firm
performance: what do we know and where do we need to go?, International Journal of Human
Resource Management, 6:3, 656-670, DOI: 10.1080/09585199500000041

To link to this article: https://doi.org/10.1080/09585199500000041

Published online: 28 Jul 2006.

Submit your article to this journal

Article views: 1790

View related articles

Citing articles: 334 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=rijh20
The International Journal of Human Resource Management 6:3 September 1995

Human resource strategies and firm performance:

what do we know and where do we need to go?

Lee Dyer and Todd Reeves

Abstract This research review focuses on the links between human resource strate-
gies and organizational effectiveness. It is likely that bundles of, or configurations
of, activities are more important in enhancing labour productivity than any single
activity. However, studies are typically limited in theoretical rigour, have quite
small samples and are typically non-cumulative. The empirical basis of strategic
human resources management is thus circumscribed.
Keywords Human resource strategies, firm performance, strategic human
resource management, labour productivity

Do human resource strategies - defined as internally consistent bundles of

human resource practices - contribute to organizational effectiveness? Maybe,
this review of the applicable research suggests, but it is not clear how or why.
Specifically: (1) bundles have been shown (in one study) to be more effective
than their individual components at enhancing labour productivity; (2) not all
bundles are equally effective, though - some configurations are clearly better
than others not only at enhancing labour productivity, but also at controlling
employee turnover and improving product quality; (3) the more effective bun-
dles, however, are by no means identical from study to study, belying the exis-
tence of a clearly superior human resource strategy and making it difficult to
explain the results; and (4) there is no convincing evidence that the explanation
lies in the contingency hypothesis - that is, in the notion that the more effective
bundles are those which fit best with extant business strategies (or organizational
contexts). But these observations are based on a limited number of studies
which are generally lacking in theoretical rigour, non-cumulative and focused on
quite narrow samples. At this juncture, then, it seems that the strategic human
resource management bandwagon, notwithstanding its conceptual allure, is bar-
relling along on a rather fragile empirical undercarriage.
Strategic human resource management (SHRM) has emerged as a, if not the,
major paradigm among scholars and practitioners in many parts of the world.
This is apparent from the recent literature on international human resource
management (e.g., Schuler, Dowling and De Cieri, 1993), as well from recent
reviews of trends in the US (Dyer and Kochan, in press), Canada (Betcherman
et al., 1994) and the UK (Lundy, 1994; Truss and Grattan, 1994).
The reasons for this emergence are many and varied, but two stand out:
increasingly intense global competition and the corresponding search for sources
of sustainable competitive advantage. For US firms, global competition emerged
in a major way in the 1970s, intensified in the 1980s and has become a way of
life in the 1990s. At first it was competition based on price, fostered in large
part by the enormous differences in wage rates between the US and many of its

0985-5192 0 Routledge 1995

 HR strategies and firm performance 657
major international competitors, beginning with the Japanese, but evolving to
the newly industrializing countries and, more recently, to the less developed
countries. With time, the focus shifted to quality, again driven in large part by
the Japanese, most prominently in the auto industry, but eventually extending to
consumer electronics, office machinery, industrial machinery and many other
industries. Currently, globally competitive prices and world-class quality are, for
many companies around the world, simply baseline; the real competitive action
swirls around customization, service, speed and innovation.
As capital and technology became increasingly available to virtually anyone
anywhere, the search for sources of sustainable competitive advantage increas-
ingly pointed inward towards organizational capability (Ulrich and Lake, 1990),
and more specifically to the strategic management of human resources (Dyer,
1993; Cappelli and Singh, 1992; Wright and McMahan, 1992). Again, the
Japanese played a major role since, rightly and wrongly, much of the credit for
that country's monumental advancements in productivity and, especially, quality
were attributed to its management processes, work systems and approaches to
people management (Nonaka, 1992). The notion was further advanced as
Japanese-managed transplants and joint ventures - particularly NUMMI (Adler,
1993) - achieved world-class levels of productivity and quality in the US with
standard technologies and indigenous employees (Appelbaum and Batt, 1994).
Additional support was provided by anecdotes and case studies of US compa-
nies which had enhanced their competitiveness through SHRM, as well as by the
emergence of several conceptual models purporting to provide explanations for
this success (for a review, see Dyer and Kochan, in press).
Thus, SHRM's spreading popularity owes much to the explicit promise of
greater organizational effectiveness achievable, according to the dominant mod-
els, through the development of internally consistent bundles of human resource
practices - human resource strategies - which are properly matched or linked to
extant organizational contexts, most notably business strategies (Dyer and
Kochan, in press; Milliman, Von Glinow and Nathan, 1991; Truss and Grattan,
1994).
The veracity of this promise is explored here. First, we briefly examine .the
logic of the underlying theories or assumptions and, then, review the applicable,
albeit limited, research. The idea is to provide answers to the following ques-
tions:
1 Why should firms bundle their human resource practices? To what extent,
and in what ways, do they?
2 Does bundling produce positive organizational results?. *
3 Is there one best bundle, or does it (as the contingency theorists' contend) all
depend?

The logic and prevalence of bundling

Human resource practices are said to be bundled when they occur-in fairly com-
plete, mutually reinforcing or synergistic sets (Dyer and Holder, 1988; Dyer and
Kochan, in press). The logic in favour of bundling is straightforward. First,
since employee performance is a function of both ability and motivation, it
makes sense to have practices aimed at enhancing both. Second, since employee
performance is an 'overdetermined phenomenon' (Hackman, 1985), it is most
658 Lee Dyer and Todd Reeves
likely to be maximized when influenced by non-independent activities which are
deliberately redundant (MacDuffie, 1995). Thus, best results should be obtained
when there are several ways for employees to acquire needed skills (e.g., careful
selection, on- and off-the-job training, etc.) and multiple incentives to enhance
their motivation (e.g., peer pressure, monetary and non-monetary rewards, etc.).
(This, of course, raises but does not answer the questions of how much and
what types of redundancies are optimal from a benefit-cost point of view.)
The logic of bundling has inspired several attempts to build typologies of
human resource strategies; that is, to construct conceptually defensible packages
of mutually reinforcing or synergistic human resource practices. Such typologies
are typically based on selected case studies. Examples include Walton's (1985)
well-known control and commitment dichotomy; Dyer and Holder's (1988)
inducement, investment and involvement trichotomy; Begin's (1991) carefully
crafted set of five 'employment policy systems'; and Dunlop's (1994) elaboration
of eight strategic types. (For a fuller review of this literature, see Dyer and
Kochan, in press).
Similarly, the logic of bundling has inspired a small number of survey studies
designed to derive taxonomies of human resource strategies (and to uncover the
antecedents and consequences of the resulting configurations). Those of which
we are aware are briefly described in Table 1.
While the typology literature (and especially Walton's [I9851 dichotomy) often
informs the interpretation of the taxonomical studies, the same has not been
true with respect to the up-front designs. As a result, these studies include a
bewildering array of input variables. Table 2 clarifies this point. Note, for exam-
ple, that the number of human resource practices measured in these studies
varies from a low of six (Ichniowski, 1990) to a high of eleven (MacDuffie,
1995). Further, there are twenty-eight different input measures across the four
studies. Only one - formal training - shows up in all four studies; only three -
incentiveslgain-sharing plans, employee participation or involvement, and due
process procedures - show up in three of the four studies. Twenty-two of the
twenty-eight human resource practices appear in only one study.
What about the resulting taxonomies? Table 3 provides a summary of these.
Arthur (1994), studying US steel minimills, derived two types - synthesized from
six clusters identified in earlier analyses (Arthur, 1990, 1992). MacDuffie (1995),
looking at auto assembly plants worldwide, identified three strategies.
Ichniowski, Shaw and Prennushi (1993), focusing on finishing operations in
twenty-one large steel mills in the US, came up with four models, while
Ichniowski (1990), using survey data from 255 large manufacturing firms also in
the US, derived nine. These differences are not surprising, given the variations in
input measures and samples noted above and the authors' choices of data ana-
lytic techniques (described briefly in Table l).
More surprising, perhaps, is the tendency of the various taxonomies to array
in reasonably comparable continua from, in the authors' terms, 'Traditional',
'Control' and 'Mass Production' strategies on one end to 'Innovative',
'Commitment' and 'Flexible (or Lean) Production' strategies on the other (as
shown in Table 3). Some similarities exist in the components of the anchoring
strategies. All four 'Traditional' models, for example, are characterized by fixed
job designs, minimal employee participation and low levels of formal training,
while all four 'Innovative' models exhibit opposite features. There are some
Table 1 Characteristics of significant studies

Author Sample Methodology

30 steel minimills in the US. Data collected by surveying HR managers (except

Arthur (1994) Covers production and business strategy data provided by top line managers).
maintenance employees. Taxonomy of HR strategies ('human resource systems')
derived using Ward's cluster analysis technique.
Dependent variables were labour productivity, quality and
employee turnover.
255 manufacturing firms in the Data collected by surveying HR directors. Taxonomy of HR
US. Unit of analysis is strategies ('human resource management systems') derived using
Ichniowski (1990) 'business' lines' (often Wong and Lane's K-th nearest neighbour clustering procedure.
corresponding to a division of Dependent variables were labour productivity and stock market
a company). Covers value (Tobin's q).
production employees.
30 steel finishing lines in 21 Data collected through standardized, but open-ended interviews
Ichniowski, Shaw, and steel companies in the US. with HR managers, operations managers, etc. Taxonomy of HR
Prennushi (1993) Covers production strategies ('HRM Systems') derived using nominate, a maximum
employees. likelihood scaling algorithm. Dependent variables were
productivity and, parenthetically, quality.
62 high volume auto Data collected by sending surveys to a contact person in each
MacDuffie (1995) assembly plants worldwide. plant, who then distributed different sections to appropriate
Covers production departmental managers. Taxonomy of HR strategies ('bundles')
employees. determined using cluster analysis of three pre-determined scales -
buffers, work system, and HRM policies. Dependent variables
were productivity and quality.
Table 2 Human resource policies and practices measured

Authors (with number of policies and practices in parentheses)

Human resources
policies and Arthur Ichniowski Ichniowski el al. MacDuffie*
practices (N=IO) (N-6) (N=18) (N=ll)

Staffing
*Tight hiring
standards
*Promotion from
0 .
within
*Promotion based
on merit
Training
*Formal training
*Type of training
Rewards & Recognition
*Labour costs
*Incentivestgain-
sharing etc.
*Knowledge-based
Pay
*Benefit costs
Work Systems
*Participation
(QC, teams, etc.)
*Decentralized
decision-making
*Flexibility
*Problem solving
*Job rotation
*Respon for quality
Communication
*Information sharing
Employee relations
*Due process
procedure
*Opinion surveys
*Social gatherings
Other
*Skill mix
*Span of control
*No of job
classifications
*Customer visits
*Status barriers
*No suggestions/EE
.
*% suggestions
implemented
*Employment security
*Unionized
.O
.
*Data reported here pertain to MacDuffie's work systems and HRM policies scales only; his
buffers scale is omitted since it pertains to the production system rather than HR practices.
 H R strategies and j r m performance 66 1
glaring inconsistencies across the anchoring strategies. For example, Arthur's
(1994) 'Control' model contains incentivelbonus plans, while MacDuffie's (1995)
'Mass Production' model is characterized by non-contingent compensation.
The data show that between 10 per cent (Ichniowski, 1990) and 50 per cent
(MacDuffie, 1995) of the firms studied followed a 'Traditional', 'Control' or
'Mass Production' human resource strategy (Arthur's [1990, 19921 original clus-
ter analysis showed 27 per cent.) Comparable figures for the 'Innovative',
'Commitment' or 'Flexible (or Lean) Production' strategy are between 10 per
cent (Ichniowski, 1990) and 25 per cent (MacDuffie, 1995). (Arthur [1990, 19921
found 23 per cent.) This means that between one-quarter and four-fifths of the
firms studied had human resource practices which were not obviously bundled;
that is, they had human resource strategies (if that is the right term) character-
ized as 'Mixed' or 'Transitional' (or which could not be classified).
In brief, while logic supports the notion of bundling human resource prac-
tices, theoretical and empirical attempts to define the components and measure
the contents and prevalence of bundles, while individually impressive, are collec-
tively deficient and confusing. As a consequence, we have but a bare inkling of
the number of firms which are actually managing at least some of their human
resources strategically, and of the types of strategies they are using.

Bundling and organizational effectiveness

Theoretically, as noted above, bundling should promote organizational effective-
ness. At a minimum, bundles should produce greater performance effects than
any of the individual human practices of which they are composed. But, it is
unlikely that all bundles are created equal: that is, some types should produce
greater performance effects than others. Here we explore both issues.
First, a word about organizational effectiveness. In research pertaining to
human resource strategy, there are several types of outcomes which might apply
(Dyer, 1984). The three most defensible are: (1) human resource outcomes such
as absenteeism, turnover and individual or group performance; (2) organiza-
tional outcomes such as productivity, quality and service; and (3) financial or
accounting outcomes such as return on invested capital or return on assets. A
fourth possibility, for publicly held firms, is stock-market performance as mea-
sured by stock value or shareholder return.
Alas, in research, trade-offs must be made. The face validity of these out-
comes goes in the order presented. That is, human resource strategies are likely
to have their most direct effects on human resource outcomes, next greatest on
organizational outcomes, and so forth. This reflects, in part, what such strate-
gies are designed to do and, in part, the complexity of factors which affect out-
comes such as profitability, not to mention stock prices. And yet, from a
strategic standpoint, human resource outcomes are probably deficient outcomes
in the eyes of most managers, whose strategic goals certainly extend to organi-
zational outcomes and probably beyond these to the bottom-line (and if they
have stock options, perhaps all the way to Wall Street). These trade-offs may
explain why the research thus far has focused primarily on organizational out-
comes, especially productivity and quality, and only marginally on the other
three types.
Table 3 Typologies of human resource strategies

Types (with representative characteristics)

No of
Author types 'Traditional' types Mixed type(s) 'Innovative' types

Control (little Commitment (moderate

employee employee participation,
participation, little moderate general training,
2 general training, low high wages, 14% of
Arthur (collapsed wages, 80% of employees on bonuses o r
(from 6 ) employees on bonus o r incentives) (47% of sample)
incentives)
(53% of sample)
Traditional-union (fixed Union-like, d o nothing, High commitment (flexible
job design, promotion complex (collectively, job design, formal training
from within based o n 58% of sample) (22% of programmes, internal
lchniowski seniority, grievance sample wasn't promotions, opinion survey
procedures with classified) o r information sharing) (10%
arbitration of sample
(10% of sample)
System 4: traditional Systems 2 and 3 (mixed System I : innovative
(strict work rules, between systems I and careful selection,
Ichniowski, incentive pay based o n 4) ( O h not reported) extensive training, team
Shaw, and quantity not quality, problem solving, multi-
Prennushi minimal employee attribute gain-sharing,
involvement, informal skill-based pay (Oh not
training) reported)
(Oh not reported)
Mass production (5% Transitional (between Flexible production (70% of
of employees in teams, mass and flexible employees in teams,
little quality control production systems on considerable job rotation,
o n s h o p floor, little all variables) (25% of extensive quality control
training, sample) on shop floor, extensive
non-contingent training, contingent
compensation, compensation, minimal
extensive status status differentials) (25%
differentiation) of sample)
(50% of sample)

'Based on MacDuffie's 3 scales - buffers, work systems and HRM policies.

 HR strategies and firm performance 663

Bundles vs. their individual components

Only one study specifically examined the bundles vs. individual components
question. Using regression models with multiple control variables, ~chniowskiet
al. (1993) found small, but statistically significant, positive relationships between
productivity (uptime of steel mill finishing lines) and most of the eighteen
human resource practices examined. Focusing in on ten of the human resource
practices which were significantly changed during the course of their study, they
found that seven resulted in productivity improvements in the range of one or
two percentage points.
The effects of the eighteen individual human resource practices and the ten
changes, however, were virtually eliminated when measures of 'human resource
systems' (the four noted in Table 3) were entered in the various regression equa-
tions. Based on this and related evidence, the authors concluded that '[slystems
of HRM policies have bigger productivity effects than the sum of any . . .
effects due to individual practices' (p. 48) and, concerning change, that
'[mJarginal changes in individual policies have little or no eflect on productivity.
Improving productivity requires substantial changes in a set of HRM policies'
(Ichniowski et al., 1993: 37, italics in original).

Bundles vs bundles
Four studies have examined the relative effects of various bundles of human
resource practices on productivity, three of which also included a quality mea-
sure. Ichniowski et al. (1993), in addition to comparing bundles against their
individual components, also compared their four systems (see Table 3) in terms
of both productivity and quality. In terms of productivity (line uptime), the
'Innovative' system significantly outperformed the two 'Mixed' systems (by about
eight percentage points) which, in turn, outperformed the 'Traditional' system
(by about three percentage points). With respect to quality ('prime yield' rates),
the pattern was nearly the same: the 'Innovative' system outperformed the better
of the two 'Mixed' systems (by about five percentage points) which, in turn, out-
performed the remaining two systems (also by about five percentage points).
Comparing steel minimills having a 'Commitment' human resource strategy
with those having a 'Control' strategy (again, refer to Table 3), Arthur (1994)
found that the former had both significantly higher levels of productivity (i.e.,
fewer labour hours per ton of steel produced) and significantly higher levels of
quality (i.e., lower scrap rates).
In MacDuffie's (1995) study, higher levels of productivity (i.e., fewer hours
per vehicle produced) prevailed in auto assembly plants with 'Flexible (or Lean)
Production' systems than in those with either 'Transitional' or 'Mass
Production' systems (which were not significantly different from one another).
Quality (J.D. Power's 1989 survey of new car buyers) was also higher in the
plants with 'Flexible (or Lean) Production' systems; in addition, those with
'Transitional Production' systems had significantly higher quality than those
with 'Mass Production' systems. The combination of work systems and human
resource policies only (i.e., omitting the buffers component, as noted in Table 2)
produced the same result with respect to productivity, but not quality. The
complete regression models explained more variance in productivity than in
664 Lee Dyer and Todd Reeves
quality. But, because more of the variance in the former was explained by the
control variables used in the analyses, the human resource strategy variables
actually explained more of the variance in quality than in productivity.
The across-industry survey reported by Ichniowski (1990) compared the pro-
ductivity effects of the 'Commitment' model against the eight others identified
through cluster analysis (see Table 3) and, once again, the 'Commitment' model
proved superior.
The only study which included a human resource outcome (Arthur, 1994)
found human resource strategy to be significantly related to employee turnover
rates. The average turnover rate in minimills with a 'Commitment' strategy was
less than half that in minimills with a 'Control' strategy (annual rates of 3 and 7
per cent, respectively). (In an interesting sub-analysis, Arthur also found that,
while employee turnover was negatively related to productivity and quality
among minimills with a 'Commitment' strategy, no comparable relationship pre-
vailed among those with a 'Control' strategy. He suggested that, in the latter
case, the relationship may be curvilinear: that is, employee turnover may have
no effect on organizational outcomes in minimills with minimal investments in
employees up to some point, but as rates exceed that point they may become
dysfunctional. At any rate, the moderating effect of human resource strategy on
relationships between human resource and organizational outcomes is certainly
an interesting avenue for future exploration.)
Ichniowski (1990) also looked at the effects of human resource strategies on
stock-market performance (Tobin's q). The results were consistent with those
reported earlier: that is, the 'High Commitment' model was more strongly
related to Tobin's q than were any of the remaining eight models. But, as would
be expected, the amount of variance explained in this variable was only about
25 per cent, compared with about 85 per cent of variance explained in produc-
tivity (with the same control variables appearing in both models).

Summary
Ichniowski et al. (1993) make a strong case for the value of bundling, although
clearly more studies in more industries among more employee groups are
needed. And, of course, the benefit-cost question raised earlier remains to be
answered.
More interesting, perhaps, is the consistent support for the superiority of the
'Innovative', 'Commitment' and 'Flexible (or Lean) Production' models.
Notwithstanding considerable variation in methodological approaches, and their
attendant difficulties, this pattern holds across three industry studies (one each
in autos, big steel and minimills) and a broad survey, and across four different
types of organizational effectiveness measures.
But, as noted earlier, while there are similarities, the contents of the
'Commitment', 'Involvement' and 'Flexible (or Lean) Production' models are far
from identical. Employee involvement anchors all four of the models cited (see
Table 3). But (as noted earlier), there is considerable variation in the conceptual-
ization and measurement of this construct - a direct measure, the prevalence of
flexible work designs, use rates of various types of teams, etc. Careful selection,
extensive training and contingent compensation are the only other attributes to
appear in majority of the four models.
 HR strategies andJirm performance 665
This variation raises an interesting issue: to what can the consistently solid
support for the 'Innovative', 'Commitment' and 'Flexible (or Lean) Production'
models be attributed? If their similarities outweighed their differences, we could
perhaps conclude that there is, indeed, a superior human resource strategy out
there. But, this does not seem to be the case. Perhaps these models are picking
up other attributes to which they are indirectly related. For example, they may
represent generally similar broad philosophies of human resource management
which have positive performance effects irrespective of the specific human prac-
tices used in their implementation. This possibility has been raised by Kochan
and Osterman (1994) (in their 'Mutual Gains' model), and even more explicitly
by Pfeffer (1994). Another possibility is that these models represent even more
fundamental and basic underlying variables, such as trust; this notion received
some support in the interviews conducted by Ichniowski et al. (1993) in the steel
mills, and was suggested as well by Kochan and Osterman (1994). Or, perhaps,
it is simply that these 'high end' models require and receive considerably more
care and attention than do the others in terms of both design and construction,
and what is being picked up in the various studies is an interesting variation on
the well-known (if not universally accepted) Hawthorne effect.

Contingencies
Is there one best way to bundle human resource practices, or does it 'all
depend'? The preceding analysis would appear to support the former view,
within the confines of the contexts studied and the limitations noted.
But even the most fervent advocates of particular human resource strategies
acknowledge that their preferred models are probably not always best. Ed
Lawler, the high priest of the 'High Involvement' model, for example, notes in
his most recent book that this approach is 'not necessarily . . . right . . . for all
environments and all societies' (Lawler, 1992: xiv). Similar positions are
expressed, more or less firmly, by other dominant model theorists (e.g., Kochan
and Osterman, 1994; Pfeffer, 1994). In brief, just about everybody who writes
about human resource strategy subscribes to a contingency hypothesis to some
extent.
And with good reason. Logically, as noted earlier, different organizational
environments (e.g., those characterized by various combinations of business
strategies, organizational structures and process technologies [Dyer and Kochan,
in press]) should require both different types of employees and different types of
behaviours from these employees. Further, different human resource strategies
should produce both different types of employees and different types of behav-
iours from these employees. It therefore follows that a good match between
human resource strategies and organizational environments should produce core
capabilities which would enhance an organization's competitive position and,
other things being equal, result in greater organizational effectiveness (Cappelli
and Singh, 1992; MacDuffie, 1995; Schuler 1992; Schuler and Jackson, 1987).
A full test of this theoretical perspective would require, in addition to reason-
able hypotheses about specific relationships: (1) typologies or taxonomies of
organizational environments and the nature of the people and behaviours
required by each type; (2) typologies or taxonomies of human resource strategies
(such as those noted above) and the nature of the people and behaviours pro-
666 Lee Dyer and Todd Reeves
duced by each type; and (3) comparisons of core competencies andlor organiza-
tional effectiveness across organizations which are characterized by high and low
degrees of fit between '1' and '2'. This is a complicated set of analyses, and so
far no one has taken it on.
A couple of partial tests, involving two of the studies reported on earlier, have
produced conflicting results.
Arthur (1990) examined the performance effects associated with fit between
business strategy (the most commonly mentioned component of organizational
environment) and human resource strategy in a preliminary analysis of his min-
imill data. He defined fit as occurring when a cost-based business strategy was
combined with a 'control-oriented' human resource strategy and when a differ-
entiation-based business strategy was combined with a 'commitment-type'
human resource strategy. Obverse combinations constituted no-fit. He then
hypothesized that the mills characterized by fit would have higher levels of pro-
ductivity (i.e., fewer labour hours per ton of steel produced) and quality (i.e.,
lower scrap rates) and lower levels of employee turnover than those character-
ized by no-fit.
Selected data from his analysis are reproduced in Table 4. Across the entire
sample (twenty-seven mills), the expected differences appeared, but did not reach
conventionally accepted levels of statistical significance. Among mills following a
differentiation-based business strategy, those with fit (eleven mills) had 25 per
cent higher productivity (p < .lo), 34 per cent higher quality (p < .lo), and 67
per cent less employee turnover (p < .05) than those with no-fit (eight mills).
(No comparable analysis was possible for mills following a cost-based business
strategy since the no-fit cell contained only one case.) Overall, these results pro-
vide modest support for the contingency hypothesis.

Table 4 An examination of the p t ' hypothesis

Full sample
Fit Nofit
(N= 18) (N=20)
Mean SD Mean SD T

Productivity (labour hrs) 2.20 .79 2.38 1.17 .48

Quality (scrap rates) ,149 ,118 .I94 .078 .97
EE turnover ,053 ,067 ,063 ,050 .41

Only differentiators
Fit No fit
(N=ll) (N=8)
Mean SD Mean SD T

Productivity (labour hrs) 1.91 .74 2.55 1.14 1.49*

Quality (scrap rates) .I28 ,085 .I94 .078 1.67*
EE turnover .022 ,014 ,067 .052 2.73**
*p<.lO **p<.025
Source: Arthur (1990: 170).
 H R strategies andjirm performance 667
A different pattern of results emerged from an earlier analysis of the data
from MIT's worldwide study of auto assembly plants (MacDuffie and Krafcik,
1992). The focus, again, was on productivity (i.e., labour hours per vehicle pro-
duced) and quality (i.e., defects per 100 vehicles). Fit was defined in terms of the
linkage between human resource strategy and process technology (rather than
business strategy as in the Arthur study). Representative results are shown in
Table 5 (MacDuffie and Krafcik, 1992). Consistent with the contingency
hypothesis, the highest levels of productivity (twenty-two hours per vehicle pro-
duced) and quality (forty-nine defects per 100 vehicles) prevailed in high tech
plants with 'Flexible (or Lean) Production' human resource strategies. Also con-
sistent with this hypothesis is the fact that performance falls off considerably in
high-tech plants with a 'Mass Production' human resource strategy. In contrast,
however, the lowest levels of productivity (forty-one hours per vehicle produced)
and quality (104 defects per vehicle) were found in low-tech plants with 'Mass
Production' human resource strategies. Further, performance in low-tech plants
was considerably higher where 'Flexible (Lean) Production' human resource
strategies prevailed. Overall, these results provide strong support for the general
superiority of the 'Flexible (or Lean) Production' human resource strategy over
the 'Mass Production' human resource strategy in auto assembly plants (irre-
spective of the sophistication of the process technology), a conclusion seemingly
out of sync with the contingency hypothesis.

Table 5 Another examination of the 'jit' hypothesis

Mass production Lean production

HR strategy HR strategy

High
tech
Low P = 41 P = 35
tech Q = 104 Q = 73
P = Hours per vehicle
Q = Defects per 100 vehicles
Source: Adapted from MacDuffie and Krafcik (1992: 220).

As always, more research is needed. It could benefit from some prior theoreti-
cal work more precisely delineating not only the key components of the organi-
zational environment, but also the human and behavioural requirements
associated with various configurations of this environment. More ,conceptual
work is also needed on the human resource strategy side (as noted above),
including some attempt (beyond the very useful, but somewhat limited efforts of
Schuler [I9921 and Schuler and Jackson [1987]) to be more precise about the
expected human and behavioural outcomes engendered by various strategies.
The natural follow-on, then, would be more complete measures of organiza-
tional environments, as well as the inclusion of appropriate intervening human
resource outcomes between these environments, on the one hand, and organiza-
tional and financial outcomes, on the other.
668 Lee Dyer and Todd Reeves

Summary and conclusion

There are solid theoretical reasons for firms to bundle their human resource
activities into identifiable strategies. How many actually do, even among a sub-
set of their employees? The available data, which are sparse, suggest that the
figure is probably no more than 20 per cent generally, although it may go as
high as 50 to 75 per cent in some industries (e.g., steel minimills in the US,
automobile assembly plants worldwide). How many bundles, or identifiable
strategies, are there? Studies have turned up as few as three and as many as nine
(although many of the 'latter appear to lack sufficient internal consistency to
constitute real strategies). The problem is that the contents of these bundles vary
quite a bit from study to study, primarily because there is so little consistency in
the measures of human resource activities from which they are derived.
Bundles seem to be superior to any of the individual human resource activities
of which they are composed in enhancing and facilitating positive changes in
productivity and, perhaps, quality (Ichniowski et al., 1993). But, not all bundles
are created equal: some are clearly superior to others when it comes to reducing
employee turnover (Arthur, 1994) and to enhancing productivity (Arthur, 1994;
Ichniowski, 1990; Ichniowski et al., 1993; MacDuffie, 1995), quality (Arthur,
1994; Ichniowski et al., 1993; MacDuffie, 1995) and stock-market performance
(Ichniowski, 1990). The superior strategies are variously labelled 'Innovative',
'Commitment' and 'Flexible (or Lean) Production'. All incorporate a dimension
of employee involvement, and a majority also involve careful selection, extensive
training and contingent compensation. It is possible, then, that, culled and inte-
grated, these components combine to form a particularly powerful human
resource strategy.
But, probably not. Notwithstanding their similarities, these models differ in
several important respects. Given this, it is perhaps just as likely that they
reflect, to some extent anyway, underlying endogenous factors such as particu-
larly salient management philosophies or careful management attention, and
thus that the observed individual and organizational outcomes are somewhat
analogous to the well-known Hawthorne effect.
Few are those who argue with the general proposition that best results are
attained when human resource strategies fit (or are linked with) extant business
strategies and, perhaps, other elements of organizational contexts (Dyer and
Kochan, in press). It is, therefore, somewhat surprising to find that this proposi-
tion has received so little theoretical or, particularly, empirical attention. We
found only a couple of relevant studies - one (Arthur, 1990) generally support-
ive of the conventional wisdom and the other (MacDuffie and Krafcik, 1992)
not.
Broadly, then, the only reasonable conclusion is that research into the effects
of human resource strategies on organizational effectiveness, while promising, is
still in its infancy. The studies are few in number, generally lacking in concep-
tual sophistication and overwhelmingly focused on a shrinking component of
the US workforce, namely blue-collar workers in heavy manufacturing. Thus,
while there is every reason to believe that the strategic bandwagon will continue
to roll, at this point it is wise to bear in mind the rather fragile empirical under-
carriage on which it currently rests.
 H R strategies and firm performance 669
Lee Dyer
Center for Advanced Human Resource Studies
I L R School, CorneN University
Ithaca, N Y
Todd Reeves
A T S T - Global Information Solutions
West Columbia, SC,
USA

References
Adler, P. (1993) 'The "Learning Bureaucracy": New United Motors Manufacturing, Inc.'
In Staw, B. and Curnmings, L. (eds) Research In Organizational Behavior, Volume 15.
Greenwich, CT: JAI Press, pp. 111-94.
Appelbaum, E. and Batt, R. (1994) The New American Workplace. Ithaca, NY: ILR
Press, Cornell University.
Arthur, J. (1990) Industrial Relations and Business Strategies in American Steel Minimills.
Unpublished PhD dissertation, Cornell University.
Arthur, J. (1992) 'The Link Between Business Strategy and Industrial Relations Systems
in American Steel Minimills', Industrial and Labor Relations Review, 45(4): 488-506.
Arthur, J. (1994) 'Effects of Human Resource Systems on Manufacturing Performance
and Turnover', Academy of Management Journal, 37(4): 670-87.
Begin, J. (1991) Strategic Employment Policy: An Organizational Systems Perspective.
Englewood Cliffs, NJ: Prentice-Hall.
Betcherman, G., McMullen, K., Leckie, N. and Caron, C. (1994) The Canadian Work-
place in Transition. Kingston, Ontario: Industrial Relations Centre Press, Queen's
University.
Cappelli, P. and Singh, H. (1992) 'Integrating Strategic Human Resources and Strategic
Management'. In Lewin, D., Mitchell, 0. and Scherer, P. (eds) Research Frontiers in
Industrial Relations and Human Resources. Madison, WI: Industrial Relations Research
Association, pp. 165-92.
Dunlop, J. (1994) 'Organizations and Human Resources: Internal and External Markets'.
In Kerr, C. and Staudohar, P. (eds) Labor Economics and Industrial Relations: Markets
and Institutions. Cambridge, MA: Harvard University Press, pp. 375-400.
Dyer, L. (1984) 'Studying Human Resource Strategy: An Approach and an Agenda'.
Industrial Relations, 23(2): 156-69.
Dyer, L. (1993) Human Resources as a Source of Competitive Advantage. Kingston,
Ontario: Industrial Relations Centre Press, Queen's University.
Dyer, L. and Holder, G. (1988) 'A Strategic Perspective of Human Resource
Management'. In Dyer, L. (ed.) Human Resource Management: Evolving Roles and
Responsibilities. Washington, DC: Bureau of National Affairs, pp. 1-46.
Dyer, L. and Kochan, T. (in press) 'Is There a New HRM? Contemporary Evidence and
Future Directions'. In Downie, B., Kumar, P. and Coates, M.L. (eds) Managing
Human Resources in the 1990s and Beyond: Is the Workplace Being Transformed?.
Kingston, Ontario: Industrial Relations Centre Press, Queen's University.
Hackman, R. (1985) 'Doing Research That Makes a Difference'. In Lawler, E.E. 111, et
al. (eds) Doing Research That Is UseJul in Theory and Practice. San Francisco, CA:
Jossey-Bass.
Ichniowski, C. (1990) 'Human Resource Management Systems and the Performance of
U.S. Manufacturing Businesses'. National Bureau of Economic Research. Working
Paper No. 3449, September.
Ichniowski, C., Shaw, K. and Prennushi, G. (1993) 'The Effects of Human Resource
670 Lee D y e r a n d Todd Reeves
Management Practices on Productivity'. Unpublished paper presented at the ILR
School, Cornell University, 13 September.
Kochan, T. and Osterman, P. (1994) The Mutual Gains Enterprise. Boston, MA: Haward
Business School Press.
Lawler, E.E. 111. (1992) The Ultimate Advantage: Creating the High Involvement
Organization. San Francisco, CA: Jossey-Bass.
Lundy, 0. (1994) 'From Personnel Management to Strategic Human Resource
Management', The International Journal of Human Resource Management, 5(3):
687-720.
MacDuffie, J.P. (1995) 'Human Resource Bundles and Manufacturing Performance:
Flexible Productions Systems in the World Auto Industry', Industrial and Labor
Relations Review, 48(2): 197-221.
MacDuffie, J.P. and Krafcik, J. (1992) 'Integrating Technology and Human Resources for
High-Performance Manufacturing: Evidence from the International Auto Industry'. In
Kochan, T. and Useem, M. (eds) Transforming Organizations. New York: Oxford
University Press, pp. 209-26.
Milliman, J., Von Glinow, M.A. and Nathan, M. (1991) 'Organizational Life Cycles and
Strategic International Human Resource Management in Multinational Companies:
Implications for Congruence Theory', Academy of Management Review, 16(2): 318-39.
Nonaka, I. (1992) 'Self-renewal of the Japanese Firm and the Human Resource Strategy'.
In Pucik, V., Tichy, N. and Barnett, C. (eds) Globalizing Management: Creating and
Leading the Competitive Organization. New York: Wiley, pp. 300-18.
Pfeffer, J. (1994) Competitive Advantage Through People. Boston, MA: Harvard Business
School Press.
Schuler, R. (1992) 'Strategic Human Resource Management: Linking the People With the
Strategic Needs of the Business', Organizutional Dynamics, Summer: 18-31.
Schuler, R. and Jackson, S. (1987) 'Linking Competitive Strategies With Human Resource
Management Practices', Academy of Management Executive, l(3): 207-1 9.
Schuler, R., Dowling, P. and De Cieri, H. (1993) 'An Integrative Framework of Strategic
International Human Resource Management', Journal of Management, 19(2): 419-59.
Truss, C. and Grattan, L. (1994) 'Strategic Human Resource Management: A Conceptual
Approach', The International Journal of Human Resource Management, 5(3): 663-86.
Ulrich, D. and Lake, D. (1990) Organizational Capability: Competing From the Inside Out.
New York: Wiley.
Walton, R. (1985) 'From Control to Commitment in the Workplace', Harvard Business
Review, March-April: 77-84. Wright, P. and McMahan, G. (1992) 'Theoretical
Perspectives for Strategic Human Resource Management', Journal of Human Resource
Managemenr, 18(2), 295-320.