Content uploaded by Sanford M Jacoby
Author content
All content in this area was uploaded by Sanford M Jacoby on Feb 20, 2017
Content may be subject to copyright.
203
Industrial and Labor Relations Review, Vol. 56, No. 2 (January 2003). © by Cornell University.
0019-7939/00/5602 $01.00
THE EFFECT OF EMPLOYER NETWORKS
ON WORKPLACE INNOVATION AND TRAINING
CHRISTOPHER L. ERICKSON and SANFORD M. JACOBY*
If innovative work practices improve performance, why does the intensity of
their adoption vary substantially across establishments? Following a lead sug-
gested by some sociological studies, the authors empirically investigate the role
of social networks (ties to other organizations) in the organizational learning
associated with diffusion of innovative work practices. Using establishment data
on formal affiliation and other network measures, they find that managerial
participation in networks—specifically, in industry and cross-industry associa-
tions, civic organizations, and the internal networks of multi-unit firms—
positively affected both the probability that high-performance work practices
and employee training programs would be adopted and, where they were
adopted, the intensity of their adoption. Furthermore, multiple affiliations
raised the likelihood that an establishment would pursue an intensive approach
to work reorganization and training.
Christopher Erickson is Professor of Management
and Sanford Jacoby is Professor of Management, Policy
Studies, and History, both at UCLA. For assistance of
various kinds, the authors thank the State of Califor-
nia Employment Development Department, Kenneth
Budman, Richard Copeland, Lisa Evans, Sophie
Flaherty, Peter Goldschmidt, and Jonathan Troper.
Helpful comments were received from Ronald Burt,
Robert E. Cole, Keith Head, and Arne Kalleberg. The
research was supported by a grant from the California
Policy Research Center.
The quantitative data used in this study are the
property of the State of California Employment De-
velopment Department. For questions regarding these
data and the computer programs used to generate the
results, please contact Christopher Erickson at the
UCLA Anderson School of Management, Box 951481,
Los Angeles, CA 90095–1481.
ver the past three decades, U.S. com-
panies have been taking a new ap-
proach to work organization, one that is
more participative, flexible, and team-ori-
ented. The new approach goes by different
names, among the most popular being
“high-performance” work practices. Such
practices are becoming increasingly com-
mon. A survey of establishments with 50 or
more employees found that around 40%
had self-directed work teams, 50% used job
rotation, and nearly 60% had quality circles
(Osterman 2000). These figures—and
those from other surveys (Gittleman et al.
1998)—suggest that, depending on the
practice, somewhere between one-fifth and
three-fifths of mid-to-large-sized U.S. estab-
lishments are using at least one high-per-
formance work practice.
Driving these changes have been the eco-
nomic gains associated with the new ap-
proach. A growing literature finds strong
O
204 INDUSTRIAL AND LABOR RELATIONS REVIEW
effects—on productivity, quality, profitabil-
ity, stock prices, and other outcomes—from
adopting innovative work practices (Becker
and Gerhart 1996; Huselid 1995; Ichniowski
et al. 1997). Much of this literature is
focused at the establishment level, where
productivity can most accurately be mea-
sured.
Yet the high-performance approach is
still far from being widely accepted. Smaller
establishments are less likely than large
firms to pursue new forms of work organi-
zation, and even many larger establishments
have yet to adopt them (Gittleman et al.
1998). Those that adopt these practices do
so with varying intensity, whether measured
by the number of high-performance prac-
tices or by the share of an establishment’s
employees using them. In California, 30%
of establishments with over 50 employees
have self-directed work teams, but only 11%
of these establishments’ nonsupervisory
employees participate in them (Erickson
and Jacoby 1998).
A question that has not been adequately
addressed is why, if innovative work prac-
tices are beneficial to performance, the
intensity of adoption varies greatly across
establishments. Osterman (1994) found a
link between adoption of high-performance
practices and a company’s overall business
strategy, such as whether it competes on
cost or quality. Pil and MacDuffie (1996)
found that adoption intensity is greater
when an establishment has already adopted
complementary HR practices (high levels
of training, for example) or complemen-
tary technology (flexible automation). The
idea of complementarity between work prac-
tices on the one hand and HR practices and
technology on the other is derived from
studies of “bundling” (MacDuffie 1995;
Dunlop and Weil 1996), which have found
that the strongest productivity effects of
high-performance work practices occur
when they are adopted intensively and when
they are coupled with complementary HR
policies and technological conditions.
The bundling phenomenon can be in-
terpreted as a form of organizational learn-
ing. As Pil and MacDuffie (1996) argued,
many firms adopt an incremental, trial-
and-error approach to work innovation.
They change one or two practices and ob-
serve the results before proceeding fur-
ther. But the biggest payoffs arise from
more comprehensive change in work orga-
nization, HR practices, and technology.
Unless an organization is aware of this fact—
and understands how to effect thorough-
going change—it will remain stuck in a
small change/low payoff equilibrium.
How does organizational learning oc-
cur, and does it affect the adoption of work-
place innovation? These questions moti-
vate the present study. Although there is a
sizable literature on organizational learn-
ing (Weick and Westley 1996), its insights
have rarely been applied or tested by those
studying workplace innovation. The orga-
nizational learning literature suggests that
many companies have access to explicit
information (knowledge about “what”) but
vary significantly in their ability to apply
this information (knowledge about “how”).
The knowing-doing gap arises when com-
panies lack access to information about
implementation, troubleshooting, and
customization. Although this tacit knowl-
edge lowers the barrier to innovation, the
market does not readily supply it (Powell
1990; Pfeffer and Sutton 1999).
One way organizations gain tacit knowl-
edge is through social ties to other organi-
zations, that is, through networks. Net-
worked organizations are those that are
cosmopolitan, externally oriented, and
characterized by multiplex, nonredundant
affiliations. With respect to the workplace,
the information that flows through net-
works can include knowledge about the
performance effects of work reform, the
benefits of bundling, and tactics to per-
suade managers and employees that inno-
vation is beneficial.
In this paper we first review the literature
on networks and innovation and develop
hypotheses regarding the effect of networks
on the adoption of human resource inno-
vations. Next we test the hypotheses, using
establishment data on formal affiliation and
other network measures to predict the in-
tensity of workplace innovation and of train-
ing. We then report on interviews with
EMPLOYER NETWORKS 205
managers and assess issues raised by the
quantitative analysis. We close with some
suggestions for policy and for future re-
search.
Networks and Innovation
Recently there has been a proliferation
of studies emphasizing the role of networks
in organizational learning (for example,
Nohria and Gulati 1994; Strang and Soule
1998). By enhancing access to knowledge,
networks promote awareness and early
adoption of an innovation. By promoting
social interaction, they generate trust and
norms of reciprocity—social capital—that
are conducive to knowledge transfer. Reci-
procity creates an incentive for informa-
tion sharing, while trust enhances the qual-
ity of the information being shared. Al-
though markets also can transmit informa-
tion through price signals and contractual
relations, networks are more adept at dif-
fusing complex information concerning
quality and know-how (Uzzi 1997; Podolny
and Page 1998).
What is the optimal social structure for
knowledge transfer? Coleman (1988)
claimed that information diffusion is en-
hanced when a network is tight or “closed.”
Closure ensures that those who do not ob-
serve reciprocity norms or who transmit
faulty information will be sanctioned. Burt
(1992), however, argued that an actor’s
informational advantage is maximized when
network ties are diverse and nonredundant.
Participating in multiple networks is con-
sistent with an information-searching ori-
entation that facilitates innovation adop-
tion (Kimberly 1981). Rather than being
incompatible, the two perspectives suggest
a tradeoff between reliable versus original
information. Cultivating ties within a closed
network enhances information reliability
and the ability to effect knowledge transfer
(Hansen 1999). Nonredundant ties to
multiple networks generate large amounts
of information, including new information
that competitors may not possess (Podolny
and Baron 1997).
Professional networks have received the
greatest attention in studies of technologi-
cal diffusion (Coleman, Katz, and Menzel
1966; Crane 1972). Engineers, for example,
regularly engage in trading of know-how:
practical information about implementing
innovations (Schrader 1991; von Hippel
1988; Carter 1989). The procedure starts
when engineers meet at professional gath-
erings. Over time, these contacts allow
engineers to make judgments regarding
the expertise and ability of professional
colleagues (von Hippel 1987). Personal
contact also permits assessments of trust-
worthiness, eventually leading to informa-
tion exchange between suitable dyads
(Heimer 1992). Partners barter informa-
tion, with professional reputation serving
as a mechanism for ensuring that reciproc-
ity obtains. The less likely it is that shared
information will affect a firm’s rents, the
more likely it is that information exchange
will occur. Hence know-how trading is
most prevalent when the traders’ firms are
not direct competitors, when the informa-
tion is not critical to competitive advan-
tage, or when the information can be
learned through experience.
Another type of business-related network
is cross-industry business associations and
civic groups. Employer participation in
civic activities has been found to create
social capital that can be drawn upon to
promote technological diffusion or to solve
collective problems ranging from infrastruc-
ture provision to work force preparedness
(Putnam 1993; Sabel 1993). Employer in-
volvement with local schools promotes pri-
vate training efforts (Rosenbaum and Miller
1997). As for single-industry associations,
although their members compete with each
other, they are likely to use similar pro-
cesses and face similar problems (Westphal
et al. 1997). Thus there is a basis for know-
how trading (Kraatz 1998).
Know-how trading also can occur inside
large organizations. If we view companies
as know-how repositories, then internal
information exchange can be understood
as a type of innovation diffusion across
intrafirm networks (Tsai and Ghoshal
1998). Large firms offset diseconomies of
scale through their information diversity,
although firms differ in their ability to ef-
206 INDUSTRIAL AND LABOR RELATIONS REVIEW
fect know-how trading across subunits
(Hansen 1999).
Finally there are private consultancies,
whose business can be thought of as a net-
work in which all ties radiate from the
consultant. Consultants disseminate prac-
tical knowledge based on what they have
learned from their clients and their aca-
demic contacts (Eccles and Nohria 1992;
Hansen and Hughes 2000). A far-flung
client base allows consultants to span in-
dustries that otherwise might not engage in
know-how trading (Rogers 1983).
A few studies have examined the role of
networks in the diffusion of HR innova-
tions. Membership in a hospital industry
association spurs TQM adoption, although
early adopters are more motivated by effi-
ciency concerns than are late adopters, who
tend to be conformists (Westphal et al.
1997). Cole (1989) contrasted the rapid
diffusion of quality circles in Japan to their
sluggish spread in the United States, where
there were problems in assessing consult-
ant reliability. Quality-oriented U.S. firms
belong to multiple organizations and to
networks organized around suppliers and
customers (Cole 1999).
Hypotheses
There are several reasons to think that
networks play a role in the diffusion of
high-performance practices. First, practi-
cal information about implementing these
practices is not usually in the public do-
main; private networks are more likely to
provide it. Second, workplace innovations
influence but are not a critical determinant
of competitive advantage. Third, human
resource management is a semi-profession,
with organizations that can provide the
social capital necessary for know-how trad-
ing. Fourth, a company’s involvement with
local schools and related activities may pro-
vide information about HR innovations.
These considerations lead to several hy-
potheses regarding the determinants of an
intensive approach to work innovation and
training. Here, intensity has two dimen-
sions: the adoption of multiple high-per-
formance practices, and their application
to a substantial portion of an organization’s
work force. We focus on intensive adop-
tion because it is the route to maximal
productivity gains and because pursuing
this path is likely to require learning from
the experiences of others.
Our main network measure focuses on
employer participation in formal associa-
tions and activities outside the firm, such as
industry, cross-industry, and professional
organizations. Formal organizations pro-
vide pre-existing channels for communica-
tion out of which more specialized net-
works develop. There is evidence that sim-
ply being a member of an association can
facilitate knowledge spillovers that non-
members do not experience (Branstetter
and Sakakibara 1997). We do, however,
have a proxy measure for participation in
informal networks, namely, how extensively
involved the employer is with local schools.
This gives us our first hypothesis:
Hypothesis 1: Employers participating in external
associations and activities are more likely to pursue
an intensive approach to work innovation than are
less affiliated employers.
The network literature construes mul-
tiple affiliations as a way of obtaining
nonredundant information and as an indi-
cator of an external orientation conducive
to learning and innovation. If so, then
multiple affiliations should be associated
with an intensive approach to work innova-
tion.
Hypothesis 2: As the number of affiliations
increases, so does the intensity of work innova-
tion.
The bundling literature suggests that
particular HR policies, especially training,
are complementary to the intensive use of
high-performance work practices. Hence,
employer-provided training is associated
with work-innovation intensity and, indi-
rectly, with external networks. In addition
to their indirect effect, networks may di-
rectly encourage training. First, network
ties—membership affiliations, school in-
volvement, and use of consultants—may
provide information on how training pro-
grams can best be designed and adminis-
tered. Second, networks might help to
EMPLOYER NETWORKS 207
establish reputation effects that inhibit the
poaching of trained employees. Poaching
is a concern for employers because compa-
nies find it difficult to separate provision of
general training from firm-specific train-
ing. Such concerns, if not addressed, will
inhibit training investments (Stevens 1996;
Acemoglu and Pischke 1999).
Hypothesis 3: Employer networks are associated with
employer-provided training.
In addition to external networks, there
are internal networks available to compa-
nies that have multiple units and divisions.
Internal networking across company units
could be a mechanism for know-how trad-
ing about work and training innovations.
Hypothesis 4: Multiple-unit organizations are asso-
ciated with greater intensity of work innovation and
training.
To estimate these hypotheses, we per-
form two sets of regressions: with high-
performance work practices as the depen-
dent variable and with training as the de-
pendent variable in equations that include
high-performance work practices as an ex-
planatory variable. The coefficient on this
latter variable then can be interpreted as
indicating the effect on training of pursu-
ing an increasing number of high-perfor-
mance work practices, independent of the
effects of the other variables in the regres-
sion.
California Workplace Survey
We developed a questionnaire for sur-
veying employers with respect to their busi-
ness policies, HR policies, and work-orga-
nization practices. To select a representa-
tive sample and to administer the mail sur-
vey, we turned to California’s Employment
Development Department (EDD), which
collects data on wages and employment
from each of the state’s 600,000 business
establishments. Nearly 85% of the estab-
lishments employ fewer than 20 individu-
als. These “micro” establishments are very
unstable; many do not last for more than
one year (Davis and Haltiwanger 1992).
Because of their instability and the diffi-
culty of surveying them, we excluded mi-
cro-establishments from our survey. On
the other hand, by relying on the EDD’s
database, we had access to small establish-
ments—those employing 20 to 250 employ-
ees—that are often excluded or poorly cov-
ered in other employer surveys. For rea-
sons of comparability and interpretability,
our survey excluded agriculture and gov-
ernment.
The remaining sub-universe in 1996 com-
prised over 96,000 establishments, account-
ing for 73% of the state’s employment. The
size and industry distributions (at the one-
digit SIC level) of these establishments are
very similar to the national distributions.
We asked EDD to randomly sample ap-
proximately 7% of this sub-universe. We
over-sampled large establishments, send-
ing surveys to all establishments employing
more than 500 workers and to half of those
employing 250–500 workers. The surveys
were accompanied by a cover letter asking
the recipient to have his or her
establishment’s senior human resource
manager complete the survey. The merged
set of survey rounds provided us with 977
responses, representing approximately 1%
of California’s nonagricultural, private es-
tablishments employing over 20 workers.
The aggregate response rate was 10%. While
this response rate may seem low, bear in
mind that our sample included many small
establishments that have particularly low
response rates. They are often ignored in
other studies that focus either on the larg-
est establishments or on the head offices of
multi-establishment companies. Thus we
are able to avoid “headquarters bias”: the
inability of the head office to accurately
aggregate work practices in multiple, local
establishments.
In our data analysis, we assigned stratum
weights to compensate for the nonrep-
resentativeness (by size class) of the sam-
pling design and to account for industry
differences in the survey response rates. A
weight was assigned to each observation in
a given size class and industry cell, indicat-
ing the reciprocal of the probability that an
establishment in that cell was sampled
(Cochran 1977).
The survey asked employers to report on
208 INDUSTRIAL AND LABOR RELATIONS REVIEW
their usage of 10 high-performance work prac-
tices: self-directed teams; total quality man-
agement (TQM); quality circles; cross-train-
ing (in skills for more than one job); com-
pensation based on “pay for knowledge”;
regular meetings of nonsupervisory employ-
ees to discuss work-related problems; job
rotation; flextime; statistical process con-
trol; and peer review of employee perfor-
mance. Summing the number of different
kinds of practices provided one measure of
work-innovation intensity, which other re-
searchers also have used (Osterman 2000).
The survey then asked employers to state
what share of their nonsupervisory employ-
ees were involved in teams, TQM, and regu-
lar employee meetings, respectively. These
coverage statistics provided our second
measure of intensity.
We asked employers to tell us about the
job-skills training they provided (Frazis et al.
1995). Job-skills training upgrades or ex-
tends employee skills, or qualifies workers
for a job. We asked employers which (if
any) of six different types of formal job-
skills training they had provided or financed
in the previous year: management and
supervisory skills; professional and techni-
cal skills; computer skills; office, sales, and
customer-service skills; food, cleaning, and
personal-service skills; and production
skills.1
We measured affiliation in three differ-
ent types of formal organizations: whether
an establishment participated in an em-
ployer or trade association for the industry;
whether it participated in a cross-industry
association or chamber; and whether it was
a member of an educational advisory board.
Because this was an exploratory study, we
used very broad measures of affiliation.
Also, because our unit of analysis was the
establishment, we did not ask respondents
to tell us about their private activities, only
about affiliations maintained by the em-
ployer.
To assess an employer’s informal involve-
ment in civic affairs, we asked about interac-
tion with schools: whether the establishment
provided funds or equipment to schools, or
had internship programs or adopt-a-school
arrangements with local schools. These
variables have been used in previous re-
search (NCEQW 1995).
We also asked whether the establishment
relied on outside consultants to assist with
training programs, quality management,
reengineering, or a combination thereof.
Finally, to proxy the existence of internal
corporate networks, we identified establish-
ments that were part of multi-unit compa-
nies.
Various control variables were included
based on findings of statistical significance
in previous research: establishment size
and age; industry; occupational composi-
tion of employment; share of part-time
workers; and union status.
Two other control variables are a
company’s business strategy and its percep-
tion of skill trends. Previous research sug-
gests that companies competing on quality
and innovativeness are more likely than
price-driven companies to adopt innova-
tive technology and to pursue an intensive
approach to work organization and train-
ing (Osterman 1994; Wood 1995). To mea-
sure strategy, we asked respondents to rank
factors in order of importance to the suc-
cess of the establishment: competitive
prices; overall quality; innovative products
or services; tailoring products to specific
customer needs; and high levels of em-
ployee skill.2 Relatedly, rising skill require-
ments are associated with an intensive ap-
proach to work innovation and training
(MacDuffie 1995). We asked respondents
about their perception of whether skill re-
quirements were rising in their establish-
ment.
1For detailed definitions of these skill sets, see
Frazis et al (1995). Production skills include opera-
tion of machinery or equipment, manufacturing, as-
sembly, distribution, installation, and inspection.
2There were no statistically significant differences
by establishment size in business-strategy factors.
Ranked first was quality (47%), followed by price
(24%) and innovativeness (11%). California employ-
ers were slightly more likely than their national coun-
terparts to emphasize innovation (Osterman 1994).
EMPLOYER NETWORKS 209
Finally, as control variables we asked
employers about career-type policies and
about supportive employment practices like fam-
ily leave and employee assistance programs.
The measure of career-type policies was the
number of the following policies in place
for nonsupervisory employees: pension
plan; internal promotion plan; paid vaca-
tion plan; profit sharing plan; and sever-
ance pay plan. The measure of supportive
employment practices was the number of
the following policies in place: employee
assistance program; family/parental leave
plan; and employee wellness program.
Research suggests that these policies and
practices are complementary to high-per-
formance work systems; they also reduce
turnover and thereby allow employers to
recoup returns on training investments.
Complementarity occurs because compa-
nies that provide an environment that is
secure, supportive, and responsive to em-
ployees are more likely to obtain employee
commitment to high-performance work
practices (MacDuffie 1995; Wood 1999).
Incidence of High-Performance
Practices and Networks
There was a wide range of usage of high-
performance work practices. The most
prevalent practice was cross-training (66%
of establishments); least prevalent were
quality circles (8% of establishments).
Adoption rates in our sample are similar to
those found in other recent surveys. For
example, the percentage of establishments
with 50 or more employees that had self-
directed work teams was 32% nationally
and 30% in our sample. For other prac-
tices, there are slight divergences: Califor-
nia employers with over 50 employees are
slightly less likely to use TQM than are
national employers (33% versus 42%), while
cross-training is slightly more prevalent
(66% versus 56%).
Of the ten practices we studied, 13% of
establishments had adopted one; 21%, two
(the mode); 18%, three; 16%, four; 10%,
five; and 8%, six or more. As for intensity in
terms of employee coverage, participation
rates were found to be lower than might
commonly be supposed, but they were con-
sistent with rates observed in other surveys
(Gittleman et al. 1998; NCEQW 1995). For
establishments with 50 or more workers,
the nonsupervisory employee participation
rate for self-directed work teams was 11%;
for TQM it was 15%; and for regular meet-
ings to discuss work-related problems it was
35%. In other words, a little bit of innova-
tion was quite common. But thorough-
going transformation—as measured by the
number of adopted practices or by the per-
centage of employees affected—was much
less common.
Regarding formal membership networks
(industry, cross-industry, and educational
boards), the most common type was mem-
bership in an industry-specific business as-
sociation (66%); the least common was
membership on an educational advisory
board (22%).
For interaction with schools, compara-
tive data show that California employers
were less likely than their national counter-
parts to provide funds or to offer intern-
ships, and were much less likely to partici-
pate in adopt-a-school programs (Erickson
and Jacoby 1998). The most common type
of involvement was providing schools with
funds or equipment.
High-Performance Work
Practices: Multivariate Analysis
To assess the determinants of work-inno-
vation intensity, we examined an estab-
lishment’s total number of high-perfor-
mance work practices and the share of its
employees covered by three major prac-
tices: self-directed teams, total quality man-
agement (TQM), and meetings to discuss
work-related problems. The results are
from Tobit maximum-likelihood regres-
sions, which are an appropriate technique
when the dependent variable is truncated
at some value, in this case zero at the bot-
tom and either the total number of prac-
tices or 100% at the top.3 For purposes of
3One can think of the latent variable as “employer
preference for innovative work practices,” or, in the
210 INDUSTRIAL AND LABOR RELATIONS REVIEW
comparison, we also include the results of
OLS regressions. All the data in the regres-
sions were weighted by the reciprocal of the
probability of sample selection of the given
establishment size/industry cell (unweight-
ed regressions generally yielded qualita-
tively similar results). The means and stan-
dard deviations for the variables used in the
multivariate analyses are presented in Table
1. The results of the OLS and Tobit regres-
sions are presented in Tables 2a (number
of high-performance work practices), 2b
(share of employees in teams), 2c (share of
employees covered by TQM programs), and
2d (share of employees participating in
regular meetings).
Networks. The effect of formal network
membership on the intensity of high-per-
formance work practices was analyzed by
constructing dummy variables for member-
ship in one, two, or three types of formal
networks, where the network types were an
employer or trade association for the in-
dustry; a cross-industry association or cham-
ber; or participation on an educational
advisory board. Interestingly, the measure
of one network affiliation, which takes a
value of 1 if the establishment participated
in one of these formal activities, is not
significantly positively associated with the
total number of high-performance work
practices in use in a given establishment
(Table 2a). That is, establishments that
participated in a single external group do
not seem to have used more high-perfor-
mance work practices than establishments
that were without external linkage, con-
trary to hypothesis 1. However, the coeffi-
cients on the measure of two network affili-
ations and three network affiliations are
positive and statistically significant at the
10% level or better. That is, participating
in two or three such activities was associ-
ated with more intensive adoption of high-
performance work practices; for example,
establishments that had three network af-
filiations used, on average and controlling
for the other determinants, about 0.8 more
high-performance work practices than did
establishments with no network affiliations.
This is consistent with hypothesis 2: as the
number of network affiliations increases,
so does the intensity of workplace innova-
tion.
Generally speaking, the same results ob-
tain in the estimations reported in Tables
2b, 2c, and 2d, where the dependent vari-
ables are the share of employees engaged
in teams, TQM, and regular meetings, re-
spectively. However, the effect of formal
network participation is generally not sig-
nificant at the 10% level for these measures
of workplace-innovation intensity until an
establishment participates in all three types
of affiliations.4
Our measure of informal social and civic
ties is the extent to which the establishment
interacted with schools. It is the sum of
three dummy variables: whether an estab-
lishment provided funds to educational
institutions, had internship programs, or
participated in adopt-a-school arrange-
ments. Being more involved with schools
was positively and significantly related to
the number of high-performance work prac-
tices an establishment adopted. But it had
a less significant effect on the share of
employees participating in teams, TQM,
and meetings.
To measure whether an establishment
had an opportunity for internal network-
ing inside the corporation, we use a dummy
variable that takes the value 1 when an
establishment was part of a multi-establish-
ment company. The variable is generally
not statistically significant in the regres-
sions. These findings do not provide sup-
case of teams, “employer preference for collective
work practices.” Different employers who choose to
have 0% of their employees in teams may vary in terms
of their “preference for collective work practices”;
that is, the 0% could reflect a wide range of intensity
of non-collective views.
4Alternative specifications and weighting assump-
tions indicate that the findings on the relationship
between the network variables and the number of
high-performance work practices are more robust
than the findings on the relationship between the
network variables and these other three measures of
workplace-innovation intensity.
EMPLOYER NETWORKS 211
port for hypothesis 4, that multiple-unit
organizations are associated with greater
intensity of work innovation.
Finally, we note from the results of like-
lihood ratio tests (χ2 statistics presented at
the bottom of the tables) that the network
variables have a significant joint effect on
all four measures of the adoption and in-
tensity of high-performance work practices,
independent of the other variables in the
regressions.5 Also, additional regressions
(not presented here) indicate that control-
ling for up to 61 two-digit industries (rather
than the 7 one-digit controls included in
the regressions presented in the tables)
generally does not drive the network vari-
ables to zero, indicating that there was also
a within-industry effect of networks on the
adoption and intensity of use of high-per-
formance work practices.
In short, we find the strongest support
for hypothesis 2, regarding multiple affilia-
tion. In most instances, having extra-estab-
lishment ties—either via participating in
formal organizations or being involved in
the community—stimulated the adoption
Table 1. Variable Means and Standard Deviations.
Variable Mean Standard Deviation
High-Performance Work Practices
Number of High-Performance Work Practices (0–10) 2.915 1.901
Share of Employees in Teams .156 .308
Share of Employees Covered by TQM .176 .335
Share of Employees Covered by Regular Meetings .398 .435
Training Practices
Provision of Any Job Skills Training (yes = 1, no = 0) .690 .463
Networks
One Affiliation .300 .459
Two Affiliations .314 .465
Three Affiliations .152 .360
Multiple-Establishment Company .275 .447
Use of Consultants (0–2) .410 .582
Interacts with Schools (0–3) .668 .859
Personnel Policies
Number of Career Employment Practices (0–5) 2.178 1.305
Number of Supportive Policies (0–3) 1.050 1.012
Employment Rising .352 .478
Employment Falling .176 .381
Share of Part-Time .178 .264
Business Strategy and Skill Trends
Rising Skill Requirements .465 .499
Innovative Products/Services (rank 0–5) 3.645 1.404
Other Controls
Establishment Age 18.740 19.288
Establishment Age < 10 Years (yes = 1, no = 0) .389 .488
Union Share .148 .355
Number of Observations 846
5The statistics are for a likelihood-ratio test on the
hypothesis that the network variables are jointly equal
to zero.
212 INDUSTRIAL AND LABOR RELATIONS REVIEW
of work innovations. Moreover, the greater
the number of such linkages, the greater
the intensity of adoption. In particular,
participating in three formal organizations
and being involved with schools were strong
predictors of innovation intensity at the
establishment level.
HR policies. Consistent with the litera-
ture on bundling, we found relationships
between a firm’s personnel policies and its
adoption intensity. Our measure of how
many of five different career employment
policies an establishment used is signifi-
cantly associated with the number of high-
performance work practices (Table 2a) and
with the share of employees in teams (Table
2b). Our measure of the sum of three
different supportive employee policies
(EAP, family leave, and wellness programs)
shows a somewhat different pattern. It is
Table 2a. Determinants of Number of High-Performance Work Practices.
Independent Variable OLS TOBIT
Networks
One Affiliation –.135 –.090
(0.85) (0.53)
Two Affiliations .374 .442
(2.28) (2.49)
Three Affiliations .808 .880
(3.46) (3.50)
Multiple-Establishment Company .178 .205
(1.12) (1.20)
Interacts with Schools .301 .318
(3.60) (3.53)
Personnel Policies
Number of Career Employment Practices .144 .150
(2.36) (2.26)
Number of Supportive Policies .212 .223
(2.49) (2.43)
Share of Part-Time 1.039 1.087
(4.26) (4.14)
Business Strategy and Skill Trends
Rising Skill Requirements .712 .783
(5.85) (5.97)
Innovative Products/Services –.102 –.103
(2.44) (2.26)
Other Controls
Establishment Age –.002 –.002
(0.54) (0.35)
Establishment Age < 10 Years .138 .174
(0.96) (1.12)
Union Share –.240 –.273
(1.09) (1.15)
Number of Observations 846 846
(Pseudo) R-Squared 0.199 0.047
χ2(5) 46.73 43.69
Prob > χ20.000 0.000
Notes: Absolute values of t-statistics in parentheses. Regressions also include a constant term and controls
for 1-digit industry, occupational weights, and establishment size. R-squared reported for OLS regression;
pseudo R-squared for Tobit regression. χ2 statistic is for likelihood-ratio test on the hypothesis that the five
network variables are jointly equal to zero.
EMPLOYER NETWORKS 213
significantly associated with the number of
high-performance work practices, but it is
not significantly associated with the share
of workers in teams or TQM programs.6 A
related variable is the share of part-time
employees in the establishment’s work
force. That variable is positive and statisti-
cally significant in Table 2a, meaning that
as reliance on part-timers increased, so did
the number of high-performance practices.
Business strategy and skill trends. The coef-
ficients on the variable measuring employer
perceptions of rising skill requirements
indicate that a perception of rising skill
requirements was positively associated with
the number of high-performance work prac-
tices and the share of employees covered by
teams and TQM programs. We also find
Table 2b. Determinants of Share of Employees in Teams.
Independent Variable OLS TOBIT
Networks
One Affiliation .025 .132
(0.85) (1.28)
Two Affiliations .020 .116
(0.63) (1.08)
Three Affiliations .117 .441
(2.63) (3.03)
Multiple-Establishment Company –.037 .022
(1.22) (0.21)
Interacts with Schools .016 .075
(1.02) (1.44)
Personnel Policies
Number of Career Employment Practices .030 .080
(2.59) (2.04)
Number of Supportive Policies –.007 –.026
(0.41) (0.48)
Share of Part-Time .013 –.027
(0.28) (0.17)
Business Strategy and Skill Trends
Rising Skill Requirements .039 .119
(1.70) (1.55)
Innovative Products/Services .006 .021
(0.72) (0.76)
Other Controls
Establishment Age .003 .006
(3.61) (2.11)
Establishment Age < 10 Years .043 .147
(1.59) (1.60)
Union Share –.006 .037
(0.15) (0.27)
Number of Observations 846 846
(Pseudo) R-Squared 0.098 0.089
χ2(5) 12.29 13.73
Prob > χ20.031 0.017
Notes: See notes to Table 2a.
6It has some positive association with the share of
employees participating in meetings, with statistical
significance at the 10% level in the OLS regression
(Table 2d).
214 INDUSTRIAL AND LABOR RELATIONS REVIEW
significant effects for the business strategy
variable: an emphasis on innovative prod-
ucts or services. Innovativeness was signifi-
cantly associated with the number of high-
performance work practices (Table 2a; the
lower the value of this variable, the more
highly it was ranked by employers in terms
of its strategic importance) and was also
significantly associated with a greater share
of employees participating in TQM, but
not in teams or meetings.
Other control variables. We included con-
trols for establishment age (in years) and a
dummy variable for whether the establish-
ment had existed for less than 10 years.
The point of using the latter variable was to
get a rough sense of whether there was a
“greenfield” effect. The coefficients on
both variables tend to be either statistically
insignificant or significantly positive in the
regressions, providing mixed support for
the oft-asserted claim that workplace inno-
vation is more likely to be introduced in
newer “greenfield” establishments.
Unionization. The literature is divided
on the relationship between unionization
Table 2c. Determinants of Share of Employees Covered by TQM.
Independent Variable OLS TOBIT
Networks
One Affiliation –.005 .013
(0.19) (0.10)
Two Affiliations .051 .152
(1.73) (1.20)
Three Affiliations .120 .453
(2.83) (2.65)
Multiple-Establishment Company –.019 .073
(0.66) (0.61)
Interacts with Schools –.002 .067
(0.15) (1.06)
Personnel Policies
Number of Career Employment Practices .014 .018
(1.29) (0.39)
Number of Supportive Policies .022 .071
(1.45) (1.12)
Share of Part-Time –.013 –.022
(0.29) (0.12)
Business Strategy and Skill Trends
Rising Skill Requirements .039 .187
(1.77) (2.04)
Innovative Products/Services –.018 –.055
(2.34) (1.75)
Other Controls
Establishment Age –.000 –.002
(0.09) (0.58)
Establishment Age < 10 Years .025 .035
(0.96) (0.31)
Union Share –.005 –.023
(0.13) (0.14)
Number of Observations 846 846
(Pseudo) R-Squared 0.068 0.049
χ2(5) 14.18 12.19
Prob > χ20.015 0.032
Notes: See notes to Table 2a.
EMPLOYER NETWORKS 215
and workplace innovation, with some stud-
ies finding a positive association (Kelley
and Harrison 1992) and others finding in-
novation more closely associated with a
nonunion approach (Kochan, Katz, and
McKersie 1986). Reflecting this divide, in
our study the share of employees covered
by collective bargaining (union share) has
positive or negative coefficients in differ-
ent regressions but is never statistically sig-
nificant at the 10% level.7
Training: Multivariate Analysis
We also examined whether network par-
ticipation was related to corporate training
programs, as conjectured in hypothesis 3.
We asked employers whether they provided
any of six different types of job-skills train-
ing (management, professional, computer,
other clerical and sales, production, and
Table 2d. Determinants of Share of Employees Involved with Regular Meetings.
Independent Variable OLS TOBIT
Networks
One Affiliation –.030 –.042
(0.75) (0.36)
Two Affiliations .074 .166
(1.74) (1.39)
Three Affiliations .129 .350
(2.13) (2.09)
Multiple-Establishment Company .038 .145
(0.93) (1.28)
Interacts with Schools .020 .107
(0.92) (1.78)
Personnel Policies
Number of Career Employment Practices .003 .010
(0.16) (0.23)
Number of Supportive Policies .039 .072
(1.75) (1.18)
Share of Part-Time .066 .158
(1.04) (0.90)
Business Strategy and Skill Trends
Rising Skill Requirements .015 .116
(0.49) (1.32)
Innovative Products/Services .013 .025
(1.17) (0.80)
Other Controls
Establishment Age .002 .004
(1.64) (1.23)
Establishment Age < 10 Years .045 .145
(1.21) (1.36)
Union Share –.063 –.184
(1.10) (1.14)
Number of Observations 846 846
(Pseudo) R-Squared 0.070 0.030
χ2(5) 15.43 14.91
Prob > χ20.009 0.011
Notes: See notes to Table 2a.
7In addition to controls for 1-digit industry and
occupational shares, we measured establishment size
with four categorical variables (these coefficients are
not reported in the tables). We did not find strong
establishment size effects, once other variables had
been controlled for.
216 INDUSTRIAL AND LABOR RELATIONS REVIEW
personal service skills). For the purposes of
this analysis, we focused on whether an
establishment provided any job-skills train-
ing: that is, the binary decision whether or
not to provide any one or more of the six
types of training. We estimated this binary
decision using a logistic regression; the
results are presented in Table 3.
Because the bundling literature suggests
a relationship between training and high-
performance work practices, we included
the number of high-performance work prac-
tices as an independent variable in the
equation. As previously noted, the coeffi-
cient on the number of high-performance
work practices can be interpreted as indi-
cating the effect on training of pursuing an
increasing number of high-performance
work practices, independent of the effects
of the other variables in the regression.
Note as well that the effects of the other
variables in the training regressions are
mitigated by the fact that they also contrib-
ute to the adoption of high-performance
work practices.
Two new variables are introduced here.
We ask whether an establishment relied on
a consultant to provide training expertise.
We hypothesize that consultants are a source
of the same kind of practical, semi-propri-
etary information that can be obtained
through social networks, and should have a
positive effect on training. Our measure of
use of consultants is whether the establish-
ment used private training/redevelopment
consultants, or quality management/
reengineering consultants, or both. We
also asked whether employment at the es-
tablishment had significantly risen or fallen
in the previous year, so as to control for
employment trends that might affect train-
ing provision.
Networks. The effect of formal network
ties on training—as measured by the three
membership affiliation variables—is mod-
estly statistically significant. As we found in
the previous section, belonging to a single
network did not have a statistically significant
effect; belonging to multiple networks had
positive and statistically significant effects.
Our measure of informal social and civic
ties—interaction with schools—is also sig-
nificantly associated with provision of train-
ing. As discussed below, the relationship is
open to multiple interpretations: that em-
ployers who were involved with schools were
more likely to come into contact with other
employers in ways that promoted training;
that school involvement made employers
confident of graduates’ quality and thus
more likely to train; or that training, in-
stead of being a dependent variable, was
causing employers to become more involved
with schools so as to reap a better payoff on
their investments.
Being part of a multi-establishment com-
pany was significantly related to the provi-
sion of training. This could be an informa-
tion effect, or perhaps it indicates that there
are economies of scale in training such that
large companies do more of it.
Finally, we found very strong results for
use of consultants, which we interpret as an
information effect: consultants offer both
technical know-how and political advice for
legitimating decisions to spend more money
on training. However, a skeptic might say
that this is simply a case of consultants
boosting their fees by promoting more in-
tensive training than otherwise would oc-
cur.
As in the workplace innovation regres-
sions, the likelihood ratio test result sum-
marized by the χ2 statistic at the bottom of
the table indicates that the network vari-
ables are jointly significant at a very high
level of statistical significance. This sug-
gests that the network variables do not play
a purely mediational role, but rather have
an impact on training that is independent
of the other variables in the regressions.
And, as in the workplace innovation regres-
sions, additional regressions (not reported
here) indicate that networks play a role
even after we control for detailed 2-digit
industry effects.
High-performance work practices. Consis-
tent with the “bundling” literature, the
number of high-performance work prac-
tices was positively and statistically signifi-
cantly related to the provision of training
(Table 3). Keep in mind that we are mea-
suring job-skills training, that is, training
that upgrades or extends employee skills or
EMPLOYER NETWORKS 217
qualifies workers for a job. This training is
different from—and was measured by us
separately from—the interpersonal and
behavioral training associated with the in-
troduction of high-performance work prac-
tices (for example, group problem-solving,
leadership, and techniques of quality im-
provement). The results suggest that the
use of an increasing number of high-per-
formance practices went hand-in-hand with
job-skills training, although it is difficult to
determine the direction of causality.
Personnel policies. Career-type employ-
ment policies were not significantly associ-
ated with provision of training once we
include the network variables. Policies to
promote a supportive work environment,
however, were strongly associated with the
provision of training. The rising employ-
ment variable did not have a significant
effect on provision of training, while the
falling employment variable had a moder-
ately negative effect. Finally, we did not
find a significant relationship between use
of part-time workers and the provision of
training.
Business strategy and skill trends. A percep-
tion that skill requirements were rising was
positively and significantly associated with
provision of training. But pursuit of a
business strategy based on innovation had
no association with provision of training.
Additional control variables. In contrast to
other studies (for example, Bishop 1994;
Lynch and Black 1998), we do not find
statistically significant establishment size
effects for training provision. Unionism is
not statistically significant. Older establish-
ments were less likely to provide training.
Interviews
Having found a relationship between
networks and workplace innovation, we
Table 3. Determinants of Provision of Job-Skills Training.
(Logistic Estimates)
Networks
One Affiliation –.248
(1.08)
Two Affiliations .444
(1.86)
Three Affiliations .751
(1.92)
Multiple-Establishment .929
Company (3.61)
Use of Consultants 1.86
(7.36)
Interacts with Schools .389
(2.80)
High-Performance Work Practices
Number of High-Performance .119
Work Practices (2.28)
Personnel Policies
Number of Career .079
Employment Practices (0.85)
Number of Supportive Policies .301
(2.21)
Employment Rising –.184
(0.92)
Notes: Absolute value of t-statistics in parentheses. Regression also includes a constant term and controls for
1-digit industry, occupational weights, and establishment size. χ2 statistic is for likelihood-ratio test on the
hypothesis that the six network variables are jointly equal to zero.
Employment Falling –.417
(1.63)
Share of Part-Time .508
(1.40)
Business Strategy and Skill Trends
Rising Skill Requirements .792
(4.28)
Innovative Products/Services .043
(0.69)
Other Controls
Establishment Age –.015
(2.17)
Establishment Age < 10 Years .032
(0.15)
Union Share –.160
(0.47)
Number of Observations 846
Pseudo R-Squared 0.254
Log Likelihood –419.649
χ2(6) 118.07
Prob > χ20.000
218 INDUSTRIAL AND LABOR RELATIONS REVIEW
wanted to learn more about the meaning of
the observed relationship. We also wanted
to study how these networks—membership
affiliations, multi-unit firms, school inter-
actions—were related to organizational
learning. Can we infer causality from the
former to the latter? Finally, we wondered
why some organizations maintained numer-
ous affiliations while others had few or no
external linkages.
With these questions in mind, we ran-
domly selected 75 establishments for tele-
phone interviews. We were able to conduct
interviews with the manager in charge of
human resources at 62 of these 75 estab-
lishments; in nine cases, we conducted fol-
low-up interviews.8 The interviews followed
a script of open-ended questions. Qualita-
tive research tools were used to analyze the
interviews: counting phenomena, noting
themes, and comparing groups. Bear in
mind that the purpose of the interviewing
was not hypothesis testing but an effort to
interpret our quantitative findings (Miles
and Huberman 1984; Coffey and Atkinson
1996).
Did membership affiliation play a role in
organizational learning? The interviews
suggest that it did. Forty-five percent of
those belonging to an industry association
mentioned it as an information source on
some aspect of HRM; for cross-industry as-
sociations and educational boards, the cor-
responding figures were 33% and 10%.
Industry associations were of two types:
those limited to HR managers, and those
including a variety of managers. From these
associations interviewees said that they
learned about industry-specific compensa-
tion issues, legal regulations, and “best prac-
tice” for work organization. Those citing
industry associations as being useful for
providing information on “best practice”
said that this occurred through relation-
ships formed in face-to-face interactions
with other managers. Typical in this regard
was a manager who participated in a
healthcare-industry association. He said
that information regularly passed between
him and other HR managers he had met
through the association because they had
known each other for years and had devel-
oped close relationships.
Cross-industry groups were also of two
types: professional associations like the
Society for Human Resource Management,
and general business groups like state and
local employers’ associations and cham-
bers of commerce. Only two managers
cited professional associations as sources of
HR information. The greatest number of
positive remarks about professional asso-
ciations referred to their value in advanc-
ing a manager’s career by helping to build
personal skills, providing professional cer-
tification, or creating a network that could
be useful when seeking a new job. More
frequently cited as a source of HR informa-
tion were employers’ associations. Single-
establishment companies were more likely
than others to cite the information value of
these associations, perhaps because they
provide expertise that larger firms are able
to supply internally. Another explanation
was provided by the HR manager of a small
optical coatings company, who said that the
employers’ association had larger compa-
nies belonging to it than the industry asso-
ciation and that “this gives me something to
look to as a model.”
Only three of the six respondents from
establishments belonging to an educational
advisory board made reference to it with
respect to HR information. One was the
HR manager of a medical equipment firm
that had adopted a panoply of advanced
work practices. He gave three reasons for
belonging to the board of his local commu-
nity college: he felt that employers should
be involved in community activities; he
wanted to stay apprised of educational de-
velopments; and he wanted to influence
curricula in ways that would benefit the
company (and presumably lower its train-
ing costs; the firm is currently hiring 200
new workers each year).
Similar motives were expressed with re-
8Eighteen percent of the interviewed establish-
ments had either one membership affiliation or none;
32% had two affiliations; and 50% had three affilia-
tions.
EMPLOYER NETWORKS 219
gard to involvement in other school-related
activities. Managers said they had intern-
ship programs or provided resources to
local schools because these helped in re-
cruiting (“the earlier the better”), kept them
abreast of hiring pool characteristics, and
allowed the company to “act out its altruism
and contribute to society.”
Although our multivariate analyses found
little support for a relationship between
high-performance work practices and mem-
bership in a multi-unit enterprise, the in-
terviews revealed that some of the most
intensive trading of information occurred
between establishments within the same
firm. Three-fourths of the managers from
multi-unit firms mentioned the opportuni-
ties for learning that arose as a result of
belonging to a multi-unit entity—“a huge
pool of internal expertise,” one manager
called it. Sometimes the learning was infor-
mal, as when an HR manager in a medium-
sized manufacturing company credited
quality-management tips received from a
colleague at a sister plant with saving his
own plant from having to shut down. In
other companies, the headquarters orga-
nized formal meetings, taskforces, and con-
ferences to integrate HR managers and get
them exchanging ideas with each other.
Half of the managers in this group said that
they adopted HR innovations as a result of
conscious efforts by the company to com-
pare units with one another and to bench-
mark the HR practices of the best-perform-
ing units.
Managers from multi-unit firms reported
high levels of engagement with industry,
professional, and community organiza-
tions.9 Organizational learning inside multi-
unit companies appears to have been a
complement to, not a substitute for, learn-
ing from entities outside the company. The
companies routinely paid any dues or ex-
penses associated with affiliation, even for
professional associations that might advance
a manager’s outside career. Some multi-
unit firms identified organizations that they
thought were beneficial to join and then
assigned different managers the responsi-
bility for participating in them.
Interviewees had diverse opinions about
consultants, with roughly a third saying
that they disdained them, a third saying
that they regularly relied on them, and
another third falling in-between. Those
who valued consultants said that they
brought in fresh ideas or were useful for
providing benchmarks about “best prac-
tice” at other companies. We found that
HR managers regularly discussed with each
other their experiences with particular con-
sultants.
Did the ties that existed at networked
companies precede or follow innovation?
Respondents at two-thirds of the establish-
ments with membership affiliations men-
tioned specific HR innovations that were
adopted as a result of prior membership;
none mentioned affiliation occurring after
adoption. With respect to school interac-
tions, however, we did uncover three in-
stances in which ties were formed after
training and work-organization innovations
had been introduced. Presumably employ-
ers became more involved in schools so as
to garner a higher return on their organiza-
tional investments.
Companies with few or no affiliations
were a diverse group, including single-es-
tablishment businesses and high-labor-turn-
over firms in the fast food and construction
industries. The interviews revealed one
common factor: a tendency to focus on
short-term management issues, whether
because of the nature of the business (con-
struction), the high turnover of employees
(food services), or the lack of managerial
capacity and specialization (small indepen-
dent firms).
Conclusions
Our study provides one answer to the
question of why some firms but not others
go down the high-performance road. We
9The survey data corroborate this: multi-unit com-
panies were more likely to be affiliated with industry
and cross-industry associations and to interact with
local schools.
220 INDUSTRIAL AND LABOR RELATIONS REVIEW
find that managerial participation in net-
works positively affected the adoption and
the intensity of adoption of high-perfor-
mance work practices and employee train-
ing programs in the firms in our sample.
The networks we examined included mem-
bership in industry and cross-industry asso-
ciations, school interactions, and the inter-
nal networks of multi-unit firms. We also
found evidence of a multiplex effect: mul-
tiple affiliations—participating in more
than one network—raised the likelihood
that an establishment pursued an intensive
approach to work reorganization and train-
ing.
From a theoretical perspective, our find-
ings underscore the importance of social
relationships in the process of innovation
diffusion. While markets matter a great
deal in providing incentives for innovation,
there are other influential factors as well.
Employer networks—social ties to individu-
als outside the establishment—play an im-
portant role in innovation diffusion and
organizational learning. Participating in
these networks offers opportunities and
creates skills necessary for trading knowl-
edge about innovation implementation and
innovation results. Many of the ideas being
shared are not available on the open mar-
ket. While consultants purvey some of this
knowledge, there is reluctance to hire them.
A skeptic might be unconvinced by our
argument that employer networks are a
source of economically useful information
that facilitates the adoption of high-perfor-
mance work practices and training. Rather,
the skeptic might argue that our various
network measures are simply picking up
legitimation effects: managers who are
networked are more susceptible to con-
formism. Wanting to appear au courant in
the eyes of their peers and the public, they
adopt faddish HR innovations to bolster
their public and professional images. They
care less or not at all about whether these
innovations have any demonstrable rela-
tionship to performance (DiMaggio and
Powell 1983)
We have several responses to the skep-
tics. First, numerous empirical studies have
consistently found a strong relationship
between adoption of high-performance
work practices and establishment-level out-
comes such as productivity improvement.
Second, most companies are not pursuing
high-performance work innovations on the
cheap. Our regressions show that as the
intensity of workplace innovation rose,
employers were more likely to invest in
training. Employee training is not faddish
and it is costly. Third, our data show a
positive and statistically significant correla-
tion between the number of adopted high-
performance practices and the share of
employees covered by those practices.10
That is, companies that were the most inno-
vative were more likely to be spreading
innovations across a larger proportion of
employees. As with training, they were
backing up innovations with real resources.
From a practical perspective, our re-
search suggests that organizational learn-
ing can be promoted by a corporate ethos
emphasizing openness and connectedness
and by having companies encourage man-
agers to be “joiners” who are involved in
industrial, professional, and civic organiza-
tions. Not only do managers who partici-
pate in multiple networks acquire useful
information, they also build skills for trad-
ing information and assessing its reliability.
Unfortunately, the employers who might
benefit the most from these activities—
smaller, single-unit companies—typically
feel that they do not have the resources to
engage in networking activities. Govern-
ment and unions might have a role to play
here in providing small business with infor-
mation on best HR practices. Indeed, it is
a familiar role for unions in such industries
as construction and apparel. For multi-
unit companies, our research suggests an
additional recommendation: strengthen-
ing corporate headquarters’ role in pro-
moting the internal exchange of knowl-
edge.
10The point correlations are .33 with teams, .38
with TQM, and .38 with regular meetings. All are
significant at the .001 level.
EMPLOYER NETWORKS 221
Our study has examined several kinds of
employer networks, but there are others
that might be relevant to understanding
the process by which firms adopt HR inno-
vations. Two of our interviewees empha-
sized the value of relationships they formed
while in graduate school. Other networks
include those that link suppliers to each
REFERENCES
other and to a purchaser, and civic organi-
zations that bring employers together to
work cooperatively not only to reform
schools but also to improve transportation,
housing, and other local institutions (Heinz
Endowments 1999). Finally, we need more
studies that analyze union contributions—
actual and potential—in this area.
Acemoglu, Daron, and Jorn-Steffen Pischke. 1999.
“Beyond Becker: Training in Imperfect Labour
Markets.” Economic Journal, Vol. 109 (February), pp.
F112–F142.
Becker, Brian, and Barry Gerhart. 1996. “The Impact
of Human Resource Management on Organizational
Performance.” Academy of Management Journal, Vol.
39 (August), pp. 779–801.
Bishop, John. 1994. “The Incidence of and Payoff to
Employer Training.” Center for Advanced Human
Resource Studies, Cornell University, Working Pa-
per 94–17.
Branstetter, Lee, and Mariko Sakakibara. 1998. “Japa-
nese Research Consortia: A Microeconomic Analy-
sis of Industrial Policy.” Journal of Industrial Econom-
ics, Vol. 46 (June), pp. 207–33.
Burt, Ronald S. 1992. Structural Holes: The Social
Structure of Competition. Cambridge: Cambridge
University Press.
Carter, Anne P. 1989. “Knowhow Trading as Eco-
nomic Exchange.” Research Policy, Vol. 18, pp. 155–
63.
Cochran, William G. 1977. Sampling Techniques. New
York: Wiley.
Coffey, Amanda, and Paul Atkinson. 1996. Making
Sense of Qualitative Data: Complementary Research
Strategies. Thousand Oaks, Calif.: Sage.
Cole, Robert E. 1989. Strategies for Learning: Small-
Group Activities in American, Japanese, and Swedish
Industry. Berkeley: University of California Press.
____. 1999. Managing Quality Fads: How American
Business Learned to Play the Quality Game. New York:
Oxford University Press.
____. 1998. “Learning from the Quality Movement:
What Did and Didn’t Happen and Why?” California
Management Review, Vol. 41, No. 1 (Fall), pp. 43–73.
Coleman, James. 1988. “Social Capital in the Cre-
ation of Human Capital.” American Journal of Sociol-
ogy, Vol. 94, supp. (July), pp. S95–S121.
Coleman, James, Elihu Katz, and Herbert Menzel.
1966. Medical Innovation: A Diffusion Study. India-
napolis: Bobbs-Merrill.
Crane, Diana. 1972. Invisible College: Diffusion of
Knowledge in Scientific Communities. Chicago: Uni-
versity of Chicago Press.
Davis, Steven J., and John Haltiwanger. 1992. “Gross
Job Creation, Gross Job Destruction, and Employ-
ment Reallocation.” Quarterly Journal of Economics,
Vol. 107, No. 3 (August), pp. 819–63.
DiMaggio, Paul J., and Walter W. Powell. 1983. “The
Iron Cage Revisited: Institutional Isomorphism and
Collective Rationality in Organizational Fields.”
American Sociological Review, Vol. 48, No. 2 (April),
pp. 147–60.
Dunlop, John T., and David Weil. 1996. “Diffusion
and Performance of Modular Production in the
U.S. Apparel Industry.” Industrial Relations, Vol. 3,
No. 3 (July), pp. 334–55.
Eccles, Robert G., and Nitin Nohria. 1992. Beyond the
Hype: Rediscovering the Essence of Management. Bos-
ton: Harvard Business School Press.
Erickson, Christopher L., and Sanford M. Jacoby,
with Richard Copeland, Peter Goldschmidt, and
Jonathan Troper. 1998. “Training and Work Orga-
nization Practices of Private Employers in Califor-
nia.” Policy Research Report, California Policy Semi-
nar, Berkeley.
Fligstein, Neil. 1990. The Transformation of Corporate
Control. Cambridge: Harvard University Press.
Frazis, Harley, Diane Herz, and Michael Horrigan.
1995. “Employer-Provided Training: Results from
a New Survey.” Monthly Labor Review, Vol. 118, No.
5 (May), pp. 3–17.
Galaskiewicz, Joseph. 1985. “Professional Networks
and the Institutionalization of a Single Mind Set.”
American Sociological Review, Vol. 50 (October), pp.
639–58.
Gittleman, Maury, Michael Horrigan, and Mary Joyce.
1998. “‘Flexible’ Workplace Practices: Evidence
from a Nationally Representative Survey.” Industrial
and Labor Relations Review, Vol. 52, No. 1 (October),
pp. 99–115.
Hansen, Morton T. 1999. “The Search-Transfer
Problem: The Role of Weak Ties in Sharing Knowl-
edge across Organizations.” Administrative Science
Quarterly, Vol. 44, No. 1 (March), pp. 82–111.
Hansen, Steve C., and John S. Hughes. 2000. “Con-
sultants, Exclusivity Agreements, and Product Mar-
ket Competition.” Working paper, Anderson School,
UCLA, October.
Heimer, Carol A. 1992. “Doing Your Job and Helping
Your Friends: Universalistic Norms about Obliga-
tions to Particular Others in Networks.” In Nitin
Nohria and Robert G. Eccles, eds., Networks and
222 INDUSTRIAL AND LABOR RELATIONS REVIEW
Organizations: Structure, Form, and Action. Boston:
Harvard Business School Press, pp. 143–62.
Heinz Endowments. 1999. Innovative Regions: The
Importance of Place and Networks in the Innovative
Economy. Pittsburgh: Heinz Endowments.
Huselid, Mark A. 1995. “The Impact of Human
Resource Management Practices on Turnover, Pro-
ductivity, and Corporate Financial Performance.”
Academy of Management Journal, Vol. 38, No. 3 (June),
pp. 635–72.
Ichniowski, Casey, Kathryn Shaw, and Giovanna
Prennushi. 1997. “The Effects of Human Resource
Management Practices on Productivity: A Study of
Steel Finishing Lines.” American Economic Review,
Vol. 87, No. 3 (June), pp. 291–313.
Kelley, Maryellen, and Bennett Harrison. 1992.
“Unions, Technology, and Labor-Management Co-
operation.” In Lawrence Mishel and Paula Voos,
eds., Unions and Economic Competitiveness. Washing-
ton, D.C.: Economic Policy Institute, pp. 247–86.
Kimberly, John. 1981. “Managerial Innovation.” In
Paul C. Nystrom and William H. Starbuck, eds.,
Handbook of Organizational Design. New York: Ox-
ford University Press, pp. 84–104.
Kochan, Thomas A., Harry C. Katz, and Robert
McKersie. 1986. The Transformation of American
Industrial Relations. New York: Basic Books.
Kraatz, Matthew S. 1998. “Learning by Association?
Interorganizational Networks and Adaptation to
Environmental Change.” Academy of Management
Journal, Vol. 41, No. 6 (December), pp. 621–43.
Lynch, Lisa, and Sandra Black. 1998. “Beyond the
Incidence of Employer-Provided Training.” Indus-
trial and Labor Relations Review, Vol. 52, No. 1 (Octo-
ber), pp. 64–81.
MacDuffie, John Paul. 1995. “Human Resource
Bundles and Manufacturing Performance: Organi-
zational Logic and Flexible Production Systems in
the World Auto Industry.” Industrial and Labor
Relations Review, Vol. 48, No. 2 (January), pp. 197–
221.
Miles, Matthew B., and A. Michael Huberman. 1984.
Qualitative Data Analysis: A Sourcebook of New Meth-
ods. Beverly Hills, Calif.: Sage.
National Center on the Educational Quality of the
Work Force (NCEQW). 1995. “First Findings from
the EQW National Employer Survey Administered
by the Bureau of the Census.” Philadelphia.
Nohria, Nitin, and Ranjay Gulati. 1994. “Firms and
Their Environments.” In Neil J. Smelser and Rich-
ard Swedberg, eds., The Handbook of Economic Sociol-
ogy. Princeton: Princeton University Press, pp. 529–
55.
Osterman, Paul. 1994. “How Common Is Workplace
Transformation and Who Adopts It?” Industrial and
Labor Relations Review, Vol. 47, No. 2 (January), pp.
173–88.
____. 2000. “Work Reorganization in an Era of
Restructuring: Trends in Diffusion and Effects on
Employee Welfare.” Industrial and Labor Relations
Review, Vol. 53, No. 2 (January), pp. 179–96.
Pfeffer, Jeffrey, and Robert I. Sutton. 1999. “Know-
ing ‘What’ to Do Is Not Enough: Turning Knowl-
edge into Action.” California Management Review,
Vol. 42 (Fall), pp. 83–108.
Pil, Fritz, and John Paul MacDuffie. 1996. “The
Adoption of High-Involvement Work Practices.”
Industrial Relations, Vol. 35, No. 3 (July), pp. 423–55.
Podolny, Joel M., and K. L. Page. 1998. “Network
Forms of Organization.” Annual Review of Sociology,
Vol. 24, pp. 57–76.
Podolny, Joel M., and James N. Baron. 1997. “Re-
sources and Relationships: Social Networks and
Mobility in the Workplace.” American Sociological
Review, Vol. 62, No. 4 (October), pp. 673–93.
Powell, Walter W. 1990. “Neither Market nor Hierar-
chy: Network Forms of Organization.” In Barry
Staw and Larry L. Cummings, eds., Research in Orga-
nizational Behavior, Vol. 12. Greenwich, Conn.: JAI
Press, pp. 295–336.
Putnam, Robert D. 1993. Making Democracy Work:
Civic Traditions in Modern Italy. Princeton: Princeton
University Press.
Rogers, Everett M. 1983. Diffusion of Innovations, 3rd
ed. New York: Free Press.
Rosenbaum, J. E., and S. R. Miller. 1997. “Hiring in
a Hobbesian World: Social Infrastructure and Em-
ployers’ Use of Information.” Work and Occupations,
Vol. 24, No. 4 (November), pp. 498–523.
Sabel, Charles. 1993. “Studied Trust: Building New
Forms of Cooperation in a Volatile Economy.”
Human Relations, Vol. 46, No. 9 (September), pp.
1133–70.
Schrader, Stephan. 1991. “Informal Technology
Transfer between Firms: Cooperation through In-
formation Trading.” Research Policy, Vol. 20 (April),
pp. 153–70.
Stevens, Margaret. 1996. “Transferable Training and
Poaching Externalities” In Alison Booth and Den-
nis Snower, eds., Acquiring Skills: Market Failures,
Their Symptoms, and Policy Responses. Cambridge:
Cambridge University Press, pp. 19–40.
Strang, David, and Sarah Soule. 1998. “Diffusion in
Organizations and Social Movements: From Hybrid
Corn to Poison Pills.” Annual Review of Sociology, Vol.
24, pp. 265–90.
Tsai, Wenpin, and Sumantra Ghoshal. 1998. “Social
Capital and Value Creation: The Role of Intrafirm
Networks.” Academy of Management Journal, Vol. 41,
No. 4 (August), pp. 464–76.
Uzzi, Brian. 1996. “The Sources and Consequences
of Embeddedness for the Economic Performance of
Organizations: The Network Effect.” American So-
ciological Review, Vol. 61, No. 4 (August), pp. 674–98.
____. 1997. “Social Structure and Competition in
Interfirm Networks: The Paradox of Embed-
dedness.” Administrative Science Quarterly, Vol. 42,
No. 1 (March), pp. 35–67.
von Hippel, Eric. 1987. “Cooperation between Ri-
vals: Informal Know-How Trading.” Research Policy,
Vol. 16, pp. 291–302.
____. 1988. The Sources of Innovation. New York:
Oxford University Press.
Weick, Karl E., and Frances Westley. 1996. “Organi-
zational Learning: Affirming an Oxymoron.” In S.
Clegg, C. Hardy, and W. Nord, eds., Handbook of
EMPLOYER NETWORKS 223
Organization Studies. Thousand Oaks, Calif.: Sage,
pp. 440–58.
Westphal, James D., Ranjay Gulati, and Stephen
Shortell. 1997. “Customization or Conformity? An
Institutional Perspective on the Content and Con-
sequences of TQM Adoption.” Administrative Sci-
ence Quarterly, Vol. 47, No. 1 (March), pp. 366–94.
Wood, Stephen J. 1999. “Family Friendly Manage-
ment: Testing the Various Perspectives.” Na-
tional Institute Economic Review (April), pp. 99–
116.
____. 1995. “The Four Pillars of HRM: Are They
Connected?” Human Resource Management Journal,
Vol. 5 (Autumn), pp. 49–59.
