THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN
RESOURCES PROJECT:
Second Interim Report
CSM-32
Clair Brown, Editor
2.
Innovative Human Resource Policies and Work Practices
Vincent M. Valvano
2.1 Introduction
2.2 Distribution of Work Practices Across Fabs
2.3 Distribution of Compensation Policies Across Fabs
2.4 Policy and Practice Comparisons Across Fab
Categories
2. 5Policies and Practices Across Occupational
Categories
2.6 Summary
2.1 Introduction
A variety of factors may influence the set of human resource (HR)
practices adopted by a wafer fab. Fab managers may have more or
less autonomy to implement human resource policies, depending on
the degree to which such policies are set at the corporate level.
Casual observation at the fabs we have visited suggests that fab
managers in the United States have relatively more autonomy in this
respect than their counterparts in Asia, where a range of policies
concerning training, work organization and employment security are
set at the corporate level. In the U.S., with a few notable exceptions,
only compensation policy is generally a corporate-level management
prerogative, while significant control over the implementation of
particular work practices resides with fab-level management.
Characteristics of the major process and product technologies in
place at a fab may also influence HR practices. Seven of the 15
fabs in the Human Resources subsample are primarily or exclusively
logic fabs, 5 are memory fabs and 3 fabs produce older bipolar /
analog products. (Two of the logic fabs also produce memory chips.)
In general, production volume for a given product is higher in memory
fabs than in logic fabs. Logic fabs are likely to have more active
die types in production at any particular time, and hence smaller
average volumes for particular products compared to memory fabs.
Consequently, equipment settings are reprogrammed and photo masks
are generally changed more frequently in logic fabs, and more tests
are carried out for a given level of production, relative to memory
fabs. This level of "customized" production that is typical
of logic fabs affects their performance. In the full CSM sample,
memory fabs tend to outperform logic fabs in most performance metrics,
including yield, labor productivity, and stepper throughput, at
least in part due to simple economies of scale. A preliminary comparison
of human resource policies and practices at memory and logic fabs
is presented here.
Participating fabs were asked about current practice and future
intentions with respect to a range of innovative human resource
policies and practices. Our goal was to get a general indication
of which practices have been widely adopted in the industry and
which have not, as well as a record of the timing of adoption. We
presented a set of innovative human resource policies and asked
fab management if such policies were currently in place, had been
previously attempted and abandoned, or were planned for future adoption.
(Table 2-1 defines the policies and practices.)
Because some of these policies apply to a specific occupational
category and others could, in practice, be applied selectively across
occupations, we asked respondents to identify practices in place
for each of the principal occupations in the fabCoperators, technicians,
and engineers. Subsequent questions in the Human Resources Questionnaire
examine the intensity of specific practices such as training and
employee participation in teams. Eventually, as our sample of fabs
increases, we hope to be able to assess the relationship between
fab HR policies and performance.
Recent research on firm adoption of innovative HR policies has focused
on the extent to which particular policies are complimentary and
might be expected to be adopted in bundles. For example, a policy
of employment security might be implemented in tandem with policies
of cross-training, continuous improvement, and knowledge pay. As
Figures 2-1, 2-2,
and 2-3 indicate, innovative training
policies are in place at nearly all of the fabs in our sample. Employment
security, job rotation, and quality improvement teams are also widely
diffused policies. However, various types of compensation policies
are more selectively applied by particular fabs (Table
2-2). Only at several Asian fabs in the sample were policies
of training, job rotation, and employment security implemented in
a way that suggested management recognized and was trying to capture
the benefits from potential complimentarities between these policies.
Given that the fabs in our sample are spread across three continents,
it is likely that national labor market institutions and policies
affect the set of HR practices in place at particular fabs. We will
focus on a comparison of Asian and U.S. fab human resource policies,
given that 12 of 15 sample fabs are located in one of these two
regions. In our sample, there is also an overlap between fab location
and technology. Four of the five memory fabs in our sample are in
Asia. The logic fabs are geographically diverse. Accordingly, it
may be difficult to separately identify the effects of location
and technology across our sample fabs. For example, most of the
memory fabs have a complex pay system in place for production workers
(as well as for engineers) that includes base hourly pay and bonus
pay, rather than the straight hourly pay that is typical for other
fabs. This policy is best viewed as generated by particular national
labor market institutions rather than as a decision made by management
at the firm or fab level.
2.2 Distribution of Work
Practices Across Fabs
Figures 2-1, 2-2, and 2-3 report the distribution of work practices
across fabs for each occupation. Clearly, some practices are widely
diffused across the fabs in our sample and others are not. Innovative
training policies are in place at all fabs. Financial support for
off-site training and course work (such as tuition reimbursement
and paid time off) is policy at every fab, and covers operators
as well as technicians and engineers. Operators at nearly all fabs
are cross-trained on more than one equipment type within their assigned
bay. In addition, most of the sample fabs report cross-training
operators on equipment types in more than one equipment bay. Technicians
and engineers commonly participate in on-the-job training of operators,
and engineers contribute to the training of technicians in nearly
all fabs. (Chapter 6 discusses training intensity across the sample
fabs.)
The use of problem-solving teams is also a prevalent practice among
the fabs surveyed. We distinguish between cross-functional teams,
which contain members from multiple work areas or functions, and
quality improvement teams or circles, which are organized by area.
Both are widely adopted across fabs, although not necessarily within
fabs (see Chapter 7). Self-directed work teams are somewhat less
widespread than the other types of teams. However, several fabs
plan to introduce such teams in the future, suggesting that self-directed
work teams will soon become as prevalent as quality improvement
teams are currently. Our case studies also suggest that self-directed
work teams, once established, do not displace other types of teams
operating in the fab. The relationship between types of teams appears
to be one of complimentarity rather than substitutability. Seven
of the fabs in our sample have active teams of all three types and
two more fabs are planning to add self-directed work teams to their
rosters of quality improvement and cross-functional teams.
Teams are a mechanism to solve complex problems which are difficult
for individual operators, technicians, or engineers to respond to
individually, in the context of their day-to-day responsibilities.
Teams facilitate information sharing and learning among employees
and allow groups of employees to focus intently on particular problems
in a structured environment. But it is also possible for learning
and problem solving to be incorporated into the range of activities
employees accomplish on the job without the presence of teams. For
example, the experience of one fab in our sample suggests an alternate
model for accomplishing the problem-solving activity that teams
are intended to foster. This fab set up and subsequently dismantled
two types of teamsCquality circles and self-directed work teamsCafter
a number of years. When we visited this fab in 1993, it had recently
established Total Preventative Maintenance (TPM) activity circles
among its operators. But a high level of problem-solving activity
was incorporated into the routine of employees without relying on
a large team infrastructure. The fab trained certain operators intensively
in statistical process control (SPC) techniques or equipment maintenance
and subsequently expected them to train other operators continuously,
as part of their job descriptions. For example, some operators were
selected to join the Equipment Engineering staff for a three to
twelve month period of training, after which they would perform
equipment maintenance, disassembly, and trouble-shooting. In addition
they became responsible for on-the-job training of other operators
in routine maintenance procedures. Because the operator work force
was sufficiently trained in equipment maintenance, there was no
equipment maintenance technician job category in this fab. In similar
fashion, operator group leaders were trained intensively in SPC
theory and techniques by the Quality Assurance Department. The group
leaders were responsible for providing SPC training to the operators
under their purview. The multi-leveled and near continuous nature
of training in the fab appeared to induce the formation of spontaneous
informal teams. One supervisor reported that when a particular problem
arises, he encourages four or five operators to join together to
determine the source of the problem. Hence, group problem-solving
activity is still important in this fab, but the high skill levels
of operators allow such activity to be carried out in a less structured
manner than is typical of problem-solving teams. In effect, intensive
work force training can make formal team structures at least partially
unnecessary.
2.3 Distribution of Compensation
Policies Across Fabs
Table 2-2 presents the distribution of innovative compensation policies
across the fabs in our sample. We indicate whether a particular
policy is in place for all principal occupations in a fab or some
subset of those occupations. The two most prevalent policies in
place at fabs are profit-sharing plans and employee stock ownership
plans (ESOPs). With one exception, these policies apply to all occupations
in the sample fabs. Also widespread are non-monetary incentives,
suggestion pay and knowledge pay. However, these three policies
are somewhat more likely to be applied selectively to certain occupational
groups. Team and individual incentive pay policies are rarely in
place for all occupations in a given fab. But about half of the
sample fabs have individual incentive pay for one or two occupations.
2.4 Policy and Practice
Comparisons Across Fab Categories: Memory vs. Logic Fabs and Asian
vs. U.S. Fabs
To get some indication of whether the policies in place at sample
fabs differ according to process/product technology or location,
core policies were identified for each of four fab categoriesCmemory,
logic, Asian, and U.S. A policy was considered core if it was widely
diffused across the fabs in a category, i.e. if more than two-thirds
of the fabs had such a policy in place or were planning to implement
it. Table 2-3 presents the core
practices of fabs in each category. All fabs, regardless of category,
have the same broad set of policies in place for operator training,
though this does not imply that actual training levels are similar
across fabs (see Chapter 6). Differences between categories are
nearly identical across the two comparison groups (memory-logic,
Asian-U.S.) for team structure, compensation policy and work organization.
Since most memory fabs in our sample are Asian fabs, and conversely,
most Asian fabs are memory fabs, the one-to-one correspondence of
core policies in the two categories is not surprising. (Only half
of U.S. fabs are logic fabs, and conversely.) Unfortunately, such
correspondence makes it difficult to identify systematic differences
across comparison groups. And inadequate sample sizes at present
do not permit comparisons that might be more informative, such as
comparing U.S. logic (memory) fabs and Asian logic (memory) fabs.
The group comparisons produce a number of interesting results. U.S.
fabs are more likely to have in place a variety of team structures
(cross-functional, self-directed work teams, and quality improvement).
Asian fabs tend to restrict their teams to quality improvement efforts.
Our site visits give no indication that more teams are necessarily
better. In fact, U.S. fabs tend to have a shorter track record with
team organization and accordingly may be trying to learn what types
of teams work best.
Asian fabs use a wider range of compensation policies than U.S.
fabs. Suggestion and knowledge pay and non-monetary incentives are
more widespread in Asian fabs.
Shift rotation for operators is widespread in Asia but not the U.S.
Job rotation for engineers is also the more prevalent in Asian fabs.
In U.S. fabs, job rotation for operators is not a core practice,
unlike in the Asian fabs.
2.5 Policies and Practices
Across Occupational Categories
There is some evidence that policies that were once applied exclusively
to one occupation are now more widely diffused across occupations
in the fab. For example, knowledge pay is nearly as prevalent for
operators as for engineers in our sample of fabs. Team incentives,
although not a widespread practice, are more common for engineers
than for operators or technicians. However, a number of fabs have
now begun to extend this policy to technicians. Shift work schedules,
once a policy applied principally to operators and technicians,
are becoming more common for engineers as well (although not each
shift is necessarily fully staffed). About half of the fabs in our
sample now also put engineers on regular night and weekend assignments.
2.6 Summary
1. Systematic differences in practices exist across fabs when they
are grouped according to their major process flows (memory or logic)
or according to region (Asia or the U.S.). Because most of the memory
fabs in our sample are in Asia (4 of 5 fabs), it is not easy to
distinguish between region-practice and process-practice correlations.
2. U.S. fabs are more likely to use a variety of team structures
(cross-functional, self-directed work teams, and quality improvement).
Asian fabs tend to restrict their teams to quality improvement efforts.
Our site visits give no indication that the number or variety of
teams in place at a fab is positively related to the amount and
quality of problem-solving activity. U.S. fabs tend to have a shorter
track record with team organization and may be trying to learn what
types of teams work best.
3. Asian fabs use a wider range of compensation policies than U.S.
fabs. Suggestion and knowledge pay policies are more prevalent in
Asian fabs.
4. Work organization differs significantly across Asian and U.S.
fabs. Shift rotation for operators is widespread in Asia but not
the U.S. Job rotation for engineers is also the norm in Asian fabs.
Moreover, in U.S. fabs, operator job rotation is not a common practice,
unlike in the Asian fabs.
End of Chapter 2
Go to
Chapter 3
Go to Table of Contents for this Chapter
Go to Table of Contents for the
CSM-HR Interim Report
|