THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN
Second Interim Report
1. Executive Summary
1.6 Employee Experience, Pay, and
There is significant variation in employee experience levels in
our sample of semiconductor fabs. One-third of our sample fabs had
a majority of operators with at least five years of tenure. At several
fabs, however, a majority of operators had less than two years of
tenure. Operator work forces with a female majority are common in
semiconductor fabs and predominate in our sample. The engineering
work forces in our sample had higher levels of fab tenure than operators.
Most fabs reported that a majority of their engineers had at least
five years of tenure.
We find a significant degree of wage-level dispersion across our
sample fabs. This is expected given that the sample fabs span seven
countries and three continents. Most fabs in our sample reported
operator career ladders with four grades and fairly flat wage paths
across grades. Career ladders for technicians exhibit more steepness
than those for operators. Engineer career ladders are longer and
the wage paths more steeply sloped, relative to those for operators
In our sample, we do not observe a positive relation between work
force experience and wage path steepness, as might be expected if
a principal goal of career ladders is to encourage skill development
and retain experienced employees.
Job assignments may or may not translate into job ladders. This
relationship depends on the use of job assignment to achieve skill
acquisition (i.e., learning a new task and multi-skilling) or achieve
skill deepening (i.e., learning how to solve problems or make repairs)
or reward workers with a preferred assignment. Job assignment may
be used by management as a strategy for developing skills, or job
assignment may be used by employees to improve their own working
conditions. Similarly, job rotation may or may not be part of a
system of multi-skilling and flexibility; instead, it may be part
of a system of job control to relieve monotony or to decrease injuries.
In the high performing fabs, we observed that job assignments were
an integral part of the training process and were used to broaden
and deepen skills as well as to reward for good performance.
1.7 Appraisal and Promotion
We asked the fabs to rank the following criteria in order of importance
for promotion: Attendance, Attitude, Output or Work Goals, Quality
Record, Skill Level, Number of Skills, Team Participation, and Tenure.
Given the technical complexity of each step of the semiconductor
process flow and the need to understand these complexities for effective
trouble-shooting, the data show that fabs value skill depth more
highly than skill breadth. Of the eight appraisal criteria, the
fabrication plants in our sample ranked Skill Level at the top for
all job categories--engineers, operators, and technicians. Number
of Skills, our proxy for skill breadth, ranked no higher than fifth.
Of the appraisal criteria, Tenure ranked last. Less than half the
fabs in our sample even consider Tenure in promotion decisions.
However, tenure and skill development are often closely related
for operators and technicians. Promotion based upon skill would
result in similar outcomes to promotion based upon tenure, with
some exceptions at the high and low ends of performance and ability.
Few operators and technicians are ever promoted to other job categories.
Only three fabs reported promoting more than ten operators a year
to group leader, supervisor, or technician. No fab reported promoting
more than four technicians to a supervisor or engineer position.
The limited number of promotions of operators and technicians outside
of their job categories indicates that barriers exist. During site
visits, we found that the primary barrier is additional schooling.
Another reason why few operators are promoted to the technician
job category is cultural. A few fabs in our sample have barriers
along gender lines: women are operators and men are technicians
or engineers. A final reason for the limited mobility of hourly
workers is a monetary reward issue. Some operators and technicians
whom we met during our site visits said that they were reluctant
to relinquish the ability to earn overtime pay for a salaried position.
In the First Interim Report on Human Resources, we found that the
practice of relying on prior experience was not an effective method
of obtaining skilled operators and technicians. This implied that
much of the important learning for operators and technicians occurs
on-the-job in a particular fab environment.
The average number of days of initial training is similar across
job categories, ranging from 21 to 27 days. The number of days of
initial training is positively correlated with performance for all
three job categories.
About 40% of each worker's time is spent in on-the-job training
during the first year of employment across all three job categories,
with an additional 5-8% of their time spent in the classroom. No
significant correlations were found between this training and performance.
After the first year, approximately 15% of each workers time
is spent in on-the-job training, with an additional 6-10% spent
in the classroom. Again, no significant correlations were found
between the amount of training after the first year and the performance
The results show that the type of training and the method is delivered
are important, rather than the time in training. It appears that
using OJT training alone isn't useful with respect to increasing
performance. For operators and engineers, using only classroom training
hinders the goal of high performance. Technicians respond positively
to solely classroom training. There is a positive correlation between
the use of both types of training and the performance metrics across
all job categories.
1.9 Skills and Work Tasks
Rapid and thorough problem-solving is a critical activity in the
semiconductor industry. The initial data collection allowed us to
draw preliminary conclusions about the relationship between HR practices
· Superior performance is associated with high technical content
in operators' job responsibilities. To more thoroughly investigate
this relation, our new questionnaire explores the distribution of
equipment maintenance and SPC responsibilities among operators,
technicians, and engineers.
· Many engineers thought that they should be devoting their
time to long-term, yield improvement projects rather than "fire-fighting"
on the production line. Our data show, however, that engineering
involvement in sustaining the production line correlates with high
fab performance. We are studying to what degree techs and operators
can substitute for engineers when problems occur.
Examination of the participation of operators, technicians and engineers
in equipment maintenance and statistical process control (SPC) activities
indicates that these findings are partially substantiated. Teamwork
is commonplace in the majority of fabs in our sample, and teams
at different fabs share many characteristics. The organization of
work tasks and the participation of workers in specific activities
seem to have a bigger impact on performance than the characteristics
Equipment maintenance and statistical process control (SPC) activities
exhibit a greater degree of variation in human resources practices
across the fifteen fabs than does participation in teams. Both equipment
maintenance and SPC are central to effective problem-solving in
a fab. The correlations for SPC activities and performance support
the hypothesis that fabs that engage in technical tasks most intensively
should exhibit the highest performance. The involvement of operators
in equipment maintenance activities is positively correlated to
performance; however, the involvement of technicians and engineers
is not. So the earlier finding that fabs that involve their operators
in equipment maintenance achieve a higher level of performance is
supported. A fab's need to solve problems quickly and permanently
requires operators and technicians to identify problems immediately
and then work with engineers to uncover root causes and implement
lasting solutions. We will continue to document how this process
is most effectively achieved.
1.10 Summary of Findings and Job Prospects
The semiconductor industry provides a lot of training across all
occupations, from operators to engineers. This training is necessary
because workers are involved in continuous problem-solving in an
industry that is continually introducing new processes or new products
and is continually automating. After an average of a month of initial
training, workers are receiving training about half the work time
during the first year with the bulk of the training on the job.
In subsequent years, about one-quarter of the time is spent in training
Training of technicians is more likely to be correlated with the
performance metrics than training of operators and engineers. We
believe that this reflects the importance of machine up-time in
determining machine productivity and the large variation in actual
machine up-time observed across fabs. Fabs also exhibited large
variations in the involvement of employees in equipment maintenance
activities (Figure 1-1).
Operators are involved in fairly high skilled procedures, including
various types of SPC (Figure 1-2) and
equipment maintenance activities. Compared to operators in traditional
manufacturing jobs, the operators in semiconductors oversee a highly
technical process and undertake relatively complex technical tasks.
Most operators are involved in data collection and monitoring, but
the level of operator involvement declines as the difficulty of
the task increases. The level of operator involvement in problem
solving is usually limited to identifying the nature of the problem
and notifying technicians or engineers. In a few fabs, operators
are involved in performing some routine maintenance. Overall, operators
perform tasks that require training and skill development. However,
operators are still limited in their skill development and career
growth, as well as wage growth, unless they become techs.
Two examples of career ladders for operator/technicians are shown
in Table 1-2. All production workers in the large Japanese semiconductor
companies are on a career ladder that combines operator and technician
tasks, training and skills. By age 40, Japanese electronics workers
have technical skills and job tasks. In the U.S., the operator jobs
are usually separated from the technician jobs , and an operator
does not necessarily (or usually) become a technician. However,
most fabs provide the opportunity for an operator to move up to
a technician job. To do this, the operator typically must return
to school to earn an AA degree in electronics, since fewer than
10% of operators have AA degrees. They also must take some home
study courses as well as undertake specific company-provided training
(including on-the-job training in certain processes). The requirements
are rigorous and require a lot of nonpaid time and commitment. At
the U.S. company shown in the example, which encourages internal
promotion, only one-third of the operators become technicians and
one-half of the technicians were promoted from operator. For those
operators who do become technicians at this U.S. company, their
career ladders looks like the Japanese career ladder, which takes
about 20 plus years and includes wages increasing 3.6 times.
Korean fabs provide an interesting contrast to both the Japanese
and U.S. cases, since operator jobs are strictly segmented from
technician jobs. Women, who live and work at the company for only
three to five years before quitting to get married, are operators;
men, who usually have long careers with the company, are technicians
(as well as engineers and managers).
Hourly earnings in the semiconductor industry have increased as
the importance of technicians has increased (Figure
1-3). However, average earnings ($14.50 in 1994) are still low
compared to UAW production and craft workers, who earned $18.28
and $21.44, respectively, in the Big Three automobile companies
American companies increased employment in the U.S. relative to
employment offshore since 1991, so that U.S. employment exceeded
offshore employment in 1994 (SIA Annual Databook: 1995, Semiconductor
Industry Association). Overall, the outlook for employment, especially
technicians and engineers, is strong even in the presence of continual
automation. Although there is the potential for long career ladders
for non-college graduates, the upgrading requires a technical education
pursued outside of work as well as work-based training. The semiconductor
industry pays relatively lower wages than unionized manufacturing
but above average wages for all manufacturing, which has lower skill
1.11 Focus Studies
Analysis of key questions are presented in the seven focus studies.
In "The Transferability of Case
Study Findings to Other Firms," Rascher and Brown address
the problem of how to know if "best practices" can
be transferred from one fab to another. They present a table
that highlights which practices are necessary or sufficient
for successful transfer.
In "The Evolution of Skill Demand
and the Nature of the Employment Relationship in a Technology-Intensive
Firm," Valvano looks at the dynamics of career and pay
patterns of managers and professionals in one U.S. company over
a 19-year period. He finds that the occupational distribution
of employment shifted towards higher-skill professionals beginning
in the mid-1980s. Real mean earnings over the period 1976-1994
are relatively stable, and there is little evidence of adjustment
of earnings levels to labor market or product market shocks.
In "How Does Knowledge Flow?
Inter-Firm Patterns in the Semiconductor Industry," Appleyard
shows that Japanese engineers rely more upon public sources(e.g.,
conference presentations and journals) for technical information
and U.S. engineers rely more upon private sources (e.g., customers
and personal contacts at other companies). She discusses the
formulation of optimal strategies of interfirm knowledge sharing
given the characteristics of the semiconductor industry.
In "Managing Creativity and
Control in Innovation," Brown analyzes the management of
the development and transfer of a new process or product. She
discusses how work organization, incentive systems, and communication
systems affect the creation, sharing, and control of knowledge.
In "Enhancing the Rate of Learning
by Doing through Human Resource Management," Hatch demonstrates
that investments in human capital, which transform laborers
into problem solvers, increase the level of learning by doing.
· In "A Systems View of Work Group Performance: An
Example from Semiconductor Manufacturing," Bowen constructs
a model of group performance that analyzes the contribution
of the knowledge, skills, and abilities of work group members
to performance and identifies which elements of the work system
inhibit performance improvement.
In "Statistical Tools for Industry
Data," Sattler uses multidimensional scaling and principal
components analysis to analyze the patterns and to present "overall
picture" graphs of the high performers and low performers
in the first wave of case study data. She also compares these
data to a subset of the data from the HR questionnaire.
End of Chapter 1
Go to Chapter 2
Go to Table of Contents for the CSM-HR