Introduction

      Management of human resources and the restructuring of work in the semiconductor industry holds important lessons for the future, since this industry is already experiencing the competitive forces and technological innovation that most industries are expected to face in the 21st century. This report explores what determines long-run competitiveness in the semiconductor industry, which is characterized by rapid technological change and global competition that result in short product cycles, declining prices, and volatile markets. The continual technological change and automation require a flexible and skilled work force. Although labor costs are a small proportion of total cost, the management of human resources potentially has a big impact on performance because of the role of labor in determining the life and productivity of the costly capital equipment. How technological change, automation, and global competition are affecting wages and working conditions, along with educational requirements, are examined in this study.

      This report first looks at the role of HR in performance, which is determined by manufacturing efficiency and quality. In the next chapter, a statistical analysis of data from sixteen fabs is used to identify key variables in effective HR systems and to explore how the component parts of these HR systems fit together. In chapter 3, we deepen the analysis through qualitative case studies to examine how employment systems actually operate in three key areas--skill development, problem solving and innovation. Chapter 4 explores the extent to which our semiconductor results are transferable to other industries. In chapter 5, we examine the wage structures in the semiconductor industry at both the national level and at the fab level to see how wage inequality has grown in the U.S. semiconductor industry since the late 1970s and to compare these changes to national trends. We use detailed personnel data that spans sixteen years at one major company to see to what extent internal compensation policies are mirroring the national trends. This company's pay and promotion system for managers and professionals is further explored in the subsequent focus study. The final focus study examines knowledge sharing and innovation across company lines in the development of equipment. This executive summary highlights our findings.

HR Systems and Firm Performance (Chapter 2)

      To measure human resource practices, we created measures of the significant components of the HR system for our sample of sixteen fabs. These measures focus on skills developed (training breadth and depth and experience), the actual skills used in job tasks, worker involvement in specific activities (both individually and in teams), and the reward system.

      The performance of a particular HR system depends on goals as well as the economic and cultural environment. In general, we found one type of high-performing HR system that achieved both quality and volume (output) goals:

• Skills-driven Employee Involvement (EI) System
  Fabs with this type of HR system have advanced SPC systems, trained operators and technicians performing SPC and equipment maintenance, decentralized teams for problem solving, and experienced engineers with long potential career ladders. These fabs were able to achieve high scores on both volume and quality, often depending upon which goal they stressed.

      However, we also found two high-performing systems that exhibited a trade-off between reaching quality and volume goals:

• Training-intensive and Experienced Engineers System
  Fabs with HR systems that rely on experienced technicians and engineers rather than advanced SPC systems and that use continuous training have higher scores on volume-related metrics (stepper throughput and direct labor productivity) with lower scores on quality metrics (defect density and line yield).
• SPC and Job-related Training System
  Fabs with HR systems that have advanced SPC systems used by a trained work force, trained technicians and operators doing equipment maintenance, and centralized teams led by engineers with high potential wage growth, have higher scores on quality (defect density and line yield) and lower scores on volume (stepper throughput and direct labor productivity).

      In developing a high performance HR system, the firm must take into account social norms and labor market institutions, the product market, and its information systems since the path to high performance depends upon all these factors.

Lessons from the Shopfloor (Chapter 3)

      The impact of environmental factors is explored in four studies, which provide a detailed analysis of our field work to understand how employment systems actually operate. In particular, we analyze the related processes of skill development, problem solving and innovation in selected fabs. In the first two studies, environmental factors such as the number of processes and products being produced and the stability of the organization are vastly different, and these dissimilarities have an impact on the functioning and effectiveness of the fabs’ problem-solving systems. In the third study, external factors are common and include high turnover of production workers, and this allows a more direct study of the effectiveness of the HR system in one environment. In the fourth study, a case study of the development process at two fabs is combined with a multi-fab study of shop floor practices to study the management of innovation.

      "The Role of Skill Upgrading in Manufacturing Performance" and "Problem-Solving Structures" are both case studies comparing a high-performing and a low-performing fab. In both cases, the environment exhibits more variation between fabs than between the HR systems themselves. However, in both studies the low-performing fab would benefit from making changes in the training and problem-solving structures so that they functioned more effectively. The first case study documents at two fabs the skill upgrading systems equipping operators to perform planned equipment maintenance and technicians to perform planned maintenance and repairs. Production workers at both fabs engage in few SPC activities. At the Japanese fab in the case study, process engineers are involved with the advanced SPC system. In contrast, the U.S. fab examined has a less-developed SPC system, which accounts for part of its lower performance even with its employee involvement in equipment maintenance. In addition, the U.S. fab faces a much more challenging environment where the organization is being restructured and where production encompasses a broad product and process mix. The Japanese fab is producing a stable product. The second case study is similar in that a high-performing U.S. fab is producing an old and stable product and a low-performing U.S. fab is in the process of reorganization and produces a large variety of products in a weak product market. Although both fabs have extensive training systems for problem solving, one U.S. fab does a much better job of involving the production workers in solving problems while the other fab relies much more on engineers.

      "Human Resource Policies in an Environment of High Labor Turnover and Rapid Technological Change" compares two semiconductor fabs that are closely matched in terms of their process technologies, product mix, and business strategy and that are located in the same region in a country outside of the U.S. The core human resource policies at these two fabs are remarkably congruent. This suggests that external factors, such as high operator turnover and the product market, have had a major influence on the choice of policy. The need to continually bring new processes into the fab and ramp up volume as quickly as possible contribute to an engineering dominant culture. Both fabs are quite good at new process introduction but they appear less successful in dealing with yield problems associated with volume processing, such as equipment-induced problems and operator errors and misprocessing. Yet both fabs have very good direct labor productivity. Even so, there are important differences in performance between these two fabs that need to be explained. One fab has much higher line yield and lower cycle time than the other as a result of effectively using automation and computer-aided manufacturing (CAM) to compensate for the constraints imposed by the external labor market. This technology tends to mistake-proof certain relatively routine steps in wafer processing, to compensate for a low-skilled and taxed operator work force, which does not receive on-going training. It also speeds processing times and speeds the processing of information, enhancing the problem-solving capabilities of the engineers. The comparison of these two fabs shows that HR systems alone cannot be expected to solve all employee-related problems. The effective use of technology must be part of the firm's strategy in meeting performance goals.

      The chapter "Innovation on the Shopfloor: Successes from the Semiconductor Industry," describes how firms structure knowledge diffusion and problem-solving activities to improve their performance in the semiconductor industry. An analysis of firm-level data from Asia, Europe and the U.S. demonstrates how semiconductor companies have structured their employment systems to solve the technically demanding problems that arise daily on the shop floor.

      In a high-tech environment, the key ingredients of a high performing human resource system include:

• incentives for creativity and systems for control in managing innovation by engineers;
• the organization of work so that line workers develop and use problem-solving skills;
• team problem-solving and the sharing of knowledge across all job categories; and
• the formation of career ladders so that skill development, responsibilities, and compensation are linked over a worker's tenure.

Lessons for Other Industries (Chapter 4)

      In this chapter we compare the semiconductor industry to other important industries to ascertain the extent to which our findings might be applicable to other industries. In particular, we focus on the industries being studied by the Sloan Industry Centers, since their research provides an in-depth analysis that is not otherwise available. These industries range from long-established, traditional manufacturing to those that produce high-technology products to those that provide services. Our inter-industry comparisons will help us to understand in what ways the semiconductor industry is unique and in what ways it shares common problems.

      Industry comparisons indicate that some of the lessons from the semiconductor industry can be useful to other industries. Like most other industries studied, the semiconductor industry faces global competition, especially from manufacturing in low-wage countries, and business cycles. In other ways, such as the pace of technological change with high R&D costs and short product lives, the semiconductor industry is like other high-tech industries in its need to create and control knowledge and to continually implement technological changes. U.S. high-technology manufacturing industries, with the exception of telecommunications, also share common employment structures, and firms pay relatively low wages for the skills and education required of their workers. Traditional U.S. manufacturing, which has typically been unionized, pays much higher wages for often lower-skilled work. In addition, these unionized industries have lower earnings disparities between production and nonproduction workers.

      We explore the usefulness to other industries of lessons from the semiconductor industry in three areas—improving shopfloor performance, managing risk and uncertainly, and regulating intellectual property. Overall, we believe that the semiconductor and other high-tech industries provide lessons to help other industries prepare for the future as the pace of technological change, uncertainties associated with global markets, costs of capital and R&D increase across all industries.

Wage Structures and Inequality (Chapter 5)

      This chapter examines wage structures and inequality in the semiconductor industry on three levels ranging from national data to detailed case study data. In the economy as a whole, wage inequality has increased rapidly since 1979: The combination of an increasing education premium, increasing experience premium, and increasing within-group dispersion results in a large cumulative increase in national wage inequality.

      Although the trend in wage inequality in the semiconductor industry is similar to the national trend, the underlying changes in inequality exhibit some differences. The growing inequality in the semiconductor industry reflects a rapidly growing managerial and professional premium and slowly increasing college premium. The semiconductor industry, however, does not exhibit growth in the experience premium or in within-group (i.e., among workers with similar education, experience, and occupation) dispersion, both of which are occurring nationally. Thus, the majority of the changes in relative wage levels are driven by the occupation premium, which is closely linked to the college premium. We believe that automation, the rapid pace of technological change, and the outsourcing of manufacturing work are responsible for the increase in the earnings of college-educated engineers and managers and a decrease in real wages for production workers.

      The cross-sectional data from the U.S. semiconductor fabs in the CSM sample describe an industry that uses a variety of compensations practices. Some fabs present their workers with long career paths and potential earnings growth while other fabs do not. In spite of variations in pay structures, semiconductor fabs overall have substantial occupational pay differences that reflect both the return to education and the return to the managerial/professional occupation. Analysis of personnel data from one firm ("NewTech") over the period 1979 through 1994 supports these results. There has been a dramatic shift in the occupational structure, driven by a major decline in the employment of production workers and a continuous increase in professional employment. By several measures, wage dispersion at NewTech did not increase during the 1980s, a time when inequality was increasing significantly in the U.S. labor market. However, by the mid-1980s, both education and occupational differentials at NewTech began to widen considerably and within-group wage dispersion increased in the 1990s. Operatives' wages lost ground to those of all other major occupational groups, including technicians. The college-high school wage differential widened considerably. Both of these trends mirror those occurring in the aggregate labor market, although they appear to have started somewhat later at NewTech. In contrast to a rising experience premium in the U.S. labor market, the experience premium for professionals at NewTech has been declining. This may reflect the importance of cutting edge skills in industries in which technological innovation is rapid.

      Using both industry-wide data and firm-specific data, we show that the increase in the return to education, the relative increase in managerial/professional pay, and increased earnings dispersion within occupations are driving increased inequality in the semiconductor industry.

Focus Studies (Chapter 6)

Pay Policies for Managers and Professionals
      In his chapter "Pay and Promotion Systems for Managers and Professionals at NewTech", Vince Valvano uses data from "NewTech" to analyze the relationship between pay and two prominent features of this firm's internal organization—job grades and job classifications (titles). There is significant movement of employees between job grades and between job classifications in this firm. The former are more strongly associated with increases in pay. However, he argues it is misleading to identify all moves between job grades with promotions or to equate the job grade structure with the firm's responsibility hierarchy. The salary ranges associated with job grades are administratively determined, so that some employees will change job grades simply because they have reached the pay ceiling associated with their current grade. He also examines transitions between professional and manager job titles. Traditionally a transition to manager has been thought to be important for continued pay growth for engineers. Valvano finds, however, that moves from manager to professional job titles occur in this firm, suggesting that the career ladder for engineers is more flexible than is typically assumed in the literature.

Knowledge Sharing
      Melissa Appleyard in her chapter "The Role of Knowledge Spillovers in Buyer-Supplier Co-Development" presents a model of cooperative technology development undertaken by a buyer and a supplier of capital equipment. In order to meet the demanding technical requirements of their production processes, semiconductor companies often spearhead technology development projects with their equipment suppliers to advance the processing capabilities of their capital stock. This chapter focuses on the tradeoffs a buyer faces when initiating an equipment improvement project. On one hand, the buyer wants the piece of equipment, or "tool," to be tailored to its particular processing needs. On the other hand, the buyer wants the modification to be incorporated into multiple generations of the tool, and equipment suppliers will more readily incorporate hardware changes that have general applicability across its customer base. Buyers will pursue equipment modifications that have some applicability to other chip producers' processing needs and may have to offer inducements to the supplier to ensure participation. Knowledge spillovers across buyers occur not because one buyer uses another's modification without compensation but because they are embodied in capital equipment purchased from a common supplier. This paper finds that the growth of and access to the supplier's stock of knowledge discourages vertical integration in the semiconductor industry.

Conclusion

      Three major conclusions emerge from the analyses summarized above. First, effective human resource systems in the semiconductor industry incorporate the development and use of skills and knowledge into all jobs so that solving problems and implementing technological change is an integral part of the production process. Engineers play a critical role in solving problems and implementing new processes in this high-tech industry. However, we witnessed a wide range of problem-solving capabilities among the operator work forces of the fabs in our sample. In most fabs, the types of problems that operators can be expected to solve are fairly simple, although operators can play a critical role in monitoring the production process, collecting data, and performing routine machine maintenance. A trade-off between quantity (efficiency) and quality goals in manufacturing sometimes exists, and also technology and employee involvement may be substitutable in providing problem solving support to engineers. We found one type of high-performing HR system that could achieve both quality and volume goals. This skills-driven employee involvement system combined both advanced information technology and automation with highly trained production workers performing SPC and equipment maintenance activities. Some fabs with high performance in volume-related metrics substitute engineering time and talents for advanced technology (training-intensive and experienced engineer system), but this option may become less viable as fabs become more automated to handle larger wafers and as the information technology system is upgraded to handle a larger array of products and processes. Some fabs exhibiting high performance in quality metrics have advanced SPC systems and trained production workers engaged in SPC and equipment maintenance activities (SPC and job-related training system). However, these fabs have not fully engaged their production workers in the problem-solving process.

      The second conclusion of the study is that the future looks bleak for those with less than a two-year college degree. The semiconductor industry in the United States has been experiencing a shift in its occupational structure as production jobs have declined while engineering and other professional jobs continue to grow. On top of declining job opportunities, semiconductor operatives' wages lost ground to those of all other major occupational groups, including technicians. The semiconductor industry has also experienced a rapidly growing managerial and professional premium and slowly increasing college premium. However, the industry has not exhibited growth in the premium for experience, which had started at a high level. At one major company, experience premiums were high across all occupations although declining for professionals. With the occupation premium, which is closely linked to the college premium, driving the changes in relative wage levels, the external labor market rather than technological change is exerting downward pressure on the wages of production workers. Particularly in the U.S., production workers in all industries face declining employment opportunities and weakening collective bargaining power. Even though semiconductor operatives are relatively high-skilled and face pressures to avoid costly mistakes while working in uncomfortable work environments, they earn about 75% of the wages of operators in the unionized automobile sector.

      The third major conclusion of the study is that human resources systems alone cannot be used to solve all employment problems. Often HR systems are expected to solve problems that should be solved by technology or product-market strategies, such as the use of automation, redesign of the product or process, or introduction of a more popular product. For example, one of the fabs in our sample effectively responded to a high level of operator turnover by introducing advanced information and computer aided manufacturing systems to reduce misprocessing errors by operators. Here the use of technology is a superior solution to implementing onerous work rules to reduce mistakes. In another example, a fab producing below capacity because of lack of demand will not have good measures of performance regardless of the HR system. In general, HR systems cannot take care of problems that are outside the control of the fab manager, including cycles in the product market, changes of ownership, and technological breakthroughs. However, high performing and flexible HR systems can facilitate the response to external shocks and technological change.

      We believe that the semiconductor industry provides a glimpse of the future while it simultaneously produces an input, the semiconductor chip, that is shaping the direction of a number of industries. Other industries can look to the semiconductor industry for guidance on how to prepare for a future that holds an increased pace of technological change combined with risks and uncertainties associated with global markets and with high capital and R&D costs.