THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN RESOURCES PROJECT:

Second Interim Report
CSM-32
Clair Brown, Editor

8. The Transferability of Case Study Findings to Other Firms
Dan Rascher and Clair Brown

8.1 Introduction
8.2 Transferability Issues - Importance
8.3 Transferability Issues - Processes
8.4 Key Variables - What Do We Need to Know?
8.5 Transferability in the Semiconductor Industry
8.6 Conclusion
RF References

8.1 Introduction

Case studies on high performance work systems attempt to illuminate the ways in which the practices operate to improve performance. These studies are particularly useful for identifying what appears to be "best practice" in the industry. Because of their limited number, however, case studies cannot be used to statistically test the relationship between outcomes and practices, as can be done with less detailed data on a large number of plants. We believe that this limits the acceptance and usefulness of case study data. One of the major goals of the human resource study of the competitive semiconductor manufacturing program is to understand what are the "best practices" in the semiconductor industry. This paper explains how to expand the importance of case study results by determining if and how it is possible to transfer these practices to other fabs in order to achieve high performance. A fab manager wants to know that if they adopt a certain set of "best practices" in their fab, will they become a high performer?

There are three domains to which we would like to establish transferability of our CSM-HR study:

• Can a fab manager duplicate the "best practices" or processes in other fabs within the same company?
• Can the conclusions from the CSM-HR study be transferred with similar outcomes to other companies within the semiconductor industry?
• Are the results applicable to firms in other industries?

Although the macro-micro linkages are important in establishing what effect the macro-environment has on the transferability across national borders, these linkages are not analyzed here. Instead, we focus on intra-industry transfer issues and briefly consider inter-industry transferability.

Why don't most case studies consider the transferability issue explicitly? Probably because no conventional methodology exists. That begs the question why not? Pertaining to the laboratory versus field argument, Locke mentions that "knowledge isn't sufficient to deduce the answers...thus the real problem of generalization is more one of induction, not deduction." This may also be true for establishing transferability from a case study to other domains. Instead of hard, fast rules that are determined from basic principles, the development of the necessary methods might be on a case by case basis. Here we are searching for some common procedures or steps that can be created and used as a guide.

The rest of this paper is organized as follows. The next section discusses the importance of establishing the transferability of case studies. Section three examines the processes involved with the transferability issue. A general discussion of key work practices is explored in section four. Section five uses the CSM-HR study to begin testing the transferability of the "best practices," followed by section six, the conclusion.

8.2 Transferability Issues - Importance

In our study of competitive semiconductor manufacturing, our goal is to examine whether the results from the study, the determination of the "best practices," are applicable to other plants within the same company, other companies, or even other industries, and how this is affected by the macro environment. How can we test this inquiry, and if the results aren't readily transferable, can adjustments be made which will allow them to be? Since there doesn't appear to be a formal process for generalizing case studies to other domains, a methodology needs to be developed.

Edwin Locke (1986) has systematically examined generalizing lab studies to the field, but not necessarily from one field to the next. This is a closely related issue, but crosses a different bridge. The lab versus field issue addresses the grades of realism in the lab versus the field. We are trying to see whether lessons learned from a particular field study, such as the CSM project, can be used as general benchmarks for other fields or fabs, as the case is here.

Suppose we find that systematic employee skill development over a period of years is part of a successful high performance work system in our sample of semiconductor fabs. Does establishing transferability mean that the empirical value of skill development must be the same for other fabs, or just the direction of the effects? Because of the complex nature of high performance systems, it is likely that most practitioners would be satisfied with establishing external validity on the level of directional effects within the same magnitude as their own case studies.

A procedure for establishing transferability of case studies would have to distinguish between testing whether a case study was transferable and making a case study transferable. If the conclusions from the CSM study are not generalizable to other fabs in the industry, is there anything that can be done? If possible, an adjustment mechanism should be developed that would allow at least some of the conclusions to be transferable to other fabs.

8.3 Transferability Issues - Processes

To establish transferability, a researcher must identify the key variables and hypothesized relationships from case study data in a particular industry. Since "best practices" may include more than one path, and may include HR systems that are not thought of as "high commitment," these variables and relationships may be more complicated than originally thought. There are likely to be several "best practices" depending on the environmental constraints such as the culture and institutions of a country. Theoretically, the number of "best practices" should be limited by the number of goals that the HR system must successfully reach.

With that in mind, where should one start in testing whether a particular case study is transferable? For example, the CSM-HR study examines 15 fabs in an industry that has hundreds. Suppose X1, X2, and X3 designate the observed systematic combinations of HR practices (denoted HR bundles or systems), and we know where these various HR bundles fall on the measure of performance outcomes. We would like to determine to what degree the environment in the population of fabs is replicated so that we can conclude that X1, X2, and X3 will result in the same performance outcomes. To do this we must classify the observed HR practices as necessary or sufficient (or not necessary).

Statistically, a sufficient condition to establish that the conclusions of a case study are generalizable is that the sample is representative of the population. However, if the case studies were not selected to be representative, can we identify, ex post, selective important dimensions in which the sample must be representative? What is needed is the identification of the essential features, or key variables. In the end, one would like the essential features of the sample fabs to be representative of the population.

There are at least two ways to view this. Suppose Fab 1 in our study has work practices bundle X1 in place and has high performance metrics. One test of external validity would show that most fabs in the population that also have these work practices in place are high performers. The percentage of these would have to be statistically significant for transferability not to be rejected. Another perspective would be that of the manager of Fab 2, which is not in the sample. If these practices are imported into Fab 2, will it become a high performer? The answer depends on whether Fab 2 contains the other essential features related to the environment or culture, which will be discussed below.

There are really two sets of important aspects. One is the relationship between the performance metrics and the HR bundles , which are trying to be transferred. Two is the set of conditions required for successful transfer. X1, X2, and X3 cause hypothesized performance outcomes under certain conditions. The transferability of the results of the CSM-HR study will be satisfied conditional on whether a specific environmental climate exists.

The particular problem with transferability within an industry from detailed case studies is to ensure that the hypothesized relationship between performance and practices has correctly identified all the key variables in the relationship. For example, if a key variable is incorrectly left out, then transfer of the system as identified will not necessarily lead to the anticipated performance outcome. If a key variable is misspecified (i.e., the specified variable is acting as a proxy for the true variable), and if the value of the variables diverge or if the relationship between them shifts, unanticipated results can occur.

For example, assume that HR bundle X1, which is associated with high performance metrics, includes industry specific work practices (SP1), e.g., operators performing certain SPC tasks. Suppose X1 also includes general HR practices (GP1), such as employment security and structured on-the-job training.

There are initially two steps to establish transferability:

• Step 1: determine whether the companies that use X1 are high performers, and
• Step 2: to what extent the remaining firms are not high performers.

A test of whether the results of the case study are transferable within the industry would involve showing that X1 is transferable. One test of generalizability is a statistical calculation of whether other semiconductor fabs that employ X1 practices are high performers. A further necessary condition of this test is that the environmental conditions are also satisfied. Only fabs that satisfy both of these conditions are eligible to be part of the test for transferability. If the success rate is statistically significant, then it might be concluded that the case study is applicable to the rest of the industry. This is similar to out-of-sample testing for ensuring model stability.

Step 2 involves checking whether the case study is transferable to firms in other industries (inter-industry) that aren't using comparable work practices or general practices GP1. To establish transferability, it might be sufficient to show that the firms satisfy the environmental conditions. Thus, in all probability, any firm that satisfies the environmental conditions could adopt practices comparable to X1 and become a high performer. Specific practices would most likely need to be replaced by practices specific to the industry that fulfill similar goals, but general practices, GP1, might be universally applicable.

What are the environmental conditions in a particular domain that are necessary for transferability? These are conditions that might either constrain or support a particular practice. They are aspects of the economy that are uncontrollable, at least in the short run, such as the labor supply, product market demand, financial institutions, cultural practices, and the availability of certain technologies. For example, the labor market for the industry might not have enough qualified technicians. Thus, either the HR system cannot be adopted without modifications, or upon adoption part of the system will not perform as expected--in this case the less qualified technicians will prevent the fab from reaching expected performance.

8.4 Key Variables - What Do We Need to Know?

HR or employment systems involve individual level characteristics such as motivation and participation; group level characteristics such as work group climate and work organization; and organization level characteristics such as organizational performance and strategy. The potential "best practices" HR system must have key variables that fulfill specific goals. Here we identify the major individual and group goals: incentives and monitoring, skill building and knowledge creation, and work organization. The HR practices that are used to fulfill these goals are given in Table 8-1.

Go to Next Section of this Chapter
Go to Table of Contents for this Chapter
Go to Table of Contents for the CSM-HR Interim Report