THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN RESOURCES PROJECT:

Second Interim Report
CSM-32
Clair Brown, Editor

12. Enhancing the Rate of Learning by Doing Through Human Resource Management
Nile W. Hatch

12.1 Introduction
12.2 Learning by Doing in Semicondictor Manufacturing
12.3 A Model of Learning by Doing
12.4 Empirical Analysis
12.5 Conclusions
REF Reference

Abstract

Learning by doing occurs in the semiconductor industry through manufacturing yield improvements that result in cost reductions. While all semiconductor firms learn through engineering analysis of production volume to improve yields, firms differ substantially in the level and rate of improvement of yields. In this paper, I find that investments in human capital transform laborers into problem solvers. increasing the level of learning by doing activities. Turnover reduces the level of human capital in the factory, resulting in lower yields and a slower rate of learning by doing.


12.1 Introduction

Since the pioneering work by Wright (1936), there has been a steady stream of research on learning-- by doing. Wright found that the direct labor cost of manufacturing an airframe fell by 20% with every doubling of cumulative output. Many studies followed to corroborate Wright's findings in a variety of industries. Subsequently, the scope of analysis was broadened as costs other than direct labor were also shown to decrease with experience and researcher began to study cumulative investment and time as alternative determinants of learning by doing (Arrow, 1962; Rapping, 1965; Sheshinski, 1967; Stobaugh and Townsend, 1975; Lieberman, 1984).


Recent research on learning by doing has focused on what the associated cost reductions imply for the market structure and the price/output path in an industry characterized by learning by doing. An assortment of empirical and theoretical research has considered the role of learning by doing in market structure and pricing (Arrow, 1962; Spence, 1981- Fudenberg and Tirole, 1983; Lieberman, 1984; Ghemawat and Spence, 1985). More recently, a number of studies have explored the relationship between learning by doing and collusion (Mookherjee and Ray, 1991), predatory pricing (Cabral and Riordan, 1994), international dumping (Dick, 1991), and infant industry protection (Head, 1992; Miravete, 1994).


In 1972, the Boston Consulting Group (BCG) advised its clients to manage strategically based on the assumptions of the learning curve. The popular implementation of the strategy was to obtain the highest market share in an attempt to obtain the greatest cost advantage through learning by doing. However, when the promised profitability didn't materialize, the learning curve fell out of favor as a management tool. The apparent failure of "strategic" management based on the learning curve led to criticism of prevailing models and to the identification of gaps in our understanding of how learning by doing works, what forces act upon it and how it influences firm behavior (Dutton and Thomas, 1984; Montgomery and Day, 1984; Alberts, 1989; Mookherjee and Ray, 1991). Some research has been conducted ,to supplement our understanding of the relationship between learning by doing and competitive advantage. 3 However, for the most part, this research has left the underlying determinants of how learning by doing occurs unexplored.


Several authors have described an array of factors that influence a firm's learning performance, but these studies rarely include the factors in a model.' The biggest obstacle to statistical analysis of the determinants of the learning curve is the proprietary nature of the required cost and manufacturing operations data. As a result little progress has been made toward identifying how learning by doing works as opposed to documenting that it exists ' One important exception is the recent paper by Adler and Clark (1991) who use detailed case study data to identify the role of indirect labor (engineering changes and workforce training) in learning by doing in several divisions of a high technology firm.

Hatch and Reichelstein (1994) show how semiconductors costs fall through yield improvements and identifies cumulative volume an cumulative engineering as joint determinants of learning by doing. Rather than serve as a proxy for manufacturing experience, cumulative volume represents the source of information about yield losses, which are eliminated through engineering analysis of production volume.

This paper, extends the Hatch and Reichelstein (1994) research to identify some of the factors that differentiate the rate of learning between factories. Even factories of similar size and age differ substantially in the levels and rates of improvement in yields. This paper identifies, models, and estimates the impact of equipment operator participation in problem solving teams and operator turnover on the rate of learning by doing using data collected in the Berkeley CSM study.

Human capital is important because ultimately it is people who must learn. Firms that invest in the human capital of the direct labor workforce are able to push responsibility for performance down to the lowest levels, incorporating more information into the decision making and problem solving activities that drive yield improvements. Empirically, the involvement of equipment operators in improvement teams improves yields. In contrast, operator turnover is extremely disruptive in yield improvement activities. This is because turnover represents human capital that is lost when the employee leaves the firm. Also, the new operators who replace them are more prone to accidents, causing additional yield problems rather than helping, to solve them.

The relationship between yield and learning by doing is developed, including an introduction of the main determinants of learning by doing, in Section 2. Section 3 develops a model of learning, by doing based on yield improvements followed by a description of the data and empirical results in section 4. Conclusions are given in section 5.

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