PLACE N RETURN BOX to roman thin chockom how your record.
TO AVOID FINES Mum on or baton date duo.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Afflnnutlvo ActioNEqual Opponunny Initiation
mm:

LINKING WORKFORCE SKILL LEVEL DECISIONS TO
THE ADOPTION OF ADVANCED MANUFACTURING
TECHNOLOGIES: A STRATEGIC CHOICE FRAMEWORK
By

Mark Page"

A DISSERTATION

Submitted tO
Michigan State University
In partial fulfillment Of the requirements
for the degree Of

DOCTOR OF PHILOSOPHY

Department Of Marketing and Supply Chain Management

1997

ABSTRACT
LINKING WORKFORCE SKILL LEVEL DECISIONS TO
THE ADOPTION OF ADVANCED MANUFACTURING
TECHNOLOGIES: A STRATEGIC CHOICE FRAMEWORK
BY

Mark Pagell

This dissertation sought to answer the following question: Is there a
relationship between the appropriateness Of the decisions management makes
regarding the skill level Of operational employees in Advanced Manufacturing
Technology (AMT) installations and adoption success ? Field studies were
conducted to collect the data from both manufacturing and human resource
managers. The data was analyzed using both statistical and visual approaches.

The key result Of the analysis is that there is no evidence that highly
skilled employees are a prerequisite for AMT adoption success. The data set
contains a number Of AMT installations which are both highly successful and
using very low skilled employees. In addition the analysis suggests that fitting
Skills to; the manufacturing environment, the level of managerial discretion over
the workforce, and the uncertainty Of the external environment, may increase the

level Of adoption success.

Copyright by
Mark Pagell
1997

ACKNOWLEDGMENTS

Like all dissertations this one was not an individual effort. I could not have
completed the work without the help, input, encouragement and occasional push
from any number Of people. First and foremost the suggestions and input Of my
committee; Alison Barber, Daniel Krause and Gary Ragatz, were invaluable.
Vlfithout them there would be nO dissertation.

The managers who participated in the study also need tO be thanked. At
each and every plant I visited people tOOk up a large slice of their valuable time,
to answer my questions. I finish the process with a much deeper understanding
and appreciation of the work these people do. I can only hope that the results of
this work provide them an adequate return on their time.

Additionally I would like to thank the rest of the members Of the faculty at
Michigan State University. Although they did not participate directly in the
dissertation they did help to prepare me for it. I would especially like to
recognize Steven Melnyk who taught me the value Of asking (and trying to
answer) interesting questions.

, My chairman, Robert Handfield, with whom I managed tO remain friends

throughout the process, deserves special thanks. He tOOk up an inordinate

amount Of his time and energy making sure that I not only did interesting work,
but that I did it well. He pushed me to do a much better dissertation than I ever
would have on my own; and for that I am grateful.

Finally I thank Megan, for putting up with the piles Of paper, the late nights
at the computer, and the five years Of living in poverty. It is to her that this work is

dedicated, as a small payment towards her commitment tO us.

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................... xii
LIST OF FIGURES ......................................................................................... xv
CHAPTER 1
OVERVIEW OF THE RESEARCH ................................................................. 1
1.1 Introduction ..................................................................................... 1
1.2 Research question .......................................................................... 4
1.2.1 Specific research hypothesis .............................................. 5
1.3 Background ..................................................................................... 6
1.3.1 The strategic choice framework .......................................... 6
1.3.2 Level of analysis ................................................................. 8
1.3.2.2 Types Of installations ........................................... 9
1.3.3.1 Studying skill levels and adoption success .......... 10
1.3.3.2 Skill definition ....................................................... 12
1.3.3.3 Operators ............................................................. 13
1.3.4 Appropriateness Of choice ..................................................... 14
1.3.5 AMT ............................................................... ' ....................... 15
1.4 Scope of Dissertation ...................................................................... 15
1.5 Results19
1.6 Structure Of Dissertation ................................................................. 20
CHAPTER 2
LITERATURE REVIEW .................................................................................. 22
2.1 Introduction ..................................................................................... 22
2.2 General criteria for successful adoption .......................................... 24
2.2.1 Technical issues that impact adoption success .................. 25
2.2.1.1 Software ................................................................ 25
2.2.1.2 Tool rules .............................................................. 25
2.2.1.3 Supplier quality ...................................................... 26
2.2.1.4 Information systems .............................................. 26
2.2.1.5 Maintenance .......................................................... 27
2.2.2 Managerial issues that impact adoption success ................ 27
2.2.2.1 Strategic planning ................................................. 28
2.2.2.2 Integration across functions .................................. 29
2.2.2.3 Top management support ..................................... 29
2.2.2.4 Process/technology champion .............................. 30

vi

2.2.2.5 Performance measurement for the system and

its Operators .......................................................... 30

2.2.2.6 Human resource issues ......................................... 32

2.2.2.7 AMT success: conclusions .................................... 33

2.3 The paradox -prevailing wisdom verses the evidence .................... 35
2.3.1 The prevailing wisdom .................. _. ..................................... 35
2.3.2 The prevailing evidence ...................................................... 39
2.3.2.1 Deskilling evidence ................................................ 39

2.3.2.2 Unanswered questions .......................................... 43

2.3.3 Concluding statements on the paradox .............................. 47

2.4 Strategic choice .............................................................................. 48

2.4.1 The strategic choice framework - its appropriateness for
studying the impact Of technology on the workforce and

the resulting skill levels. ...................................................... 49

2.4.2 Mixed change studies: authors who have concluded
choice but not explicitly used this framework ...................... 52
2.4.3 Explicit tests Of the strategic choice framework .................. 56
2.4.4 Strategic choice conclusions .............................................. 60
2.5 Factors that explain strategic choices ............................................. 60
2.5.1 External factors that influence choice ................................. 62
2.5.2 Internal factors that influence Choice .................................. 62
2.5.2.1 General business factors ....................................... 63
2.5.2.2 Product/process factors ......................................... 64
2.5.2.3 Industrial relations factors ..................................... 65
2.6 Choosing a strategy: two generic strategies ................................... 66
2.6.1 Commitmentstrategies..................................‘ ..................... 67
2.6.2 Control strategies or costs reduction strategies .................. 68
2.7 Summary Of literature review .......................................................... 69

CHAPTER 3

PRE-RESEARCH AND RESEARCH DESIGN ............................................... 71
3.1 Pre-research ................................................................................... 71
3.1.1 The case studies ................................................................. 72
3.1.1.1 Boxes .................................................................... 73
3.1.1.2 Cars ....................................................................... 75
3.1.1.3 Composites ........................................................... 77
3.1.1.4 Case study conclusions ......................................... 80
3.2 Proposed model of choice of worker Skills and AMT success ....... 84
3.2.1 Justification Of factors ......................................................... 86
3.2.2 Definition of Factors ............................................................ 90
3.2.2.1 Change I complexity .............................................. 90
3.2.2.1.1 Product! process change ..................... 92
3.2.2.1.2 Environmental complexity .................... 93
3.2.2.3 Managerial discretion ........................................... 93
3.3 Hypotheses .................................................................................... 97

vii

3.4 Detail Of measures to be used ........................................................ 101

3.4.1 Skill level ............................................................................. 101
3.4.2 Change Icomplexity ............................................................ 108
3.4.2.1 The internal environment: product/process
change .................................................................. 108
3.4.2.2 External environment: environmental
complexity ............................................................. 1 12
3.4.3 Managerial discretion .......................................................... 114
3.4.4 Appropriate skill level .......................................................... 116
3.4.5 Successful adoption ............................................................ 120
3.5 Methodology ...................................................... ' ............................. 1 24
3.5.1 Data collection .................................................................... 124
3.5.2 Sample selection ................................................................ 126
3.5.3 Limitations Of research ........................................................ 128
3.5.4.1 Statistical analysis ................................................. 130
3.5.4.2 Data analysis: visual displays ................................ 133
3.6 Validity issues ................................................................................. 134
3.7 Reliability issues ............................................................................. 137
CHAPTER 4
ANALYSIS ...................................................................................................... 138
4.1 Overview and chapter contents ...................................................... 138
4.2 Preliminary data analysis ................................................................ 139
4.2.1.1 Construct validation sample ..................................... 141
4.2.1.2 Data preparation - outliers and missing values ....... 143
4.2.1.3 Analysis - Correlations .................... 144
4.2.1.3.1 Product/process change .......................... 145
4.2.1.3.2 Product/process change and
environmental uncertainty ...................... 146
4.2.1.4 Discussion Of results Of construct validation study... 147
4.2.1.5 Possible sources of measurement error in
construct validation data .......................................... 149
4.2.2 Construct validation study and remaining analysis ................ 150
4.3 Primary sample: preliminary analysis ............................................. 151
4.3.1 Sample selection ................................................................... 152
4.3.2 Potential biases in the primary sample .................................. 153
4.3.3 Comparison Of Primary sample tO Construct
Validation Sample ................................................................ 156
4.3.3.1 Product/process change variables ........................... 156
4.3.3.1.1 Batch size ................................................ 157
4.3.3.1.2 Churn ....................................................... 159
4.3.3.1.3 Families .................................................... 160
4.3.3.1.4 Master production schedule deviation ..... 162
4.3.3.1.5 Product/process change ......................... 163
4.3.3.1.5 Environmental uncertainty ....................... 163

viii

4.4

4.5

4.3.4 Discussion of similarities and differences between the

primary sample and the construct validation sample ............. 164
4.3.5 Detection of outliers in the primary sample ............................ 168
Analysis Of primary data set ............................................................ 170
4.4.1 Correlation matrix primary sample ......................................... 170
4.4.2 Testing H1A: the relationship between skills and adoption
success ................................................................................. 173
4.4.2.1 Correlation Of skill level and adoption success ........ 174
4.4.2.2 Regression Of skill level on success using a
polynomial term ........................................................ 174
4.4.2.3 Regression Of skill level on success using a
polynomial term ........................................................ 176
4.4.2.4 Summary Of tests Of H1A .......................................... 176
4.4.3 Testing H1B: the relationship between appropriate choice
and adoption success ............................................................ 177
4.4.3.1 Operationalization Of fit ............................................ 177
4.4.3.2 Correlations between appropriate skill level and
success .................................................................... 180
4.4.3.3 Categorization Of success ......................................... 183
4.4.4 Drivers Of Skill choices ........................................................... 186
4.4.4.1 Moderated regression .............................................. 186
4.4.4.2 Visual representations Of the drivers Of
skill choices .............................................................. 189
4.4.5 Conclusions Of primary data analysis .................................... 194
Additional Analysis .......................................................................... 195
4.5.1 Additional quantitative analysis ....................... i .................... 195
4.5.1.1 Using the individual product/process change
items rather than the index .................................... 195
4.5.1.2 Self assessment of success .................................. 197

4.5.1.3 Additional measures Of the external environment. 199
4.5.1.4 Installation age, skill level and adoption success .. 203

4.5.1.5 Size Of plant .......................................................... 205
4.5.2 Comparisons Of CNC and FMS .......................................... 206
4.5.2.1 Comparisons Of CNC and FMS - key variables
and indices ............................................................ 207
4.5.2.1.1 Comparing product/process variables
for CNC and FMS .................................. 207
4.5.2.1.2 Comparing external environments for
CNC and FMS ....................................... 210
4.5.2.1.3 Comparison Of total preparation time
for CNC and FMS .................................. 211
4.5.2.1.4 Comparisons Of levels Of success for
FMS and CNC users ............................. 211
4.5.2.1.5 Comparison of the size Of FMS and
CNC installations ................................... 212

4.5.2.1.5 H1A: Relationship between Skills and
adoption success for CNC and FMS ..... 213

4.5.2.1.6 H1 B - The relationship between
appropriate choice and adoption

success for FMS and CNC .................... 214
4.5.2.1.7 H2 - H4: Drivers .of choice ..................... 218
4.5.2.2 Conclusions Of comparisons Of FMS and CNC ..... 224
4.5.3 Industry Effects ................................................................... 225
4.5.3.1 Machine TOOls ....................................................... 227
4.5.3.2 Stamping Dies ....................................................... 229
4.5.3.3 Auto Parts ............................................................ 232
4.5.3.4 Diesel Power ......................................................... 234
4.5.3.5 Injection Molds ...................................................... 237
4.5.3.6 Office Furniture ..................................................... 240
4.5.3.6 Industry conclusions .............................................. 241
CHAPTER 5
DISCUSSION OF RESULTS .......................................................................... 244
5.1 Overview and chapter contents ...................................................... 244
5.2 Level Of support for proposed hypotheses ...................................... 244
5.3 Skill Level Of the Workforce and Adoption success ........................ 246
5.4 Skill choices and appropriate choices ............................................. 251
5.4.1 Drivers Of skill choices ........................................................ 251
5.4.2 Alternative operationalizations Of appropriate skill level ..... 252
5.4.3 The drivers Of skills and the appropriate Skill level .............. 253
5.5 Industry effects ............................................................................... 255
5.6 Differences between FMS and CNC ............................................... 257
5.7 Summary of key conclusions: ......................................................... 259
5.8 Limitations Of the research .............................................................. 259
5.9 Future research and conclusions .................................................... 263
5.9.1 Future research ..................................................................... 263
5.9.2 Conclusions ........................................................................... 267
APPENDIX A
FACTORS THAT IMPACT THE SUCCESS OF AMT ADOPTION ................. 269
APPENDIX B
VARIABLES THAT IMPACT THE STRATEGIC CHOICE OF SKILL
LEVELS ......................................................................................................... 272
APPENDIX C
INTERVIEW PROTOCOL ............................................................................... 275
APPENDIX D
CONSTRUCT VALIDATION INSTRUMENT ................................................... 285

APPENDIX E

FACTOR ANALYSIS ....................................................................................... 287
APPENDIX F

CORRELATIONS CONSTRUCT VALIDATION STUDY ................................. 289
APPENDIX G

STATISTICAL CHARACTERISTICS OF CONSTRUCT VALIDATION
VARIABLES .................................................................................................... 291
APPENDIX H

STATISTICAL CHARACTERISTICS OF PRIMARY SAMPLE VARIABLES... 292
APPENDIX I

CORRELATIONS FOR PRIMARY SAMPLE DATA ........................................ 293
APPENDIX J

ASSESSING FIT FOR THE 30 RESPONDENTS ........................................... 295
APPENDIX K

SORTED FIT MEASURES ............................................................................. 296
APPENDIX L

DESCRIPTIVE STATISTICS BY TECHNOLOGY ...................... . ................... 297
BIBLIOGRAPHY ............................................................................................. 298

xi

Table 2-1

Table 2-2

Table 2-3

Table 2-4

Table 2-5

Table 3-1

Table 3-2

Table 3-2

Table 4-1

Table 4-2

Table 4-3

Table 4-4

Table 4-5

Table 4-6

Table 4-7

Table 4-8

Table 4-9

LIST OF TABLES

Human Factors Leading to AMT Success .................................... 36
Deskilling Evidence ...................................................................... 40
Unexplored Questions .................................................................. 44
Authors who have Concluded Choice .......................................... 53
Explicit tests Of Choice ................................................................. 57
Case Study Conclusions .............................................................. 81
Appropriate skill level ................................................................... 1 19
Appropriate Skill Level ................................................................. 178
Correlation Matrix for Product/Process Variables ........................ 145
Correlations Between EU and Product/Process Variables ........... 147
Sample Breakout by Industry ....................................................... 152
Comparison of Batch Sizes .......................................................... 158
Comparison Of Churn ................................................................... 160
Comparison of Families ............................................................... 161
Comparison of MP8 Deviation ..................................................... 162
Comparison of Product/process Change ..................................... 163
External Complexity: Individual Items and Summed Index .......... 165

xii

Table 4-10 Relationships Between Variables Hypothesized tO Drive Choice 171

Table 4-11 Polynomial Regression ................................................................ 176
Table 4-12 Categorization Of Variables .......................................................... 179
Table 4-13 Distribution Of Variables by Category ...... A ..................................... 180
Table 4-14 Respondents Per Cell .................................................................. 180
Table 4-14 Correlations for FIT and success ................................................. 183
Table 4-14 FIT1’s ability to correctly categorize success ............................... 184
Table 4-15 FIT2’S ability to correctly categorize success ............................... 184
Table 4-16 FlT3’s ability to correctly categorize success ............................... 185
Table 4-17 FIT5’s ability tO correctly categorize success ............................... 185
Table 4-18 Step 1 Of Moderated Regression ................................................. 187
Table 4-19 Step 2 of Moderated Regression ................................................. 188
Table 4-20 Descriptive Statistics by Discretion .......................... ~ .................... 194
Table4-21 Regression Of Product/Process Variables on Skills ..................... 196
Table 4-22 Correlations For Environmental Measures ................................... 201
Table 4-23 Dependent Variable: Total Preparation Time ............................... 202
Table 4-24 Total Preparation Time ................................................................ 203
Table 4-25 Correlations WIth Age Of Installation ............................................ 204
Table 4-26 Skills to Success for FMS ............................................................ 214
Table 4-27 FIT and Adoption Success For FMS ............................................ 216
Table 4-28 Correlations Between Fit and Success - CNC ............................. 217
Table 4-29 Ability of various FIT variables to Categorize for CNC ................. 217
Table 4-30 Key FMS Variables Sorted by Skills ............................................. 218

xiii

Table 4-31 Drivers Of Choice Regressed on Skill Level for CNC ................... 221

Table 4-32 Machine Tools .............................................................................. 227
Table 4-33 Stamping Dies ......................................... _ ..................................... 230
Table 4-34 Auto Parts ................................................................................... 232
Table 4-35 Diesel Power ................................................................................ 235
Table 4-36 Injection Molds ............................................................................. 238
Table 4-37 Furniture ....................................................................................... 240
Table 5-1 Primary Results .............................................................................. 245
Table 5-2 Secondary Results .......................................................................... 246

xiv

Figure 1-1
Figure 1-2
Figure 2-1
Figure 3-1
Figure 3-2
Figure 3-3
Figure 3-4

Figure 3-5

Figure 4-1
Figure 4-2
Figure 4-3
Figure 4-4
Figure 4-5
Figure 4-6
Figure 4-7
Figure 4-8

Figure 4-9

LIST OF FIGURES

A complete model Of AMT adoption success ............................... 16
Conceptual model Of skills and AMT adoption success ............... 18
Schematic Of Literature review ..................................................... 23
A complete model Of AMT adoption success ............................... 85
Basic conceptual model ............................................................... 87
Conceptual Model Of Skills and AMT adoption success ............... 88
Interaction of managerial discretion and product change ............ 102
Interaction Of managerial discretion and environmental

complexity .................................................................................... 103
Skill Level and Adoption Success ............................................... 175
Product Process Change and Skill Level .................................... 189
Environmental Uncertainty and Skill Level .................................. 190
Skills to Product Process Change: High Discretion ................... 191
Skills tO Product Process Change: Low Discretion .................... 191
Environmental Uncertainty tO Skills: High Discretion ................. 192
Environmental Uncertainty to Skills: Low Discretion .................. 193
Skill Level and Self Assessment Of Adoption Success ............... 199
Skill Level and Adoption Success - FMS .................................... 213

Figure 4-10 Skill Level and Adoption Success - CNC .................................... 215

Figure 4-11 Product Process Change and Skill Level (FMS) ......................... 219
Figure 4-12 Environmental Uncertainty and Skill Level ................................. 220
Figure 4-13 Product Process Change and Skill Level (CNC) ......................... 222
Figure 4-14 Environmental Uncertainty and Skill Level (CNC) ...................... 223
Figure 4-15 Environmental Uncertainty tO Skills (CNC) ................................. 223
Figure 5-1 A Process Map Of the Appropriate Level of Training .................. 254

Chapter 1
OVERVIEW OF THE RESEARCH

1.1 Introduction

Flexibility is Often seen as one Of the key dimensions on which firms will
compete in both the near and long term future (De Meyer et. al. 1989, Gem/in
1993, Swamidass and Newell 1987). Gerwin (1982, 1993) identified 5 different
dimensions Of flexibility ; product mix flexibility, parts flexibility, routing flexibility,
design change flexibility, and volume flexibility, each Of which has two aspects;
range and time. A production system is more flexible if it can handle a broader
range of possibilities on one or more dimensions, or if changesican be made
faster. Not only are there many types of flexibility but there are many ways to
achieve flexibility including holding excess capacity, using cross trained workers
(V. Smith 1994), and Advanced Manufacturing Technologies (AMT). The ability
of AMT to mitigate the cost/efficiency tradeoffs inherent in many Older
technologies (Nemetz and Fry 1988) has lead tO a large amount of practitioner
and academic interest in these technologies.

AMT is a term that covers a broad range Of computer controlled
automated process technologies, including everything from stand alone robots,

flexible manufacturing systems (FMS), and Computer Integrated Manufacturing

(CIM) systems. Robots and Computer Numerically Controlled (CNC) devices
can often be substituted for existing non-computer controlled equipment with little
or nO change tO the shop floor. Flexible Manufacturing Systems (FMS) and CIM
systems are different in the sense that simultaneous Changes to the entire
production system are Often required during the installation process.

Despite the attention that FMS has received in the literature the
technology has not been widely adopted (Handfield and Pagell 1995; Mansfield
1993 a; 1993b) especially when compared to other types Of programmable
automation such as CNC machines. Mansfield (1993a) explains the slow
diffusion Of FMS using a model that predicts the rate Of adoption as a function Of
the number Of firms already using the technology and the size Of the investment
required. Mansfield’s study Of 12 innovations found that in general, diffusion
tended to be faster for those technologies that were more profitable and required
a smaller investment. This model is used by Mansfield to partially explain the
diffusion pattern of FMS.

Mansfield’s conclusions make sense intuitively. FMS is expensive and
may not perform well when measured on Return on Investment (ROI) criteria.
Nevertheless, the benefits Of FMS extend beyond lower costs, and Mansfield’s
model does not fully explain why firms who might install FMS for non-financial
reasons do not. A broader model Of FMS diffusion provided by Handfield and
Pagell (1995) proposes five attributes that affect the rate Of adoption: relative
advantage, compatibility, complexity, trialability, and Observability. These criteria

are more expansive but share an important trait with Mansfield’s model; the

number Of adopters Of a technology is related tO the number Of people already
using it successfully.

A meta analysis (across 75 innovation studies) on the diffusion of various
innovation characteristics (Tornatzky and Klein 1982) found that relative
advantage, compatibility and complexity issues are most important in the
diffusion literature. The relative advantage Of FMS (mitigation Of the tradeoff
between flexibility and efficiency) is Often not exploited because systems run only
a few parts in fairly high volumes (Venkatesan 1990). FMS is technically
complicated and not as easy to install as a robot. Finally, FMS may not be
compatible with the existing structure and infrastructure of firms. Compatibility
issues, especially those relating tO management of the workforce are Often
blamed for the high number Of AMT installations that do not reach full adoption
or fail completely.

There is some evidence that companies successful at implementing AMT
use commitment strategies to manage their human resources (Arthur 1994).
There is also evidence that companies using AMT use policies that are
consistent with knowledge work (Snell and Dean 1994). This body Of work
assumes that highly skilled workers are a key tO successful adoption Of AMT.
Nevertheless, these studies are not conclusive and ignore a significant body Of
literature in fields such as Sociology and Industrial Relations. Many researchers
in these fields posit that the way in which a company uses a technology is not
determined by the technology itself but rather by the firrn’s environment. This

body of literature is described in chapter ll.

A number of authors (Spenner 1983, 1988, Smith 1992, Form et. al. 1994)
have summarized the literature on the impact Of technology on the workforce and
concluded that technology’s impact is a result Of the way that management
reacts tO their environment rather than anything inherent in the technology. This
view is Often referred to as strategic choice (Child 1972). This dissertation uses
the strategic choice framework to determine the appropriate Skill levels Of AMT
operators in a specific environment.

The strategic Choice framework is used to explain the skill level Of
operators in CNC and FMS environments in order tO provide insights into how
the choice Of worker skills impacts the success Of technological adoption. A
model Of AMT adoption success based on factors internal and external to the
firm was developed to address the paradox dealing with the interaction Of human
resources. The Objective of the dissertation is tO help explain why successful
firms use different levels Of operator skill with AMT when theory dictates that a
high skilled workforce is a prerequisite for success. Multiple case studies were
conducted using structured interviews compare companies doing similar work,
with different technologies, as well as companies doing different work with the
same technology.

1.2 Research question

AMT success has been examined from a number Of different perspectives
and a complete model of the factors that drive successful adoption is beyond the
scope of this research effort. This study focuses on the relationship between the

skill level of the Operational employees and the success companies have with

installing FMS and CNC equipment. The basic research question which guided
this investigation was:

Is there a relationship between the appropriateness of the

decisions management makes regarding the skill level Of

operational employees in AMT installations and adoption success?

A theoretical model of the relationship between the level Of
product/process change, the complexity Of the firms external environment and
the nature Of industrial relations at the installation will be used to explain the skill
level that management chooses for Operation Of the AMT.

The model was tested through the use Of in depth structured interviews
conducted at the site Of the technological installation. Qualitative and
quantitative data was collected from multiple employees at the installation site.
Qualitative data was used tO refine measures and constructs, as well as tO
provide insights intO anomalous and/or unexpected findings. The quantitative
data was statistically analyzed in order to further understand the linkage between
Operational employee skill level, strategic choices, and adoption success.

1_.2. 1 Specific reseth hypothesis:

The following specific research hypotheses have been generated to

address the relationship between managerial choices regarding the skill level Of

the workforce and AMT adoption success.

H1a: The skill level Of the operational employees is positively related to
successful adoption of CNC and/or FMS.

H1b: Firms who choose appropriate levels Of skills for Operational employees Of
CNC and/or FMS will be more successful adopters than firms who choose
inappropriate skill levels.

Hypotheses H1a and H1b are competing hypotheses. H1a tests the
proposition that high skilled workers are a prerequisite for the successful
adoption Of AMT. H1 b tests the strategic Choice framework.

H2: The level and unpredictability Of product/process Change is positively related
tO the skill level management chooses for the operators Of CNC and FMS.

H3: The level and unpredictability Of environmental uncertainty is positively
related to the skill level management chooses for the operators Of CNC and
FMS.

H4a: Managerial discretion does not moderate the relationship between
product/process Change and skills.

H4b: Managerial discretion does not moderate the relationship between
environmental uncertainty and skills

Before describing the model associated with the hypothesis, it is
necessary tO define the underlying paradigms and terms that will be used
throughout this dissertation.

1.3 Background
1.3.1 . The strategic choice franfiwork

Child (1972) first developed the paradigm of strategic choice as an
organizational model. He viewed existing paradigms that explained variance in
organizational structure based purely on environment, technology, and or firm
Size as constraints tO the decisions that are made by the dominant collations in a
firm (top management). Child’s original formulation has been applied tO a
number of different studies in the organizational theory literature.

The strategic choice framework has also been applied in sociology and
Industrial relations to explain the outcomes Of changes in technology (From et.

al. 1994), different uses Of teams in similar environments (Mueller 1994), and

changes in bargaining strategies (Birecree 1993). In all Of these studies one or
more environmental factors such as technology, environment, or size has been
controlled, yet firms have made different choices when approaching the same
problem.

In this literature the problem is approached from the perspective that
management has some discretion over what they do, even if there are
environmental factors that limit choice. An important part Of Choice is the set Of
political factors that impact peoples’ decisions (Astley and Van De Ven 1983).
The Industrial Relations/ Sociology approach is not as structured as the
approach tO strategic choice in the organizational theory literature in the sense
that the environment is considered one driver Of Choices while the organizational
theory literature usually separates environmental drivers from choice.

Multiple explanations for organizational structure exist ( see Hrebiniak and
Joyce 1985 for a thorough treatment); including strategic choice, environmental
determinism, and institutional theory. These different perspectives are Often
seen as competing, with environmental determinists attaching little importance to
managerial decision makers (Astley and Van de Ven 1983). On the other hand
the proponents Of choice (in the organization structure/strategy domain) pay little
attention to the constraints that may be imposed by environment (Whittington
1 988).

The dichotomy between determinism and choice has lead some
researchers to propose a continuum where choice ranges from unconstrained at

one end to heavily constrained by the environment at the other end (Whittington

1988, Hrebiniak and Joyce 1983, Bluedorn et. al. 1994, Pennings 1987). The
proponents of what is deemed a constrained choice framework in the structure
strategy literature are similar to the users of Choice frameworks in Industrial
relations/Sociology. However a key difference exists In the level Of analysis
usually used in the two disciplines. Structure/Strategy researchers Often deal
with the entire firm or an entire industry (i.e. Dess and Beard 1984), while much
Of the Industrial Relations / Sociology literature deals with a small part Of a firm,
or a single work group (i.e. Zicklen 1987).

This dissertation borrows from both the Industrial Relations I Sociology
and Structure/Strategy disciplines to develop a conceptual model Of adoption
success, but the level Of analysis will be much more in line with the Industrial
Relations] Sociology literature. Therefore the notion Of choice used in this
dissertation is the choice framework used in these fields (which is very similar tO
the constrained choice framework used by authors such as Bluedorn et. al.
1994). Management is assumed to choose a skill level for their workforce based
on factors both internal and external to the firm. Further adoption success is
assumed to be related to how well management has matched their choice tO the
environment in which they compete.

1 .3.2 Level Of analysis

The unit of analysis employed in the study Of AMT adoption is the system
or installation level, rather than the plant or firm level. A plant or firm level study
is not appropriate for a number Of reasons. First firm level data may not explain

what is happening at the actual AMT level. Secondly firm level data on

performance may be related tO many extraneous factors unrelated to the
performance Of the AMT. Plant level data can also be troublesome because
some AMT installations may be the entire plant while others may be a small part
Of a plant. Installation level data is useful because it provides the Opportunity tO
study policies that may be specific to the AMT.

1.3.2.2 Types of installations

This study examined installations using two types Of process technology;
Flexible Manufacturing Systems (FMS) and Computer Numerically Controlled
equipment (CNC). The two types of technology were chosen for a number Of
reasons. FMS, in and Of itself is Of interest due to its slow diffusion. FMS is also
interesting because it is usually posited that the technology requires higher
Operator skills. However authors who theorize that high skills are a prerequisite
for adoption success disregard the literature that indicates that many firms use
FMS with low skills.

Firms using CNC technology are included for several reasons. First the
inclusion of a second form Of technology makes conclusions far more
generalizable. CNC equipment is Often seen as being at the Opposite end Of the
AMT spectrum from FMS. FMS is an integrated process technology, while CNC
is Often installed as a stand-alone machine. The two technologies span a gOOd
deal of the range Of AMT Options.

CNC is also far more popular than FMS SO including CNC allows for a
larger sample size. Including CNC also makes the results Of the study more

useful for practitioners. Finally, and perhaps most importantly, the inclusion of a

10

second form of technology provided a way to test for the impact Of the specific
technologies on choices.
1. 3;. 1 Studying skill levels and adoption success

The success of AMT adoption is Often posited tO depend on the way
managerial issues are approached (Snell and Dean 1992). Most authors
examining the relationship between human issues and the success Of an
installation look at the type Of human resource policies used, or the overall
human resource strategy. Skill levels are usually explicitly studied by authors
such as Kelly (1990) who are interested in the impact Of the technology on
worker skills, not on the outcomes Of the adoption from a business perspective.

However, skill levels are implicitly studied by the authors who look at the
types of HR policies and strategies firms use. Walton and Sussman (1987) note
that a commitment strategy is one where employee’s knowledge and skills are
upgraded. Arthur (1992) notes that a commitment maximizing HR strategy is
one with a large percentage Of high skilled workers. The same authors note that
control policies are linked tO low skill workers with little autonomy. Snell and
Dean (1992) couch the debate in the human capital perspective. They note that
firms can invest in operators skills and knowledge tO enhance productivity. The
specific policies they study are usually linked to commitment strategies.

The preceding discussion may imply that HR policies or strategies are the
appropriate lens to study the management of the workforce. However, skill
levels Of the workforce are more appropriate then HR polices for three reasons.

First, is the content Of commitment. Most authors agree on the content Of control

11

strategies. However, the work Of WOOd and Albaese (1995) shows that the
policies that make up a commitment strategy vary from firm to firm. This lack Of
a clear definition makes it difficult to measure what type Of strategy a firm is
following. However, clear and validated measures Of Skill exist.

Measuring skill is also justified based on the definitions Of “control” and
“commitment”. Almost all authors agree with Arthur (1992) who notes that
control policies are linked to low skill requirements. Commitment polices are not
as well linked to high skill requirements. However the policies usually associate
with commitment ( extensive training, self management, high wages and high
levels Of autonomy) generally point to a larger investment in the workforce.
Based on the theory Of human capital (Snell and Dean 1992) a larger investment
in people should lead tO increased knowledge and skills. The link between skills
and HR strategy is not perfect. However, low skills are part Of the definition Of a
control Strategy, while higher skills are, all things being equal, usually part Of the
definition Of a commitment strategy.

The final justification for looking at skills rather than the type of policy
relates to the multi-disciplinary nature Of the dissertation. Skill levels are
discussed in all Of the literature that was used tO build the model and develop the
measures for the dissertation. Specific human resource policies and strategies
are not. By examining skills it is possible tO weave the multiple disciplines
together in a manner that would be more difficult if only HR strategies were

examined.

12

Skills are then a rough proxy for the type Of HR policies and strategies the
firm uses. Skill levels are easy tO define and measure. Low skills are generally
linked to a small investment in human capital and a control strategy. High skills
are usually linked to a larger investment in human capital and a commitment
strategy.

1.3.3. 2 Skill definition

This study crosses a number Of diverse disciplines describing the
relationship between technology and the workforce. Studies in industrial
relations and sociology Often focus on the impact Of technology on the workforce
in terms Of worker’s skills. These studies look for evidence Of skilling (increase in
worker skills due to the installation Of technology) or deskilling (decrease in skills
due to the installation Of the technology) with a focus on change. Studies in
human resources management Often lOOk at sets Of polices that are used tO
manage the workforce. Walton (1985) introduces the notion Of two generic
strategies to manage the workforce: control and commitment. Control strategies
are composed Of human resource policies that result in little autonomy or control
Of work fOr the operational workers. Commitment strategies are composed Of
human resource policies, which make the operational employees a main human
resource in the firm through intensive training, enriched job designs and a
management philosophy that places control Of work in the hands Of those who
actually do the work. Studies have looked at commitment in general (WOOd and
Albaese 1995) or policies linked tO commitment strategies (Snell and Dean

1994)

13

Commitment strategies, and their related policies are equated with using a
high skilled workforce, while the concept Of control has been used to describe a
low skill or degraded workforce. The focus Of this study was not on the Changes
wrought by AMT, but rather why top management chOose a specific way Of
managing the workforce vis-a-vis the Skills the workers possess. When
discussing skills it was not assumed that skills have always increased or
decreased, so the terms “low skill” and “high skill” were preferable tO “deskilled”
or “skilled”. However tO synthesize studies from various fields it was occasionally
be necessary to use other terms. WIth this in mind the terms commitment,
upgraded, Optimistic, and skilling were all used tO denote a high skilled
workforce, while deskilled, pessimistic, degraded, downgraded, and control were
used to denote a low skilled workforce.

The skill level Of the workforce refers to “job complexity: the level, scope,
and integration Of mental, interpersonal, and manipulative tasks required in a jOb“
(Spenner 1980 ). A low skill job, all things being equal, is composed Of simple
tasks that are easy to learn and to do, while a high skill job is harder to learn and
do.
1.3.3.3 Operators

For the purposes Of this work the operators were defined as those
employees most responsible for the day to day running Of the equipment. In
other words the Operator is the person most likely tO be at the machine.
Operators may perform a variety Of tasks (maintenance, programming, quality

control and the like) but their primary function is to run the machines.

14

Employees who are responsible for maintenance, programming, and or
quality control who dO not normally run the machines, but do occasionally work
on them are not Operators for the purposes Of this work. Nor are managerial
employees who occasionally fill in for absent employees.

Finally in situations where a firm had different policies for certain shifts
information for all shifts was collected. The most prevalent set Of policies was
used in the analysis.

1 .3.4 Appropriateness Of choice

There are a number Of criteria on which to judge the appropriateness Of
choices. For instance Bravennan (1974) argued that management always
lowers skills in order to control the workforce. He views this as inappropriate
because low skilled work is Often degraded work that retains little dignity. In a
similar vein it can be argued that a high skill environment with enriched jobs that
provide the workforce with a large degree Of intrinsic motivation is always an
appropriate choice because the workforce is more satisfied.

This study applied the term “appropriate choice” from a different
perspective; business efficiency. The study assumed that, all things being equal,
higher skill workers will cost more than low skill workers. If skills are wasted the
firm is not being as efficient as it could be, hence the choice was inappropriate.
When a firm chooses a skill level appropriate for its environment, it has made an

efficient choice where there are no skill shortages or surpluses.

15

1.3.5 AMT

FMS is part Of a broader continuum Of computer controlled automation
that ranges from simple stand alone equipment such as robots at one end to
completely integrated CIM systems at the other. Stand alone equipment is ofth
installed one piece at a time to create islands Of automation, while more
integrated technologies such as FMS and CIM are installed as complete
systems. Stand alone equipment such as CNC and Robots have diffused rapidly
(Mansfield 1993) while integrated technologies have not.

FMS is defined as an automated batch manufacturing system
consisting Of NC machines, linked by automated material handling devices, that
perform the operations required tO manufacture parts” (Stecke 1985).

1 .4 Scope of Dissertation

A complete model of adoption success for AMT would encompass a very
large nUmber Of variables (see Figure 1-1) and be difficult if not impossible tO test
statistically for a number Of reasons. The first problem is one of sample size.
Not only would it be extremely expensive and time consuming to examine all the
possible factors that impact AMT adoption success, but there are a very limited
number of FMS installations worldwide tO study (Handfield and Pagell 1995).
The second problem involves interpretation Of results; a large and complex
model would contain many confounding factors with an impact on success.
Finally some of the factors that impact success have already been tested (i.e.

Beatty 1993).

16

Figure 1-1 - A complete model of AMT adoption success

 

Technical Factors

 

 

 

Supplier Quality
Software
Tools
number
control
Information
Systems
maintenance
timing

 

 

 

 

 

Successful

 

 

Managerial Factors _. Adoptlon

 

 

 

 

 

Integration
across functions
with suppliers
with customers
High skill workforce
Technology strategy
Fit with business strategy
Process champion
Maintenance
who performs it
Industrial relations
Performance measurement
F MS
Operators
Human resource policies
selective staffing
comprehensive training
developmental appraisal
externally equitable
rewards

 

 

 

 

 

17

This dissertation focused on studying the relationship between managerial
Choices of operator skill levels Of CNC and FMS installations and adoption
success. Prior studies have adopted a deterministic view Of operator skills
(Walton and Sussman 1987, Saraph and Sebastion 1992), a technological
determinist view (Braverman 1974) or examined only the skill level Of the
workforce and not adoption success (Davies 1986). This study integrates:

. the choice Of worker skills from the industrial relations literature (Spenner
1988, Form et. al. 1994)

. the determinants Of organization structure and job design from the
organizational structure/strategy literature as well as the human resource
literature

. the successful management of technology from the operations management
and engineering literature

This dissertation presents a framework that explains AMT adoption success as a
function Of the appropriate managerial choices regarding the skill. level Of the
Operational workforce.

The model guiding the research is presented in Figure 1-2. This model
posits that managers make decisions as to the skill level required Of the
workforce based on three broad factors; the level of product/process change for
the examined technology, the complexity Of the environment that the plant
Operates in, and the type Of industrial relations the company has. Organizational
structure and strategy researchers have done a multitude Of studies that Show
that organizational environment impacts the choice Of structure (i.e. Burns and
Stalker 1961) impacting job design and fields (industrial relations, human

resource management, sociology, Operations management) look at the impact Of

18

Figure 1-2 - Conceptual model of skills and AMT adoption success

 

lProduct/Process Change

 

lManagerial Discretion l_

 

 

Unsuccessful
Adoption

Goals of AMT not
achieved

 

 

 

4

Inappropriate
skill level

 

 

Environmental Complexity

 

 

 

I

 

Skill Level of Workforce
Total preparation time for

 

Operator jobs

 

 

 

Appropriate skill level

 

Successful Adoption:
Goals Of AMT achieved

 

 

 

19

internal factors on polices, with many authors positing a worker skills. However
there is some disagreement as to the impact external factors have on specific
choices regarding the workforce (Wood and Albanese 1995). Many studies in a
number Of strong link between internal product proceSs factors and worker skills
(i.e. Kelly 1990). Finally the type Of industrial relations a company has have
been shown to impact choices Of policies when the environment and technology
is controlled for (Mueller 1994). This study resolves a gap in the literature
between the proponents Of choice and the proponents Of a single right way to
manage the workforce Of AMT.

1.5 Results

The answer tO the research question proposed at the beginning Of this
dissertation is; yes there is a relationship between the appropriateness Of the
decisions management makes regarding the skill level Of the operational
employees in AMT installations and adoption success. This conclusion was
reached because the analysis Of the data indicated full or partial support for all Of
the proposed hypotheses.

Hypothesis H1A is supported because there is nO evidence Of a
relationship (linear or curvilinear) between adoption sucCess and skill level. In
addition the sample contained a number Of successful companies using low
skills. Therefore there was enough evidence that low Skills could be used in
successful installations to test the remainder Of the proposed hypotheses.

H1 B is also supported. Installations where management made an

appropriate choice were more successful than installations where inappropriate

20

choices were made. The best measure of appropriateness matched skill levels
to the internal environment composed of managerial discretion and product
process change. The original formulation Of appropriateness which included
both internal and external factors was significantly related tO success ( p = .095)
but did not have as strong an effect as the formulation based only on internal
factors (p = .035).

The analysis also provided support for H2 and H3; increases in
product/process change and environmental uncertainty were associated with
higher levels of skills. H4 received partial support. It was hypothesized that
managerial discretion would moderate the relationships between skills and
product/process change and environmental uncertainty but instead managerial
discretion had a direct effect.

There is an important disconnect between the drivers Of skill choices and
the most appropriate choice Of skills. The analysis suggests that mangers chose
skills based on internal and external factors. However, the analysis also
suggests that the highest level Of success will be achieved by matching skills
only to the internal environment. Therefore a key secondary conclusion Of the
analysis is that managers in general could improve performance if they made
infrastructure choices based only on the internal environment.

1.6 Structure of Dissertation

The following chapters Of the dissertation review the relevant literature,

develop hypotheses and models, discuss the research design and methods,

detail the analysis, and discuss the results. Chapter 2 reviews the literature on

21

AMT adoption and the specific literature on the impact Of humans in AMT
systems. In doing so we will expose a paradox between the adoption literature
which contends that high skill are a prerequisite for success, and industrial
relations and sociology which posit that AMT is used with a variety Of skill levels.
Chapter 2 also reviews the relevant literature on strategic choice and the use Of
the strategic choice framework. Chapter 3 summarizes the pre-dissertation case
studies, develops hypotheses, models, and measures, and explains the data
collection and data analysis methods used. Chapter 4 discuses the analysis
performed on the data. Finally Chapter 5 discusses the results as well as future

research and conclusions.

Chapter 2

LITERATURE REVIEW

2.1 Introduction

The following chapter will review the literature relevant tO building a model
Of AMT success based on the appropriateness Of managerial choices regarding
the skill level Of the people who Operate the technology. Figure 2.1 is a graphic
representation Of the path that this chapter will take tO get from factors that
impact adoption success to the factors that drive managerial choices.

The chapter begins by reviewing the literature on adoption success/
failure. Although a large number Of factors have been shown to impact the level
Of success Of an installation, a number of authors have focused on the
importance Of managerial issues. A key managerial issue is the way the
workforce is deployed. Authors such as Walton and Sussman (1987) have
suggested that commitment policies are a prerequisite for the success Of an
installation. However, there is a large body of evidence that indicates that
successful firms use workers Of varying skill levels with AMT.

Section 2.3 Of this Chapter explores the apparent paradox Of a portion Of

the literature suggesting that a high skilled work force is a necessity for adoption

22

23

Figure 2.1 - Schematic of Literature review

 

Section 2.2
Factors impacting AMT Success

 

Section 2.2.2
Managerial Issues

 

Section 2.3
Paradox
Involving The
kill Level of th
Workforce

       
     
  
  
  

Section 2.4
Using the Strategic
Choice Framework to
Explain the paradox

   
   
    
  
     
 
  
  

Section 2.4.2
Authors who

have concluded
choice of skills

  

 

  

Section 2.4.3
Explicit tests of choice of
skills

 

  
  
   

Section 2.5
of all factors that im act
Section 2.6

Generic Choices

  

success while another portion suggests that a number Of different skill levels are
used with AMT. Section 2.4 Of the literature review looks at a specific framework

to explain the apparent paradox, strategic choice. This section begins by

24

exploring the appropriateness of using a strategic choice framework to explain
the paradox presented in section 2.3. The second sub-section reviews studies
Of the impacts Of AMT on worker skills that have concluded that management
makes a choice as to the skill level of the workforce. The final sub-section looks
at studies that start with the notion Of choice and look for the causes Of choice.

Section 2.4 deals with studies that addressed the issue Of technology and
worker skills, and or studies that used the strategic Choice framework in the field
Of labor and industrial relations. Section 2.5 lOOks at a much wider body Of
literature on the impetus for skill choices. The section concludes with a
classification scheme for the drivers of managerial choice, as they have been
identified in the literature.

Section 2.6 introduces two generic strategic choices firms can make
based on the factors reviewed in section 2.5. The final section Of the literature
review Summarizes the way the strategic choice framework will be used tO
address the paradox presented, in a model that explains AMT adoption success
as a result of management making appropriate strategic choices regarding the
skill level Of operating employees
2.2 General criteria for successful adoption

There is a large body Of work that suggests some Of the requirements for
successful adoption Of AMT (see Appendix A). These include the presence Of; a
process champion, a properly designed information system, new cost accounting

procedures, improved supplier quality, interfaces with other functions, a

25

technology strategy, and human resource policies that lead to a high skilled
workforce.

The research covers a broad range Of disciplines and topic areas that can
be subdivided into managerial issues and technical issues (Shani et. al. 1992).
Technical issues include:

0 problems with software used for integration Of various parts Of
the system
tool rules
quality of supplied inputs

the information system
machine maintenance.

2. 2. 1 Technical issues thatfiimpact adoption success
2.2.1.1 SOME;

Many firms have trouble with their software, especially the software that is
used tO link machines or process (such as Computer Aided Design (CAD)) from
various suppliers (Fry and Smith 1989, Gupta et. al. 1993). This problem can be
avoided if a firm uses a single supplier for the entire system. In Beatty’s 1993
study, companies who followed a “stereo components” approach to system
design Often found that they had bought the best Of everything but had the worst
system. Companies who followed the best overall system approach did not
always have the most cutting edge pieces Of equipment but they had systems
that worked.
2.2.1.2 TOOI rules

Tooling is another area that impacts the success Of an installation

(Meredith 1987, Fry and Smith 1989, Gupta et. al. 1993). Gupta et. al. (1993)

26

note that the planning and control Of the tools necessary for the various
machines critical due to the costs Of tooling. Additional costs occur in the
absence of a tool, or when a tOOl breaks. The nature of tool problems has lead
to a number of mathematical models to determine part routings and the like (for
example see Tawegoum Castelain, and Gentina 1994) to minimize disruptions
and or costs.
2. 2. 1.3 Supplier quality

The handling Of suppliers is a managerial problem, but the impact Of poor
supplier quality is a technical problem that has major implications for AMT,
especially FMS (Boer, Hill, and Krabbendam 1990, Knorr and Theide 1991,
Mieskonen 1991, Boer and Krabbendam 1992). FMS by nature require inputs
that are of higher quality than the inputs for stand alone machines and or non-
programmable machines. Stand alone equipment may allow formore
opportUnities for an Operator to make adjustments due to variances in inputs. In
an FMS this adjustment is much more difficult if not Impossible. An additional
problem in an FMS is that poor quality inputs will go through a greater number of
steps before the problem is Spotted, increasing the costs Of rework and or scrap.
2.2.1.4 Information SLstems

Information systems are really a unique type Of software problem. Rather
than being responsible for how various machines work together to process work,
the information system transmits the data about the work tO a large number Of
employees, suppliers, customers and machines in various locations.

lnforrnation systems, like supplier quality, have elements of both managerial and

27

technical problems. The types Of data recorded and their availability are a
managerial problem. But the system itself must link with other hardware and
software and this linkage is a technical problem. The information system and the
data it transmits are Often a vital link in successful adOption (Badiru 1990, Attaran
1992). I
2.2.1.5 Maintenaacg

Maintenance is critical in a system that is complicated and integrated.
Breakdowns do not cause rerouting or some other inconvenience, but complete
shut-down Of the line. The integration Of an AMT system can be quite
problematic when the system breaks. This is why many authors (Majchrzak
1988, Boer, Hill, and Krabbenam 1990, Jackson and Wall 1991 , Diaz 1991) point
to preventive maintenance as a key to the successful Operation Of an AMT.
Authors such as Majchrzak (1988) and Jackson and Wall (1991) take an
additional step and recommend that most maintenance should be performed by
those closest to the machines, the operators.
2. 2.2 Managerial issues that impact adoption success

Technical issues are very important tO the success Of AMT. However a
number Of authors have found that the most difficult problems lie not in the
system itself, but in its management (GeMin 1982, Meredith 1987, Adler 1988,
Boer, Hill, and Krabbenam 1990, Gupta et. al. 1993, Bessant 1993, ElangO and
Meinhart 1994). Managerial problems may be more complex than technical
problems and are more likely tO inhibit the long term success Of an installation.

Managerial issues cover a broad spectrum Of decision areas including;

28

strategic planning

integration across functions

top down management Of the innovation

3 process champion

changes in performance measurement systems
a variety Of management Of people issues .

2. 2. 2. 1 Strategic planning

In order for an installation tO be successful the AMT should not be treated
as just another piece Of equipment but should instead be treated strategically
(Adler 1988, Vineyard 1993). The technology should have well defined goals
that fit the firm’s overall strategic goals (Vineyard 1993). An installation that does
not have a strategic vision (Adler 1988), or a well defined goal(s) will have
trouble meeting needs that remain undefined and may Change over time.

Part Of the strategic plan should be the manner in which the technology
will be introduced. Some authors (Gerwin 1982, Shani et. al. 1992, Miller and
Gilbert1992) suggest that automation should be installed in phases such as
converting to cells with the existing technology before installing new technology
(Vonderembse and Wobser 1987). Phased approaches may be difficult with
FMS, and they may not matter. Beatty (1993) found that a phased approach
worked as well as a turnkey approach in the installations studied. Phasing may
help to allay psychological fears (Miller and Gilbert 1992) or build confidence with
the technology, or it may lead to some Of the software integration problems
previously mentioned.

A final strategic issue that may impact the success Of the installation is the

actual reason for installation. Groth (1993) notes that an innovation can be

29

pushed or pulled into an organization. Innovations that are pushed are Often
installed as a test case either for a parent company or for the government (see
Nobel 1984 for a discussion Of NC tools and their pushed development).
Innovations that are pulled are usually installed due to a specific market need Of
the facility doing the installation and tend tO be more successful.
2. 2. 2.2 Integration across functions

One Of the reasons that a technology strategy is needed is due to the
integration that AMT can create across the firm. Pavitt (1990) notes that a firm
needs a technology strategy to integrate all functions whose work might be
affected by a new technology installation. AMT can have an impact across a
wide variety Of functions both within the Operations arena: manufacturing, quality
control, maintenance, purchasing and across the firm: marketing, human
resources, accounting (Goldhar and Lei 1994). Firms who view the technology
as purely the domain of manufacturing will not be prepared for other potential
changes such as: the need for greater supplier quality, Changes in product
_, mixes, the need for Changes in job design, Changing human resource practices,
and the need for new performance measures, until after the technology is
installed and problems have arisen.
2. 2. 2.3 Top management support

In order to insure that the technology is installed strategically and with
input from all functions some authors (Shani et. al. 1992, Gupta et. al. 1993)
have concluded that top management support is necessary for the installation to

be successful. Due to the cost Of the equipment the decision to implement

30

cannot be purely bottom up. Moreover the integration required to implement
many AMTs requires someone at the top to get all parties on-board.
2. 2. 2.4 Process/technology champipp

Top management at some companies not only. supports the installation,
but may be the process champion. In firms where top management does not
take a championing role (one that goes well beyond just supporting the AMT) a
separate champion will be needed. In Beatty’s 1993 study all successful
installations had a process Champion; a person who sold the project to all those
who would be impacted and made sure that installation went as planned. In
addition to Beatty’s findings, other authors ( GenIvin 1982, Meredith 1986) have
supported championing the technology throughout the entire adoption and
installation process as a prerequisite for success.

Technology strategies that integrate across functions, which are
supported by tOp management and/or a process champion are prerequisites for
an installation to be successful. However, there are a host of other issues, such
as performance measurement and a variety Of people management issues that
impact the manner in which the adoption process develops once the technology
is actually in place.

2. 2. 2.5 Performance measurement for the system and its operators

Performance measurement issues can be broken into two issues; 1)
measurement of the AMT’s performance and 2) measurement for the employees

Of the system. People issues are really human resource issues, while the

31

measurement Of system performance usually falls under the banner of cost
accounfing.

Skinner (1986) noted that traditional costs accounting procedures are
problematic for systems where labor is no longer a major cost. By focusing on
labor costs (which may account for a very small part of total manufacturing costs)
managers spend an inordinate amount of time on a minor problem ignoring real
problems and real system performance. This problem is exacerbated for an
FMS where there may be little or no direct labor. GenNin (1981) notes that even
when accounting and manufacturing try and work together it may be difficult to
establish standards against which to judge behavior. Therefore a number Of
authors (Maskell 1989, Coulthurst 1989, Son 1990, Currie 1992) have suggested
alternative performance measures for AMTs. Maskell suggests that companies
should measure quality, delivery, processing times, flexibility and costs if they
wish to be world Class. Son (1990) suggests using a ratio Of total output value
over the sum of productivity, quality and flexibility costs. The author posits that
this is a much more useful measure than just straight productivity because it
encompasses a number Of inputs rather than just direct labor. Fry (1992)
suggests that actual physical measures such as number of units Of inventory be
used instead Of standardized dollar measures so that managers can readily
interpret what is happening on the shop floor. Each author suggests a
seemingly different approach, but a common theme runs throughout these
studies; different types Of AMT performance measures are needed to better

reflect the nature Of the new technologies.

32

Measurement of people has to change as well. Individual incentives for
quantities produced are impossible to determine when all employees are treated
as indirect labor. Individual incentives will prove dysfunctional if employees are
expected to work together in teams to keep a system running (Gerwin 1982,
Saraph and Sebastion 1992). Instead, it is suggested that employee
performance be judged on a group basis to reflect the fact that workers are now
team oriented
2. 2. 2.6 Humem resource issues

The way employees are measured is only one of a number Of human
resource issues that have been linked with successful adoption of AMT. A
number Of authors have suggested that a commitment strategy (Walton 1985,
Walton And Sussman 1987) is necessary for successful adoption of FMS.
Although there is no single set of policies (Wood and Albanese 1995, Osterrnan
1994) that define a commitment or the strategic human resources model (Tichey,
Fombrun, and Devanna 1982) , the common themes are well trained workers
who;

. work in groups that are at least semi-autonomous if not fully self

managing

are Often paid salary rather than hourly

may be paid for the skills they possess in order to encourage learning
that should lead tO increased flexibility

a are selected more for cognitive and perceptual abilities than specific

job skills in an attempt to select employees who will be able to learn
and change as the process and products change

0. and are Offered long term security in order to instill a sense of identity

with the firm rather than their jOb - a worker is not just a machinist, they

are an integral part Of the companies long term ability to survive and
prosper.

33

(Gerwin 1982, Jaikumar 1987, Walton and Sussman 1987, Majchrzak 1988,
Adler 1988, Gupta 1989, Gupta and Yakimchuk 1989, Tranfield et. al. 1991,
Arthur 1992, Boer and Krabbendam 1992, Saraph and Sebastian 1992, Gupta
et. al. 1993, Arthur 1994, Pricket 1994, Goldhar and Lei 1994).

Analysis of the strategies used with AMT is only one Of three ways that
related human issues have been studied. A large bOdy Of literature in the fields
Of Industrial Relations and Sociology has investigated how the introduction Of
new technologies has impacted worker skills. This work has been largely
inconclusive (for reviews see Spenner 1983, Spenner 1988, Smith 1992, Form
et. al. 1994). A large number Of studies point tO AMTs lowering worker skills,
and an equally large number of studies show an increase in worker skills, or nO
Change in skills. A final area Of research has begun to examine the specific
policies companies use with AMTs (Dean and Snell 1991, Snell and Dean 1992,
1994, Wood and Albanese 1995, Osterman 1994, Arthur 1992) as well as
examining the performance Of specific policies or sets of policies (Arthur 1994).
This last area Of research builds Off the papers Of Walton (1986), Walton and
Sussman (1987), Gerwin (1982), and many others who have proposed a high
skill (or upgraded workforce) for the successful Operation Of AMT.

2. 2. 2. 7 AMT success: conclusions

The large number Of factors that have been found to impact the success
Of AMT installations can be divided into technical and managerial factors.
Although technical factors pose problems they can be overcome. However,

many authors have concluded that managerial issues are much more difficult to

34

solve and have a large impact not only on installing the technology, but success
Of the technology over time. A key managerial issue is the type Of human
resource strategy used to manage the Operators Of the technology.

Many Of the studies that deal with what human resource strategies are
used with AMT seem to indicate that successful users Of the technology are
using commitment strategies. Yet there are numerous studies in a variety Of
fields (to be discussed in section 2.3.2) that indicate that firms use AMT with a
large variety of worker Skills, rather than just the high skills generally linked tO
commitment strategies.

These inconclusive results lead to the following question: are firms using
low skilled workers with AMT because they are naive, or are there sound
business reasons tO follow a workforce strategy that runs counter tO the
prevailing wisdom ? The next section explores the apparent paradox (Poole and
Van De Ven 1989) that exists when one body Of the literature suggests that a
highly skilled workforce is a prerequisite for success, while another suggests that
many firms ignore this approach. The term paradox is used in the rhetorical
sense tO present two Opposing theses both of which are accepted in their
respective fields but are contradictory when examined simultaneously (Poole and
Van De Ven 1989). Using this paradox as a starting point, the relationship
between the skill level Of the workforce and adoption success will be explored

using the strategic choice framework.

35

2.3 The paradox -prevailing wisdom verses the evidence
2. 3. 1 The prevafling wisdom:

Researchers in different fields Of management study take divergent views
of the interrelation between technology and people. Many Operations
management researchers (Melnyk and Narasimhan 1992, Robinson 1991,
Saraph and Sebastion 1992) assume that attributes Of the technology dictate the
types of people policies needed for successful Operation. These authors all
conclude that AMT requires Operators who are cross trained, highly flexible, and
generally highly skilled.

The above authors assume that the technology dictates the skills and
resulting human resource policies required by AMT. However, this view Of the
workforce mirrors many authors writing about workforce policies regardless Of
the technology. Walton (1985) introduces the concept Of commitment as a new
paradigm to replace control type strategies.

These convergent views Of the proper skill levels Of the workforce are
echoed in a large number Of works on the human factors that lead to successful
AMT installations which are summarized in table 2-1.

GenNIn (1982) suggests that workers should be organized into teams and
given a broad range Of tasks in order to increase satisfaction. Workers
functioning under the “Old” control paradigm Often have narrow, low satisfaction
jobs which usually provide no motivation. Gerwin proposes that the additional
motivation Of these expanded jobs will lead to successful adoption. Gupta and

Yakimchuk (1988) note that management will be well served by including

 

36

Table 2-1 - Human Factors Leading to AMT Success

Factor

Explanation

Author(s)

 

Commitment Strategies /
High Skilled Workers

a human resource
strategy that makes full

use Of human capital is ‘

required for adoption
success

Walton and Sussman
1987, Adler 1988, Gupta
1989,Boerand
Krabbendam 1990, Boer,
Hill and Krabbendam
1992, Saraph and
Sebastion 1992, Arthur
1992, 1994

 

 

 

Specific Policies Explanation Author(s)
Associated with High
Skills
Teams Of cross trained Broader more enriched Gerwin 1982, Gupta,
workers jobs should lead to Chen and Rom 1993
greater satisfaction and
performance
Autonomy workers who can Gupta and Yakimchuk
program their own 1989
equipment will have a
better understanding Of
the shop floor
Selective Staffing Operators Of AMT are Snell and Dean 1992,

chosen based on ability
to learn and think as well
as to work with a team,
rather than just for
Operational skills

1994

 

Comprehensive Training

training in skills beyond
just Operation Of the
equipment is given to
foster teamwork as well
as learning

Snell and Dean 1992,
1994

 

Developmental
Performance
Appraisal(PA)

PA is used tO foster
learning and tO increase
skills, rather than as a
punishment tOOl

Snell and Dean
1992,1994

 

 

Externally Equitable
Rewards

 

Operational employees
are paid based on what
they would receive for
doing similar work at
other installations

 

Snell and Dean 1992,
1994

 

 

 

37

employees (usually in the guise Of a union) in the technological introduction
process as well as in decisions regarding Operations of the equipment. They
specifically note that while centralized programming of equipment may seem tO
be a method Of saving money, it takes an important Skill out Of the Operators
hands and in the long run will reduce commitment to the success Of the
installation.

Walton and Sussman (1987) build Off of Walton's (1985) work by linking
commitment strategies with successful adoption Of AMT. They note that control
strategies exist, but conclude that success demands a commitment strategy.
Adler (1988) also concludes that firms can follow a low skill workforce approach.
However, to do so is to follow an Older management paradigm based on control
he does not recommend for advanced technologies. Gupta (1989) also presents
both the upgrading and downgrading strategies and concludes that a high skilled
workarce is the proper choice for FMS.

Boer and Krabbendam (1990), Boer, Hill, and Krabbendam (1992) and
Saraph and Sebastian (1992) all suggest that a high skilled workforce is a
necessity for the successful adoption Of FMS. Saraph and Sebastion give a
single set of policies that they deem necessary for successful installation Of FMS.
Gupta, Chen and Rom (1993) found that team approaches and employee
commitment were keys to a successful adoption Of FMS.

Snell and Dean (1992, 1994) did a large survey study Of workplace
practices with AMT. They found that AMT was positively related to selective

staffing, comprehensive training, developmental performance appraisal and

38

externally equitable rewards for operational employees. Their findings suggest
that in general, computer automation in the workplace was related to specific
policies that point to a commitment strategy.

Arthur (1992) found that within a specific induStry (steel minimills) two
main human resource strategies existed: cost reduction (control) and
commitment. His findings suggest that companies who follow a cost leadership
business strategy will use a control strategy for the workforce, while firms
following a variety Of differentiation strategies will use commitment type
workforce policies. In a later work, (1994) he concludes that in general
commitment strategies lead tO higher performance than control strategies.
However, the issue Of business strategy as a confounding factor is ignored.
Additionally, many authors (De Meyer et. al. 1989 and Roth and Miller 1994)
imply that the typology of strategies suggested by Porter (1980) and used by
Arthur does not mirror the way firms are presently competing across cost, quality
and flexibility simultaneously.

Arthur’s conclusions may not be very strong but they do provide some
backing to Walton’s original assertion (1985) that commitment strategies are
necessary for success. Wood and Albanese(1995) provide additional support tO
the perceived importance Of commitment strategies with their findings Of a large
number Of firms following policies that could broadly be interpreted as
commitment. A question not addressed in this work is system performance. Are
firms following a commitment strategy because it will lead tO higher

performance? Or are firms adopting these policies due tO non-economic (and

39

potentially irrational or harmful) pressures such as bandwagon effects, fads,
fashions, a standard response to ambiguity, or a need for professional status
(Abrahamson 1991, Abrahamson and RosenkOpf 1993, DiMaggio and Powell
1983, Scott 1987)? 1

2.3.2 The prevaflng evidence:

Commitment strategies, high skilled workforces, and the like get a great
deal Of support from the literature. Authors like Walton, Adler, and Gupta
conclude that firms who wish tO be successful with AMT require a highly skilled
workforce. These conclusions, although supported by some findings, are
suspect for two reasons; 1) there is a large pool Of evidence that many firms use
AMT in a deskilled or downgraded environment, and 2) many Of the works that
provide support for the use of a commitment strategy in the workplace leave
many questions unanswered.
2.3.2.1 Desang evidence:

Braverman’s (1974) work on the impact Of technology on work (specifically
NC machines) has directed an entire generation Of Industrial Relations
researchers. Bravennan’s main premise was that management used technology
to control and degrade work. Jobs are simplified in order to wrestle control from
the workers. Braverman’s’ thesis is based heavily on Marxist theory and was
developed before flexible automation was even considered at most plants.
However, there are many researchers who have found evidence Of deskilled

workforces Operating CNC and FMS (see Table 2 - 2).

40

Keefe (1991) notes that the debate concerning the impact Of NC (and

CNC) equipment on worker Skills (started by Braverrnan) is ongoing and

unsettled in the Industrial Relations/ Sociology literature. Many authors have

found evidence Of deskilling, while an equal number have found evidence of skill

upgrading in conjunction with AMT introductions. As Spenner (1988) notes,

these studies are difficult to compare due tO differences in philosophy as well as

methodology. For example Braverman’s Marxist theology impacts the evidence

he does and does not consider (Attwell 1987).

The conclusion Of a large number Of authors who have reviewed the

literature on NC/CNC equipment and its impact on workforce skill levels has

been

Table 2-2 - Deskilling Evidence

 

Author Equipment Rationale
Hazlehurst, Bradbury and NC the installation Of NC equipment
Corlett 1969 leads tO higher and lower skill

levels depending on the plant

 

Braverman 1974

NC machine tools

management deskills the
workforce tO maintain control Of
production

 

 

 

 

 

 

Spenner1983, 1988 NC / CNC both skilling and deskilling occur
Shaiken, Herzenberg and FMS management lowers worker
Kuhn 1986 skills to maintain control Of the
shop floor
Kelly 1986, 1990 NC ICNC deskilling occurs in certain
sfiuafions
Thomas 1991 FMS Management lowers skills
because they view workers as a
problem to be solved
Adler 1991 FMS FMS is used with varying skill

 

 

levels but the tOp performer
uses fairly low skills

 

 

41

that the changes in skills brought about by technology vary (Spenner 1983,
1988, Form 1994). This mixed change, contingency, or strategic choice view
posits that companies use AMT with a variety Of different skill levels. This
conclusion echoes the early findings Of Hazlehurst, Bradbury, and Corlett (1969)
and is supported by recent work by authors such as Kelly (1986, 1990), and
Keefe (1991). For many authors the debate has moved from what is the impact
Of technology on worker skills tO what causes the impact of technology tO vary (
Form et. al. 1994).

The use of NC/CNC equipment with a low skilled workforce is not
consistent with the prevailing wisdom vis-a-vis the management Of AMT. But
generalizing from these studies to more integrated forms of AMT such as FMS
may not be possible. First, many Of the Industrial Relations/Sociology studies do
not worry about different levels Of automation and mix early adopters Of what is
now outdated NC equipment with CNC users. Secondly it is possible to install a
single piece of CNC equipment as a direct substitute for an existing non -CNC
piece of equipment while making minimal Changes tO the shop floor. However,
there is also evidence Of low skilled workers being used with more complicated
forms Of AMT (specifically FMS).

Shaiken, Herzenberg, and Kuhn (1986) did case studies Of a variety Of
AMTs, (including FMS) and found wide-spread deskilling. The FMS was nO
exception. The authors concluded that management at the FMS installation was
following the classic approach predicted by Braverman, in attempting to maintain

control of the shop floor by keeping jobs simple.

42

Thomas (1991) also studied a number Of installations and found that in
the FMS case management was following a deskilling strategy because they
viewed the shop floor workers as a problem to be solved. Not only didn’t they
want the workers to have control but they also did nOt trust supervisors.
Supervisory jobs depend on the size Of the workforce; the more employees there
are the more supervisors are needed. Supervisors would be likely to push for
policies that increased the size, if not the skills Of the workforce. In order tO
control the shop floor management not only deskilled the Operating employees,
they eliminated supervisors control Of jOb design (the reasoning was that
supervisors were likely to pursue designs that required more workers).

Adler (199) studied three FMS installations: one with what he deemed a
traditional structure, one with a neO-traditional structure (somewhere between
control and commitment but much closer to control) and one with a full blown
commitment strategy. Not only did two Of the three installations studied follow
“the Old paradigm” but the most successful company was the neo-traditional
company.

These studies, along with others (such as Blumberg and GeMin 1984)
are far from conclusive but they do show that not all firms are using a high skilled
approach. Additionally, all Of these works are based on in-depth case studies.
Unlike many other studies Of workplace skills based on census information from
The Dictionary of Occupational Titties (Spenner 1980) or a mail survey (Kelly
1990) that generalizes across industries and technologies, case studies tell us

what specific companies are doing with the technology, rather than presenting

43

summary statistics. Finally, Adler’s study has the intriguing conclusion that a firm
that is still basically traditional (control oriented) is the top performer in the study.
This finding makes the assertion that commitment (high skills) is a prerequisite
for success highly suspect.

2. 3. 2.2 Unapswered guestions:

The studies in the previous section form a part Of the empirical evidence
of how AMT is used. There are also a number Of anomalies in some Of the
studies that point to an upgraded workforce. Although these authors have
generally concluded that a high skill workforce is necessary, there are areas they
have not explored that are summarized in Table 2-3

Jaikumar (1986) examines in detail the tasks performed by the operators
Of an FMS, all Of which (expect for actual machine loading and unloading)
require a fairly high level Of skill. But the article also notes that many Of the tasks
that the workers performed (such as test cutting and examining statistics on
performance) were not required although they added value. These non-required
tasks needed more skill than the required tasks such as machine tending. Non-
required tasks that may add value do not have tO be performed by the Operators
but could instead be performed by management. Additionally, as Zicklen 1987
notes, new or fairly new technologies may require more input due to learning
difficulties. The question that Jaikumar does not address is; would Operators
continue to do non-required tasks ad-infinitum, or only until the system was

debugged? A second area that is not addressed is whether doing such tasks

44

Table 2.3 - Unexplored Questions

Author
Jaikumar 1986

Unexplored question
Are the non-required tasks that
Operators perform going to be part Of
their jobs forever or only until the
equipment matures ?

 

Adler 1990

Top performing FMS has neo-
traditional jobs rather than a
commitment strategy

 

Jackson et. al. 1989

Few statistically significant differences
between mass production sites and
flexible production sites

 

Jackson et. al. 1989

Only 49 % Of the employees in flexible
production facilities were reported to
need a large amount of skill variety

 

Smith et. al. 1992

Even in successful installations there
are significant numbers Of companies
not following policies linked to higher
skills

 

Dean and Snell 1991, Snell and Dean
1992, 1994

Findings from early studies not
integrated into later work using same
database. Later studies find no linkage
between AMT and compensation
practices '

 

 

Arthur 1992, 1994

 

Ignores the third Of his small sample
that suggests that any link between
human resource strategy and success
is very tenuous at best

 

would add value in an environment where there was not a large amount Of shop

floor variability or a more mature technology ?

Adler’s 1991 work also raises questions about the notion of a high skilled

workforce being a prerequisite for FMS success. The most successful firrn’s

workforce policy was described as one that could reap the benefits of the

specialized formal job assignments characteristic Of traditional work groups yet

nurture motivation through some informal flexibility in job assignments and long

 

45

term promotion opportunities. This neO-traditional firm had far more in common
with the traditional, low skilled control strategy firm than the team based
commitment strategy firm.

Jackson et. al. (1989) compared the practices at organizations using
flexible production to organizations who used mass production. The study has
two interesting findings from the standpoint Of policies used with AMT. The first
is that generally there are few statistically significant differences between mass
production sites and flexible production sites. This seems to conflict with all Of
the studies that describe flexible production as a new paradigm that will need
new methods (see for example Jaikumar 1986). Secondly, the authors ignore an
interesting result when they do find a statistically significant difference between
the skill variety of mass production versus flexible production. The interesting
finding is that only 49% of the hourly (non-managerial) workers in flexible
production settings require a large amount Of skill variety. The fact that less than
half Of the employees using flexible production methods needed a large amount
Of skill variety is in sharp contrast with the works on commitment.

Smith et. al. 1992 raise similar questions. The authors study only
successful installations yet their results indicate that many Of the polices they
deem necessary for success (such as mutual adjustment rather than direct
SUpervision and in depth training) are not being followed by a sizable minority Of
firrns (approximately 15% for each policy). The way the data is presented makes

it difficult to come to any conclusions as tO what is actually happening among the

46

surveyed firms but it is clear that there are some anomalies among these
successful firms.

The work Of Dean and Snell (1991) and Snell and Dean (1992, 94) while
seemingly supportive Of a high skill workforce for AMT is far from conclusive
when studied as a whole. Dean and Snell (1991) conclude that integrated
manufacturing (which includes AMT, JIT, and TQM) is not related to job design
but rather to the “firm’s context” which determines job outcomes. This
conclusion is not integrated into the two later works which attempt to link specific
policies with the use Of AMT. Snell and Dean 1992 find that AMT is positively
related to the use of selective staffing, comprehensive training and externally
equitable rewards for operational employees. The study does nothing tO link the
polices with organizational success. This study also raises serious questions
about either the previous findings or the validity Of the more recent findings. The
third study (1994) makes it even harder to reach firm conclusions about the
authors’ body Of work. The authors begin with the premise that AMT requires a
compensation strategy that is appropriate for the high skilled flexible workers one
should find in these installations. Yet the authors find that there is nO
relationship between AMT and compensation practices. In order tO find such a
relationship the authors lOOk for a link between what they deem knowledge work
and the appropriate compensation system (one which emphasizes group based
incentives, salary and Oddly seniority based pay). The findings suggest that

when a firm uses knowledge workers they will use policies consistent with

47

knowledge work. Nevertheless, there is no strong evidence Of a link between
AMT and knowledge work.

Arthur (1992, 1994) also leaves some interesting items unexplained. His
first paper linked commitment Industrial Relations strategies with differentiation
business strategies. His second paper linked commitment strategies with
business success. Taken together, the conclusion would then seem to be that
low cost strategies are not a method of success. This conclusion is not as easy
to explain when one considers the distribution Of firms following each type Of
strategy. In the sample of 29 there is one firm who follows a commitment
Industrial Relations strategy with cost minimization, and 8 firms who follow a
control (cost reducing) Industrial Relations strategy with some form Of
differentiation. These firms, who are defying the logic proposed by the author
represent approximately one-third Of the sample, yet are not well explained and
they are used in the analysis done for the second paper.

2. 3.3 Concluding statements on the paradox

All Of these works are far from conclusive. When they are combined with
the studies Of CNC and FMS that indicate that all firms are not following a
commitment strategy (high skilled workforce) the question that arises is: are
these low skilled firms truly outliers or is there something systematic that can
explain why a firm would defy the dominant logic? The strategic choice schOOl
offers an explanation Of why firms would defy the dominant logic. Strategic
choice also provides a framework from which to build a model of the relationship

between skill level of the workforce and adoption success with CNC/FMS. The

48

following section looks at studies that have examined the factors that drive
managerial choices regarding the skill level Of the workforce.
2.4 Strategic choice.

Although many Of the works that deal with people polices for AMT either
ignore the notion Of choice, or dismiss the low skill Option as a poor one, choices
linked to the business environment have long been an important part of the
management literature. Burns and Stalker (1961) noted that a company’s
environment should impact the type Of organizational structure chosen. Later
authors such as Lawrence and Lorsch (1967), Duncan (1972) , Van De Ven and
Delbecq (1974) echoed and expanded on the themes Of Burns and Stalker.
Finally, Child (1972) gave the moniker “strategic choice” to the tenet Of an
organization making decisions based on their specific environment, rather than a
single right practice (as dictated by Walton 1985), or due to a belief in
technological determinism.

The strategic choice framework is a popular way for researchers tO look at
a variety Of problems across a number Of fields. The organizational structure
and strategy literature has long viewed choice as one Of the possible frameworks
from which to study the different outcomes that firms have in similar
environments (Bluedorn 1993, Bluedorn et. al. 1994, Hrebiniak and Joyce 1985).
Studies using the strategic choice framework have shown up in Operations
Management journals such as Management Science (i.e. Romanelli and

Tushman 1986).

49

Choice is also used in the Industrial Relations/Sociology literature by
authors trying to deal with the growing body of evidence regarding the impact Of
AMT on worker skills. There have been a number Of studies and reviews that
have concluded that the changes found in the AMT workplace are not a result Of
the technology, nor management’s need to control as theorized by Braverman
(1974). Rather, changes are a function Of choices based on management’s
perception Of factors in their environment.

The strategic choice framework used in this dissertation will be the choice
framework used in the fields of Sociology and Industrial Relations. This use Of
the strategic Choice framework is very similar tO the constrained choice
framework used by authors such as Bluedorn et. al. (1994) in the Structure /
Strategy literature. The use Of the framework in this context implies that
management will choose a skill level for their workforce based on factors both
internal and external to the firm. Adoption success will be related to how well
management has matched their choice tO the environment in which they
compete. The following section summarizes studies on the impact Of technology
on Operator skills, which concluded that management makes a conscious choice.
The ensuing section summarizes studies that explicitly test the notion Of strategic
choice in settings that have a bearing on this dissertation.

2.4.1 The strategic choice framework - its appropriateness for studying the
impact of technology on the workforce and the resulting skill levels.

The lack Of a clear pattern in the type Of human resource systems used in

conjunction with AMT has lead a number of researchers to conclude that

50

managerial choice should be the framework from which future studies Of the
impact Of technology on the workforce are carried out.

Spenner (1983, 1988) reviewed much Of the work concerning the impact
Of technology on workers skills, and concluded that there was no dominant
pattern Of Change. His pertinent conclusion was that varying methodologies as
much as anything else determined the outcome Of different studies. Aggregate
studies suggest little or no change in worker skills while case studies suggest a
much more volatile workplace. Spenner concludes that upgrading and
downgrading are not the only lens through which tO examine the relationship
between technology and worker skills, and adds a third category deemed mixed
Change. Spenner defines mix change as the impacts of technological change
are mixed and offsetting for a variety Of reasons that include different policies in
various stages Of a technology’s life cycle, managerial Choice and other factors in
the environment.

Kochan, McKersie, and Cappelli (1984) make a much stronger argument
for the use of the strategic choice framework in a paper that deems choice a
useful framework for future studies of industrial relations. They hypothesize that
strategic choice is a framework that recognizes the role played by management
in shaping workplace practices such as the skill level Of the workforce. The
conclusion that strategic choice is a useful framework is based on a number Of
changes in the industrial relations environment that cannot be explained by

existing rule based paradigms (see Dunlapp 1958).

51

Child (1985) looks at the evidence on the interaction Of technology and
skill levels and concludes that there is no prevailing pattern in the way
companies integrate technology and worker skill levels. He then states that the
inherent flexibility Of many Of the new technologies inCreases (rather than
decreases) managerial discretion in the choice Of workplace practices. From this
conclusion he hypothesizes four generic workplace strategies that management
can chose from: elimination of direct labor, contracting, polyvalence, and
degradation Of skills. Blacker and Brown (1987) propose a similar typology but
restrict management choices tO two strategies that are similar tO control and
commitment strategies.

Burnes (1989) takes a different path to arrive at similar conclusions. He
starts with the premise that no single theory (such as contingency theory) can
adequately explain the various findings in the literature regarding the use of
advanced technologies. He then goes on to explore specific factors that might
impact the choice Of job designs and concludes that there are a large number Of
internal and external factors that influence the level Of skills possessed by the
workforce.

Form et. al. 1994 provide the most concise argument for a strategic
choice approach (which they deem a contingency approach):

Our contingency view is thus premised on a disparate set of ideas

and research findings that do not easily “fit" the pessimistic,

Optimistic, or mixed effects scenarios. While there is a superficial

resemblance to the mixed-effects argument, the contingency

approach demands specification of the factors that create divergent
workplace outcomes.

52

Rather than looking for deskilling (the pessimistic outcome) or high skilled
workforces (the Optimistic outcome) the authors believe that research should
progress to find the reasons for these outcomes.

The work Of all Of these authors point in an important direction: the
identification Of the impetus for workplace skill levels. Work Should be directed at
finding out what drives managers to chose specific skill levels. The following two
sub-sections review studies that have either found evidence Of managerial
choice with the skill level Of Operators in AMT installations or explicitly tested
managerial choice.

2. 4.2 Mixed change studies: authors who have concluded choice but not
explicitly used this framework

A numbers Of authors have gone looking for evidence Of skilling or
deskilling in AMT installations and found instead that management can choose
outcomes. These studies are summarized in Table 2-4 Perhaps the earliest
study to deal with AMT was Hazelhurst et. al. (1969) who studied the impact Of
NC machines on worker skills. Their findings indicate that NC machines may
change the skill mix needed by the worker. The type of skills required were a
function of the way the machines were used and the environment they were
used in. Skill level was not related tO management’s need to control, nor
anything inherent in the technology.

Buchanan and Boddy (1983) analyzed a number Of case studies and

concluded that the decisions made about jobs, work organization and structure

53

Table 2-4 - Authors who have Concluded Choice

_mn- Technology muss:—

Hazelhurst et. al. 1969 NC machines NC machines may change the
skill mix but there is nO pattern
Of skilling or deskilling

 

Buchanan and Boddy Various AMT management orientation
1983 determines skill levels rather
than technologgy

 

Lund and Hansen 1983 Various AMT skill levels are a result Of the
elements management
combines to make jobs

 

Kelly 1986 NC machines management makes a decision
about what tasks workers will
perform

 

Davies 1986 Various AMT organizations chose what their
outcomes will be

 

Keefe 1991 NC machines no net change in skills over the
30 plus years this equipment

 

 

has existed
Milkman and Pullman Auto Plant In the same plant some
1991 employees experienced skilling

while others experienced
deskilling - the differences
where due to the complexity Of
products and processes

 

 

 

were not determined by technology but by management orientations such as the
goals, assumptions and values Of those who make these decisions.

Lund and Hansen (1983) conclude that advanced technology affects the
skill levels of jobs but that the changes are a result Of what elements are
combined to make up jObs. They postulate that advanced technology will lead tO
a bi-modal distribution with a large number of low skill jobs, as well as a large
number Of high skilled jobs. However they are quick to point out that these

impacts are not driven by the technology and can be altered by management.

 

54

Kelly (1986) reviewed 11 studies that dealt with the introduction Of NC
machines in locations throughout the world. The study revealed mixed results:
some firms upgraded skills while others downgraded them. The author
concluded that the way the workforce was used was not an end result Of the
technology but rather, management discretion about what tasks workers would
perform, and what tasks workers would not control. Davies (1986) reaches a
similar conclusion noting that organizations chose what their outcomes will be.

Keefe returned tO the question of NC machines and worker skills first
addressed by Hazelhertz et al. in 1969. His study Of the impact Of thirty years Of
NC machining concludes that the net effect on workplace skills is minimal, with
little or no change occurring. The author challenges both the upgrading and
downgrading schools Of thought, concluding instead that a type Of skill leveling is
occurring. This phenomena is characterized by the disappearance Of both very
high and very low skill jobs, leaving a fairly homogeneous group in the middle.

Finally Milkman and Pullman (1991) found evidence Of skills changing in
both directions in the same plant at the same time. Their findings suggest that
product complexity has a major impact on changes in skill requirements Of
operational employees. In the auto plant studied, production had shifted from a
fairly complex Cadillac (luxury car with many features) to a simpler Chevrolet
(fewer features, fewer high technology inputs). As a result, workers needed
fewer skills not because of the robots installed but because there was less
variance in their tasks. At the same time the skilled crafts had increased their

Skill levels because they needed to deal with new and more complex equipment.

55

This finding fits very well with the original conclusions Of Burns and Stalker
(1961) that the way a company is structured should be in response to their
environment. The production employees needed low skill jobs to deal with their
environment while the skilled trades needed higher skilled jobs to deal with
theirs.

All of these studies on the impacts of technology on worker skills did not
start with the strategic choice framework. Instead, these authors concluded that
management was making a conscious Choice based on the findings that a
variety Of skill levels were being used by users of AMT.

This growing body Of evidence does not suggest a single right method Of
managing people, technological determinism or that management is driven by
the desire to control the workforce. Rather, the evidence has lead to studies that
either explicitly test the notion of choice or suggests choice is the framework
from which future studies Of the impact Of technology on skill levels should be
built. The following two sections will first explore studies that have used the
choice framework explicitly in the Industrial Relations / Sociology literature.
Then, the growing body of work suggesting that choice is the proper framework
from which to study the impact Of technology on the workforce as well as
workforce skill levels and human resource policies in general will be reviewed.
This framework will then be used to build a model Of what drives the choice Of

human strategy (skill level Of the workforce).

56

2. 4.3 Explicit tefiof the strategic choice framework:

Conclusions that management chooses outcomes have been appearing in
the literature on AMTs from some Of the earliest studies on the impact Of
technology on worker skills (Hazelhurst et. al. 1969). But there have been
relatively few studies that have used the choice framework explicitly. The
following section (as well as Table 2-5) summarizes these studies as they relate
to the issues of the skill level Of Operators that management chooses for AMTs.

Osterman (1987) builds a model Of what drives a firm's choice Of
employment systems. He concludes that firms chose their organization Of work;
hence the skill level Of Operators in manufacturing settings based on physical
technology, social technology, the characteristics Of the labor force and
government policies.

Zicklen (1987) looks at the impact Of NC (and CNC) machines on worker
skills from a unique perspective, the standpoint Of the employees. The study’s
stated purpose was not tO determine if skill levels had increased or decreased.
Rather, it was an exploratory analysis Of the factors that impact the relationship
between technology and skill levels Of Operators. The author concludes that it is
management’s approach to NC equipment that impacts the way it is used.
Management chooses the skill level of the workforce based on the nature Of
production (batch size), the extent to which the technology is perfected and the
size of the shOp.

Schluer (1989) builds on the theory Of strategic choice to examine the

drivers Of human resource or industrial relations decisions. The author

57

Table 2-5 - Explicit tests of Choice

Variable

Explanation

Author(s)

 

Physical Technology /
Nature of Production

Complexity of equipment,
amount Of change on
shop floor (batch size,
cyclicality, number Of
products)

Osterman 1987, Zicklen
1987, Birecree 1993

 

Labor Force

Size, Union(s), nature of

Osterman 1987, Kelly

 

 

Characteristics relationship between 1990, Smith 1992,
union and management, Birecree 1993
labor market
Maturity Of technology Newer technologies Zicklen 1987
require higher skills than
debugged nature
technologies
Size Of Shop Larger firms tend tO have Zicklen 1987, Kelly 1990

more specialization

 

Business Strategy

Low cost producers use
low skills to control costs
while differentiators need
higher skills to deal with
more variety

Schluer 1989, Arthur
1992

 

 

 

Point in PLC Introduction and Growth Schluer1989
Products require more '
skill than mature
products
Managerial Philosophy Managers who wish to Muller 1994

 

control the workforce will
use lower skilled workers

 

 

concludes that the firm’s strategy and the point in the product life cycle impact

human resource decisions. For volatile (growth) products a more flexible

workforce is needed. Companies who compete by differentiating through

innovation also need more flexible workers. Low cost producers and/or firms

with mature or declining products will be more concerned with containing costs.

Firms with mature products will not need flexible workers who can solve

problems because the products are generally debugged by this point.

 

58

Kelly (1990) returns to the question of the impact Of new process
technology on the workforce. But rather than looking for skilling or deskilling she
explicitly looks for the factors influencing decisions on the skill level of the
workforce. She finds that small plants that are non-uniOn are most likely tO follow
a skill upgrading approach while large union firms (especially those with a
seniority system) are likely to use a lower skill approach.

Smith (1992) also begins with the concept Of choice when looking at the
impact of technological change. He concludes that technology does not impact
worker skills but rather, outcomes are due to managerial Choice and negotiation
(on the part Of the union).

Arthur (1992) develops the same hypothesis as Schuler, namely that
strategy impacts the choice of industrial relations strategy. He concludes that in
general firms following a cost leadership strategy tend to use control (cost
reduction) industrial relations systems, while firms following differentiation
strategies tend to follow commitment type industrial relations studies.

Birecree (1993) does an in-depth case study of the changes in
International Paper Co.’s bargaining strategy with their union after the
introduction Of new process technology. Before the new technology was
installed the company was very accommodating to the union due to a fear Of lost
production during strikes. But the new technology allowed a small crew Of
managers to run the machines during a strike. Not only could management run
the machines but training new workers took less time. Finally, there was a new

company that had pre-trained strike breakers available on a temporary basis.

59

Three changes: new technology, shorter training periods and the availability Of
strike breakers were the forces behind management’s change to a more
aggressive l adversarial relationship with the union.

Muller (1994) did a pair of case studies Of two cOmpanies that had very
similar external environments as well as technology. Operator teams at Ford Of
Europe were compared to operator teams at General Motors Of Europe.
Because the products produced by both companies competed in the same
market and were built using very similar technologies, the differences in structure
(Of two successful business units) were hypothesized to be explained by other
factors. The author’s conclusion was that the different types Of union
relationships at the two companies drove different policies. Ford’s managers
were concerned abut maintaining control and gave the teams little autonomy
over a small range Of items. General Motor’s teams were far more autonomous
over a larger range of tasks.

These works are far from conclusive and some of them leave many
questions unanswered. Arthur (1992) does not explain the nearly one third of his
small sample that does not fit his typology. Schluer (1989) uses Porter’s
typology of strategy which seems in conflict with later works (DeMeyer et. al.
1989). However, these works show a revival in thinking about workplace
practices as the result Of strategic choice rather than the technology (Melnyk and
Narasimhan 1992), managerial desire to control the workforce (Braverman
1974), or a single right way to manage people (Walton 1985). These studies

were conducted using the strategic choice framework because Of the large and

60

still growing, body Of work which concludes that choice is a useful framework for
future studies Of the impact Of technology on the workforce and the resulting skill
levels Of the Operators. The following section reviews the studies that examine
the appropriateness Of a strategic choice framework to Study the linkage
between technology and worker skills.

2. 4.4 Strategipchoice conclusi_op_s

The concept Of choice has been present in the literature on the impacts Of
AMT on worker skills since some Of the earliest works on the subject. Yet it has
taken many years for the concept Of choice to be a viable way Of studying this
phenomena. Labor and Industrial Relations researchers, as well as their
counterparts in the field Of Sociology have started to move from looking for
evidence Of skilling or deskilling to addressing the causes Of managerial
decisions. This shift in thinking is due to the large body Of evidence that shows
that firms use AMT with a wide variety of worker skills. The following section
builds on the studies that have suggested choice Of worker skills, and addresses
the various drivers of choice in general.

2.5 Factors that explain strategic choices:

The main premise of this dissertation is that successful adopters Of AMT
use a workforce Of the skill level appropriate for their environment. This premise
implies that management makes a decision based not on a set Of best practices
(such as Walton’s commitment strategy) but rather on factors in their
environment that they perceive as having a large impact on the way the firm

Operates. This view may seem intuitive, yet works by authors such as Saraph

61

and Sebastion (1992) imply that there is only one way to manage the workforce.
This single outcome view was also prevalent in the literature that sought
evidence Of skilling or deskilling. Braverman’s (1974) premise that management
uses technology to lower skills in an attempt tO control the workforce drove much
research in the Industrial Relations field. But Braverman’s theory has not held
up to empirical study (for an in depth discussion of the faults with Braverman’s
thesis see Attwell 1987). The Opposing view that management should use a
commitment strategy to get the most from their employees has also not held up
tO analysis. This led a number of researchers tO start looking for the
determinants Of management’s choice Of worker skill levels.

There is also a large body Of literature in a number of disciplines that
examines the drivers Of managerial decisions. The relevant work that looks at
decisions other than the Choice Of worker skill levels will be reviewed in this
section. 1

This section integrates the previously discussed studies (sections 2.4.1
and 2.4.2) Of strategic choice in a Labor and Industrial Relations context as well
as studies in a wide variety of fields such as Organizational Strategy and
Structure that have yet to be discussed. Appendix B summarizes the findings Of
the studies that either implicitly or explicitly look at what drives managerial
choices. The forces behind managerial choices can be broken into two broad
categories to make discussion easier: factors internal tO the firm and factors
external to the firm (Burnes 1989, Kimberly and Rottman 1987, Venkatraman

and Prescott 1990, Jarvenpaa and Ives 1993, Mueller 1994, WOOd and Albanese

62

1995). The internal factors can be further sub-divided into: general business
factors, product/process complexity and industrial relations.
2. 5.1 Extemi factors that influence choice

The external variables can be broadly defined as the complexity Of the
environment the firm competes in. The skills available in the labor market may
have an impact on management’s decision as tO worker skills (Buchanan and
Boddy 1983, Davies 1986, Bessant 1989, Flynn et. al. 1995 ), because labor
markets that do not have skilled workers will require management tO invest a
significant amount Of money in training. Additionally, tight labor markets may put
a very high premium on high skilled labor.

Labor markets are only part of a finn’s external environment. Factors
related to the firrn’s industry make up an important factor. The amount of Change
(both the level and the predictability) have been cited as very important
deterrnents of a firm’s structure. Many authors (Burns and Stalker 1961,
Lawrence and Lorsch 1967, Child 1972, Bluedron et. al. 1994) have noted that
complex firm environments lead to more organic structures. Organic structures
tend to be associated with more flexible and more highly skilled workers than
mechanistic structures.

2. 5.2 IntemaL factors that influence choice

A number Of authors have identified factors internal to a firm which may

impact the level Of skills chosen for the workforce. These factors can be broken

into three broad categories:

63

a general business factors

. product process factors

0 industrial relations factors.
2. 5. 2. 1 Generai business factors

General business factors include things such as'firrn Size, firm
performance and the time since the last technological change. In general, these
factors have not been well linked to skill level decisions and in some cases the
evidence is contradictory. Kelly (1990) concludes that large firms are more likely
to follow deskilling. Dean and Snell (1991) find that large firms are more likely to
use policies consistent with a commitment strategy. Dean and Snell note that
this may be because larger firms have more resources tO invest. But the same
authors hypothesize that larger firms might be less likely to follow advanced
policies because change is difficult in large organizations. The size Of a firm has
been an important contingency variable in the organization structure/strategy
literature (i.e. Bluedorn 1993). The lack Of a strong conclusion as to how size
(and perhaps age Of firm (Kimberly and Rottman 1987)) impacts a specific sub-
group within a firm makes it a variable that should be controlled for in the study.

Firm performance is also poorly linked to the skill level Of the workforce.
Dean and Snell (1991) hypothesize that firms who are doing well will not change
because they do not feel the pressure to change. But their findings suggest that
firms who are performing well have the resources to change to more “advanced”
(commitment) policies. A linkage between firm performance and Operator skills

for AMT seems tenuous at best. Firm performance is hard tO link tO individual

actions and policies. Additional firm performance Often reflects what has

64

happened in the past, not what will occur in the future. Finally, an assumption
that firm performance is a driver Of workforce skill level decisions seems to imply
that successful firms would follow one strategy and less successful firms would
follow another (a finding not supported in the literature (Adler 1990)).

The time between major changes in process technology is the one
general business factor that seems well linked to policies used by the firm.
Walsh (1991) and Schroeder, Congden, Gopinath (1995) found that the longer a
firm goes without process change the harder it is tO institute other types Of
change when technology evolves. This implies that firms who have established
social relations and/or work practices will attempt to continue them even if
technology is changed. This may mean that a firm will continue with a skill level
even if it is inappropriate for the environment they are now in.

2. 5.2.2 Product/process factors

Product/process factors have a much stronger link to the skill level Of the
workforce than the general business factors. _A number Of authors have noted
that the complexity of products and processes (or the level Of conversion
uncertainty) as well as the complexity Of tasks are major drivers of workforce skill
levels (i.e. Kelly 1990). The more things change and the more uncertainty in the
direction and timing Of those changes the more flexible the workforce needs tO
be with flexibility in the guise of higher skill levels. Product/process Change also
captures the general business factor of time since last change in the number and

timing Of process changes.

65

Some of the other product/process factors that have been identified do
not apply when dealing with a single technology or have been challenged in
other parts Of the literature. Kelly notes that NC share Of the workplace (percent
Of plant that is numerically controlled) impacts overall firm level skills but this
work will be aimed at the AMT installation level rather than the firm level. Zicklen
(1987) notes that the maturity Of a technology impacts the level Of skills used,
with more mature technologies requiring lower skills. Zicklin’s finding contradicts
the findings Of Kelly (1990) and Meredith (1987) both Of whom found that later
adopters of AMT seem to be using higher levels Of skills. These findings are not
only contradictory but they may be confounded by the number of firms who are
generally moving toward commitment strategies (i.e. Wood and Albanese 1995).
2. 5. 2.3 lndustLal relations factors

The final set Of factors fall under the category Of Industrial relations. Kelly
(1990), Chaykowski and Slotsve (1992) and Wood and Albaese (1995) all found
that unions can have an impact on the level Of workplace Skills because as
Gupta (1989) notes unions tend to be less flexible. Additionally Kochan,
McKersie, and Cappelli (1984) note that union members tend tO be Older on
average than non-union members Of the workforce. Older workers Often do not
like their workplace changed and are Often unwilling or unable tO learn the new
skills which may be associated with an upgraded installation. Additionally,
unions with seniority systems have an even greater impact on the level Of skills
(Kelly 1990). But as Osterman (1987) notes, seniority systems are not just the

province Of unions. Finally, a few authors have noted that the type Of relations,

 

 

66

rather than the existence of a union (Davies 1986, Thomas 1991, Smith 1992,
Mueller 1994), has a major impact on management’s decision as to the skill level
Of the workforce.

A large number Of factors have been proposed as impacting
management’s choice Of skill levels for Operators Of AMT. It is possible tO divide
the factors into an external factor Of environmental complexity and internal
factors Of general business, product process complexity and industrial relations.
The general business factors are not seen as having a large impact on the skill
levels Of AMT operators (although controls for size and age may be appropriate).
Product/process variables can generally be looked at as the level and variability
Of change in the production environment. Industrial relations also play a role in
the decision as to skill levels but the mere existence Of a union does not seem tO
be a strong predictor Of skill levels, unless one also examines the type Of
relations and whether or not a seniority system exists for the AMT.

These variables will be used to build part Of a model that predicts the
success Of AMT adoption based on the appropriateness Of the skill level Of the
workforce. The literature review, up till this point, has covered literature on what
drives managerial choice. The following section reviews the advantages and
disadvantages Of two generic Choices that firms can make based on the factors
already introduced.

2.6 Choosing a strategy: two generic strategies
Management is faced with two basic choices Of human resource

strategies for Operators Of AMT; 1) Commitment (Walton 1985) or upgrading;

 

67

increasing worker skills or maintaining a high level Of skills and 2) Control
(Walton 1985), cost reduction (Arthur 1992), or downgrading; decreasing worker
skills or maintaining a low skilled workforce. Much Of the literature presents
downgrading as a negative choice (Braverman 1974, Form et. al. 1994) and
upgrading as a positive Choice (Adler 1988, Jaikumar 1986, Form et. al. 1994).
However, both strategies have advantages and disadvantages associated with
them.
2. 6.1 Commitment strategies

Companies following an upgrading strategy are following a philosophy
based on the premise that " technology provides an Opportunity to free up and
enhance the work practices Of people so that the skill required tO master new
responsibilities as the firm repeatedly Changes its product lines can be
developed." (Gupta 1989). Many Of the arguments for upgrading are based on
theories of intrinsic motivation and making the most Of human capital. The
arguments for upgrading include (Shaiken Herzenberg & Kuhn 1986, Walton and
Sussman 1987, Osterman 1987 , Gupta 1989, Jackson and Wall 1991, Snell and
Dean 1992, Receive 1993):

. high skilled workers can recognize and correct production

problems which becomes increasingly important as the costs Of
errors increase

. potential for fewer managers means potentially lower overhead
costs
more task significance
skilled workers can improve the system
enriched jobs should provide intrinsic motivation

higher skilled workers should be more flexible; able to handle
change.

68

. employees who are involved in change processes tend to be
less resistant to the change

0 in situations Of high technological uncertainty operator control
(Of maintenance) leads tO a reduction Of downtime

Upgrading, or skilling is usually presented as the best way tO do things. In many
cases authors discussing HR practices seem to assume that this is the only
acceptable strategy (Adler 1988, Jaikumar 1986). However there are negative
consequences associated with the upgrading strategy including (Walton 1985,
Osterman 1987, Adler 1988, Gupta 1989, Gupta and Yakimchuk 1990)

increased payroll costs

dependence on scare human resources

limited control of workers

turnover costs

workers can take the training and then shop their skills in

another market or firm

0 management must make a large investment in time and money,
while also risking becoming obsolete

. some workers do not want to be skilled especially Older and

unionized workers.

2. 6.2 COntroI strategies or costs reduction strategies

The downgrading strategy can be seen as a strategy that is used tO
enhance managerial control: "...the purpose Of technology is tO remove the
constraints on managerial authority vested in worker skill and autonomy, work
rules and strong independent unions". (Gupta 1989 ). Downgrading assumes
that managers and supervisors will deal with all major problems. Keeping
responsibility and control in the hands Of management makes it easier tO monitor
and control employee performance. Additionally, employees performing simple
jobs need lower skills and can be paid less (Gupta 1989, Walton and Sussman

1987).

 

 

 

69

Downgrading is usually viewed negatively. However, there are some

advantages to following a downgrading strategy including (Walton and Sussman

1987, Adler 1988, Martin 1988, Gupta 1989, Hunter, Schmidt, and Judiesch

1990, Shakien, Herzenberg and Kuhn 1986):

low skill workers are easier and less costly tO train

low skill workers are easier tO find and to replace

workers are easier to control and monitor because tasks are

simple

centralized control Of programming machines

. programmer not the one doing work so there is nO incentive
tO program machine to run slower than necessary

. Operators doing their own programming have almost total
control Of theirjObs, hence production process

. reduces redundancy and avoids different Operators making
the same part differently - reduces variance

does not require high skills in labor market

low skill workers may be more receptive to the technology

because they do not see it as a threat tO their jobs

differences in operator performance are much lower in low skill

environments -less variation among workers

Arguments against downgrading include the following ( Gupta 1989, Walton and

Sussman 1987, Adler 1988, Braverman 1974):

low skill workers cannot or will not recognize production
problems

low skill jobs are boring and do not motivate

little understanding Of place in system - no task identity
potential for increased management costs

degradation Of work

2.7 Summary of literature review

Success or failure of AMT installations have been linked tO a number Of

factors. Managerial factors are usually seen as more troublesome and more

likely to harm adoption than technical factors. An important managerial issue is

70

the type Of skills that the workforce will require. Many authors posit that high
skilled workers are a necessity for adoption success. However there is a large
body Of literature on the impacts Of AMT on Operator skill levels that shows that
there is nO prevailing skill level choice.

Not only is there no prevailing choice in the skill levels used by firms,
there is no overwhelming evidence that a high skilled workforce will actually lead
to better performance. The authors who propose that a commitment strategy is
a key tO adoption success seem to be ignoring a large body Of work that shows
that firms use a variety Of skill levels in successful AMT installations. On the
other hand, a majority Of the works that have looked at worker skill levels
explicitly have not linked skill choices with adoption success.

This work will build on the existing literature to build a model of adoption
success that is based on the notion Of choice Of worker skills, which explicitly

considers the impact Of these choices on adoption success.

Chapter 3
PRE-RESEARCH AND RESEARCH DESIGN

The goal Of this dissertation is to identify the factors that influence
managerial choices regarding the workforce for AMT that lead to installation
success. Factors were chosen based on the literature as well as three
exploratory case studies. The first part of this chapter documents the three case
studies. Then the literature review and the case study findings are used tO
develop a model Of AMT success based on managerial choices regarding the
skill level Of the workforce. The second portion Of the chapter describes the
research design and methodology. 1
3.1 Pre-research

Pre-dissertation case studies served a number Of purposes. First, while
many Of the parts Of the proposed model have received extensive study, the
impact Of skill level choices on success has received almost no attention in the
literature. Pre-research gave the Opportunity tO lOOk at what drives choices and
how choices impacted success. Second, pre-research brought out the
importance Of several previously unreported factors. Finally, the pre-research

allowed for some early testing of constructs and measures.

71

72

3.1.1 The cas_.:e studies

Three case studies were conducted from May to July Of 1995, in order tO
explore the factors which influence managerial Choices Of workforce skill levels
as well as the effect Of choices on success. Although Some standard questions
were used, these initial cases were seen as much more useful for building theory
than for any type Of theory verification. The pre-research was based on the
method Of grounded theory development (Straus and Corin 1994, Glasser and
Straus 1967), controlled opportunism (Eisenhhardt 1989) and the interative type
Of qualitative research described by Miles and Huberman (1994). As data was
collected, constructs, variables and measures were refined and reformulated in
order to build a more realistic and robust model Of the impact Of choices made by
managers regarding the skill level Of the Operators Of AMT and adoption
success.

The three company cases were conducted in different industries with
varying levels of automation and computer control. The companies were chosen
in an attempt to maximize variance in settings and industries but still enable
some comparisons. There are two union firms (one large and one small) located
in the Mid-west. Two companies are mainly suppliers to other industrial
concerns, while the third makes end products. Finally the companies use a
variety Of processes. In order tO protect the anonymity Of the companies the
three installations will be refereed to as; Boxes, Cars, and Composites. Table 3-

1 summarizes the attributes Of the three firms.

73

3.1.1.1 Boxes

Boxes is a small family owned firm that makes custom heart shaped
candy boxes. The company has recently relocated from a unionized plant in a
North-eastern industrial city to a non-union site in a rural South-eastern state.
Although they do not use any type Of AMT at this time the case is included for
two reasons; 1) In discussions with management it was made clear that if and
when AMT is installed the workforce practices would not change, 2) some Of the
factors that drove the skill level decision at this plant seem to be generalizable to
an AMT environment.

Boxes seemingly makes an incredible array of products. They make heart
shaped boxes in a variety of sizes, with an almost endless variety of possible
finishes. The finishes range from a simple all red box to extremely intricate three
dimensional covers in multiple colors. This vast range of sizes and finishes
make it seem as if Boxes has a very complex product line, which is indeed the
case from a planning and material ordering standpoint. But from the standpoint
Of the line worker they have only one product, heart shaped boxes. All boxes go
through the same basic steps regardless of finish. NO matter how complex the
design all boxes are hand assembled on a simple worker paced assembly line
where each worker is responsible for a very small portion Of the work. In other
words, the company has a single product family that is fairly standard from the
operators” perspective.

The company’s pattern Of demand also influences the decisions they

make regarding operator practices. The candy business is cyclical, as well as

74

seasonal. Most sales Of candy in heart shaped boxes occur in the two weeks
proceeding Valentines Day in February with few if any sales the rest Of the year.
Additionally, candy sales fluctuate a great deal from year to year, especially for
the producers of exclusive candies (for whom a custom box is more likely).
These fluctuations make it difficult for the customers that Boxes supplies to plan
very far ahead. Most orders that are tO be delivered around Christmas (for
Valentines Day candy) are placed between April and September, with the
majority being placed after June.

Hiring does not begin until orders are received. The company must also
order cardboard, and a variety Of finish materials for each order. Once people
are hired, materials have arrived and designs are finalized, the firm is left with
little time to actually make the boxes. This is exasperated by the fact that all
orders arrive at about the same time of year. The nature Of the products (nearly
custom) also means that inventory cannot be used to smooth demand. The end
result is very uneven production volumes, with 2 shifts being the norm in the
summer and fall and almost no production in the winter and spring. In order to
achieve this type Of volume flexibility the company hires and fires a seasonal
workforce each year to meet their needs.

A final issue for the company is the union, or rather the lack of a union. A
major reason for the relocation of the plant, to the South-east was to avoid what
management deemed restrictive union practices. These practices caused
excessive costs in an industry where even the most custom of products are

bought with price as the highest priority. The move to a predominately non-union

75

part of the country was carried out to enable the use of varying workforce levels,
in order to maintain volume flexibility.
The management at Boxes has chosen a low skill workforce for three reasons:
. the use of a seasonal workforce
. the fear of production employees having power to impact costs
a a simple product mix.
Boxes does not need high skilled workers nor is it likely that high skill workers
would be willing to work in a seasonal situation. Additionally, the company
wishes tO retain control of the workforce, which is easier to accomplish when
workers do not have important knowledge of the production system.
3.1.1.2 Cars

Cars is one Of many plants in a multinational automobile company. The
plant that was studied is located in the Upper Mid-west and currently makes four
different body styles in both two and four door models. Although all four body
styles use the same assembly line, the area of interest was the robot welding
line, where three Of the four models’ bodies are welded. This area is interesting
for a number of reasons. First, the body of the fourth car model is being welded
by hand and then placed on the line so it is possible to compare the earlier
production process with the present one. Secondly, the robot welding line meets
the definition Of an FMS (a number of robots that are linked by an automated
material handling system). Additionally, the robot welding line consists Of a large
number of robots that are reprogrammed almost every year, that process very
high volumes Of cars (over 300,000 each year) and thus require high levels Of

maintenance. Finally, although the robot line requires far fewer people tO staff

76

then the non-automated line, it still has a large number Of operators whose jobs
can be studied.

The robot welding line was installed for a number Of reasons. First, it is
far less labor intensive than the process it replaced. The model that is welded by
hand requires many more employees, who set parts in jigs, perform welding
duties and then have to move the parts. The robot line only requires employees
to load parts and monitor the equipment (which is augmented by a Computer
Measuring Device). The movement, setting, and welding Of parts is done by the
robots. Not only does the robot line require fewer people it also produces higher
quality parts. The car bodies produced on the line have over 3000 welds so the
possibility for error using human labor is very high. The robot line is also more
flexible. All six bodies (3 models in both 2 and 4 doors) run on the same line.
This flexibility is somewhat deceiving because there are stations where two or
three robots are in the same station. The body is directed tO the proper robot
and skips over the robots that are for other body styles.

High volumes of very standardized parts pass through the equipment, but
changes and adjustment are possible. Each robot has a control panel near it to
allow for decentralized programming. The operators do not reprogram the
equipment nor are they responsible for maintenance. Both Of these chores are
the responsibility of skilled trade unions. However the skilled trades only enter
and debug programs. They are not responsible for the actual writing Of the

programs that control the robots.

77

The sheer number of robots makes an in-house maintenance staff a
necessity. The robot line runs at 20 cars per hour faster than the rest Of the
assembly process because it is often shut down. When it is not shut down the
line needs little human intervention. The workforce that operates the robot
welding line at Cars are low skilled. They load sheet metal parts onto the line
and may do some rough alignment. There are also a few jobs, such as minor
deburring, that require a human. In general all the operators on the line do
simpler (although much safer, quieter and cleaner) jobs than the employees on
the non automated line.

The skill level of the workforce can be attributed to the nature of
production and the company’s unions. Cars has a very simple high volume
product line (from the standpoint of the robot line there are only 6 products which
have many welds in common). The work Of the robot line is predictable and
stable, with most Of the design changes occurring during model year
changeovers. The company’s unions also limit the level Of employee decision
making. Separate electrical unions are responsible for maintenance and
program entering. Finally the company has a history Of poor labor which may be
driving management to limit electricians to program typing rather than writing.
3.1.1.300mpgitep

Composites is a small company (fully owned by a larger holding company)
located in the upper Mid-west that produces a large range of composites for the
aerospace, automobile, and prosthetics industries. The company admits to

being about the most expensive composite maker in the country ( if not the

78

world) which is Offset by their design capabilities. Composites is an engineer to
order (ETO) firm and uses a CNC lathe linked tO a CAD system, computer
mediated measuring device (CMM) and other non-computer controlled
equipment.

The choices Composites made regarding their workforce were driven by
the types of products they build and the way they use computer automation. The
products cover a large range Of volumes, sizes, shapes and materials. The
company’s most important product (by dollar volume) is test cells for jet engines.
They sell about 30 of these cells a year. The cells range in size from a diameter
of a few feet tO over 20 feet. The cells are available in both composite and
metal. Because of the low volumes, as well as the range of sizes and materials,
the process to make these cells cannot be automated. Instead, dies are cut on
the CNC lathe, and then the cell is hand formed (if composite). The work
requires extremely tight tolerances and highly Skilled workers, who have a great
deal of autonomy over their work.

The company’s other products also cover a diverse range of shapes and
sizes. They make automobile parts, prosthetic devices and a number Of
aerospace parts that are very complex (including parts that have a square
Opening at one end, a round opening at the other, and a smooth interior). Part
volumes and sizes vary tremendously, but even the higher volume products are
produced in fairly small numbers. This requires a great deal of flexibility from

both the processes and the workers.

79

The company remains flexible by having a high skilled workforce capable
Of doing many tasks, a large pool of general purpose equipment and a few
central pieces Of AMT. Products are designed on a CAD system that is directly
linked to the CNC lathe. Dies for the composite pieces Can be designed and
programmed all at once. The CMM system gives the company a tool to check
incredibly tight tolerances accurately, for even the most complex shapes
produced.

The workforce at composites is highly skilled (the president describes the
shop floor as more like a tool and die shop than anything else) yet they do not do
some tasks that are Often linked to a high skill workforce. The CNC lathe and the
CMM are complex pieces Of equipment and the company does not have the
expertise tO do anything beyond the simplest maintenance. The CMM “ is great
when it works”, but when it breaks, the company needs outside technical
support. The same holds true for the CNC lathe, which although not as finicky
as the CMM is still beyond the capabilities of the company to fix. Programming
of the equipment is another task that the operators do not perform due to the
direct link from the CAD system.

The company’s president views the introduction of the various AMT as the
difference between being competitive and being out of business. In the past few
years the company has made the transition from being mainly being a MTO firm
serving a single customer market, to being an ETO firm who serves a large
number Of different customer markets. This change has allowed the company to

survive and even prosper even as many companies tied to aerospace and other

80

defense related industries have suffered. However, this transition has not been
painless. The company is unionized and it had begun tO hire a younger, better
educated workforce before the downturn. Unfortunately the seniority system
dictated that some of the best and brightest employees Were also the last hired,
hence the first fired. Nevertheless, relations with the union are generally good
and the company has been able to work with the union (and the vast majority of
the union members) to introduce the new technologies.

3.1.1.4 Case study conclusions

The case studies provide useful insight into some Of the drivers Of skill
level decisions which are summarized in Table 3-1. The case studies also
provide insight into potential measures Of some of the constructs that have been
posited as important in the literature.

All three companies were making Choices based on factors in their .
specific environment. Boxes was driven to use a low skill workforce by the
product they made. The inability to use inventory to smooth demand led to the
company using a large seasonal workforce to achieve volume flexibility.
Additionally, their product which seems very complex is fairly simple from the
standpoint Of the operators. Cars has a low skilled, but permanent workforce, on
the robot assembly line that makes a few standard products. Finally,
Composites has a complex product line with many low volume parts that are
engineered to order.

Composites uses their high skilled workforce to deal with the variety Of

products made as well as the product’s complexity. Both Cars and Boxes have

Table 3-1 -Case Study Conclusions

_—-'E'£-

 

 

81

 

 

 

 

 

 

 

method of worker paced machine paced Job shop
production assembly line assembly line
technology stand alone Robots connected CNC lathe
equipment and by transfer line connected to CAD
conveyors (FMS) system and CMM
skill level Of very low low very high
workforce
company size/ small (<500)/ very large small / small
plant size small (>100,000) / large
(4000)
union status non-union and will union with a union with good
do anything to history of labor relations
stay that way strife that has
recently abated
product market OEM - large end products - mainly OEM -
fluctuations in cyclical - with hard to predict for
demand major shifts in any one product,
demand and for all
products
number of distinct 1 7 >50
product lines
batch size [ 500-50000 ? 1-50

 

 

 

 

simple product lines and low skill workers. These results provide preliminary

backing to the importance Of product/process complexity having an impact an the

skill level management chooses for the workforce.

Each company also has unique labor relations. Boxes moved to get away

from a restrictive union that they felt made them less competitive. Cars has a

long history of animosity that has recently been tempered by the threat of foreign

competition taking not just line jobs but managerial jobs as well. This new spirit

of working together has led to some collaboration but there is a long a bitter

 

82

history to overcome. Finally, Composites has convinced their union of the need
to remain flexible if the company is to survive. The various types Of relationships
have also influenced the skill level decisions. Boxes has chosen to remain non-
union and to keep all control Of production in the handsiof management. The
need for control works well with jobs that are very simple, and job holders who
are interchangeable and temporary. Composites has good relations which
allows for the large degree Of task flexibility they need to compete. Finally, Car’s
history of poor relations leaves electricians as program typers, rather than
writers.

Each company has chosen work force practices that are linked to how
they compete and the products they make. Additionally, each company has a
different philosophy regarding interfacing with unions. These findings seem in
line with findings in the literature that indicate that product / process complexity
as well as the existence of a union impact the choices management makes
regarding the skill level of the workforce.

The findings also point to some potential measurement problems. First is
the measure of worker skill. Many authors (i.e. Kelly 1990) have used the locus
Of programming control as a proxy for skill. It is argued that machine
programming captures both the complexity and autonomy dimensions Of skill
(Spenner 1988). But the case studies suggest that programming is only one Of
many tasks that a worker can perform. The employees at Composites did not
program the CNC equipment, yet they were the highest skilled workers present.

The employees responsible for programming at Cars were not the direct

83

operators Of the equipment but they did work along side them. A simple
measure of locus of programming control would imply that non-managerial
employees at Cars programmed the robots. But in reality the skilled trades only
enter programs but do not write them. In this case the locus of programming
measure not only fails to capture the full range of skills, it is misleading.
Therefore the measure of skill will have to be inclusive enough to cover a variety
Of skills rather than a single complex task.

The case studies also point out some important product/process
variables. The pattern of demand at Boxes was responsible for many of the
things that the company did in regards to the workforce. Additionally, Boxes
pointed to the difference between number Of products and number of product
lines. Boxes and Composites had similar numbers of total products, but Boxes
had a single product line (serving a single customer group) while-Composites
made a 'wide variety Of product families for a large number of different customer
groups. A measure of product complexity would have to be based not on total
number of products or Options but rather, on product families or number Of
distinct customer groups.

Finally, the cases pointed to the complexity of labor relations. A simple
dichotomous union variable does not capture the different types Of relationships
and attitudes the companies have with respect to unions.

Together the case studies suggest that a variety of factors may influence
managerial decision. They also emphasize the importance Of designing

measures that capture the complexity Of the workforce issues studied.

84

3.2 Proposed model of choice of worker skills and AMT success

The pre-dissertation case studies as well as the literature review were
used to build a model Of AMT success based on the choices that management
makes regarding the skill level of the workforce. The framework of strategic
Choice is used to address the paradox Of a large body of literature suggesting
that a high Skilled workforce is a prerequisite for adoption success while an
equally large body Of literature suggests that firms use AMT with a variety of skill
levels. Both the literature review and the case studies indicate that Choice may
be one way to address the paradox. This section builds a model Of adoption
success based on the choices management makes regarding the skill level Of
the workforce to address the paradox.

Figure 3-1 Shows a complete model Of the factors that impact the success
Of an AMT adoption. This model is far too large to tested simultaneously and
many of the success factors (such as a process champion) have already been
tested (Beatty 1993, Meredith 1987). Technical factors such as software
integration (Meredith 1987, Fry and Smith 1989, Gupta et. al. 1993) across
various components Of the system are important, but they have been examined
before. The solutions to many technical problems, while not simple, exist.
Finally, it is managerial issues in general that most authors conclude to be the
most problematic part Of AMT installations (GenIvin 1982, Meredith 1987, Adler
1988, Boer, Hill, and Krabbendam 1990, Gupta et. al. 1993, Bessant 1993, and

Elango and Meinhardt 1993).

85

Figure 3-1 - A complete model of AMT adoption success

 

Technical
Factors

Supplier Quality

Software

Tools

 

 

 

 

number

control
Information
Systems
maintenance

timing

 

 

 

 

Successful
Managerial Factors :IAdoption
I

 

 

 

 

 

 

Integration
across functions
with suppliers
with customers
High skill workforce
Technology strategy
Fit with business strategy
Process champion
Maintenance
who performs it
Industrial relations
Performance measurement
F MS
Operators
Human resource policies
selective staffing
comprehensive training
developmental appraisal
externally equitable
rewards

 

 

 

 

 

86

The management Of human resources is the most interesting and perhaps
the least understood Of all the management issues. The number of studies that
indicate that a highly skilled “knowledge workforce” is a necessity for AMT
success (i.e. Walton and Sussman 1987, or Saraph and Sebastion 1992) are in
Sharp contrast with the studies that Show that even in successful AMT
installations this pattern is not always followed (see Adler 1989, or Smith et. al.
1992). Figure 3-2 shows a general model that explains the success Of AMT
adoption occurs when management chooses a workforce with a skill level
matched to important factors in the environment. This is explained later in the
discussion of measures.

Figure 3-3 shows a detailed model Of the interaction proposed in figure 1-
2. This model proposes that firms that are successful adopters of AMT will
Choose workforce skill levels that are appropriate for an environment that is
broadly defined by three constructs: product! process change (internal issues),
environmental complexity (external issues) and Managerial Discretion The
constructs are justified and then explained below.

3. 2.1 Justification Of factors

The factors that are included in the model are not all Of the factors that
have been identified as driving managerial Choice. However, the factors that will
be examined were Chosen based on the literature review and the pre-dissertation
case studies. Both suggest that these three unique factors have the greatest

impact on managerial choices.

87

Figure 3-2 - Basic conceptual model

 

 

Business
Environment

 

 

 

 

 

 

 

 

 

Inappropriate
; Skill Level of
Workforce
Appropriate

 

 

Successful Adoption

 

Unsuccessful
Adoption

 

 

88

Figure 3-3 - Conceptual model of skills and AMT adoption success

 

Product/Process Change

floor

 

Level of uncertainty on the shop

 

 

Managerial Discretion
Degree of flexibility
management has with the
workforce

 

 

 

 

Unsuccessful
Adoption

Goals Of AMT not
achieved

 

 

 

Inappropriate
skill level

 

 

Environmental Complexity
Level Of uncertainty in markets
where products compete

 

 

 

 

 

Skill Level of Workforce
Total preparation time for

 

f

Operator jobs

 

 

 

Appropriate skill level

 

Successful Adoption:
Goals Of AMT achieved

 

89

The pre-dissertation case studies indicate that the level of complexity on
the shop floor and the type of industrial relations (Managerial Discretion) the firm
adopts affects worker skill levels. Boxes chose low skill workers because Of the
need for volume flexibility, the strategy to control the wOrkforce and the single
product line produced. Composites chose high skilled workers as a function Of
the complexity and variety Of their products as well as their relationship with the
union which allowed for flexibility Of worker tasks. Cars chose low skills because
Of the high volume, low variety nature of the product, as well as their experience
of poor labor relations.

The literature review supports the contention that product I process
Change and Managerial Discretion impact the choices of worker skills. The
literature review also indicates that a firm’s external environment affects the
choices made by management.

The general business variables, such as firm size and firm are not part Of
the model. Firm size has been hypothesized as having varying effects. Snell
and Dean (1992) note that large firms may have the resources to try “advanced”
programs, yet they may suffer from inertia and be difficult to change. The lack of
consensus regarding the effect of size makes it difficult to hypothesize how size
might effect choices as well as adoption success. This is complicated by the
issue of performance. A firm that is large and performing poorly has fewer
resources than an equally sized firm performing well. However, a large firm may

have greater access to resources than a poorly performing small firm.

90

The potential for interaction is further complicated the adoption success
variable being linked to performance. It may be difficult to make the jump from
adoption success to firm performance in large companies. However, small firms
installing an unsuccessful AMT may risk their survival. ”Perfonnance cannot be
both an independent variable and a dependent variable so it is not part of the
model. Size is also excluded from the model, although it will be controlled for by
correlating size and skill level as well as size and adoption success.

3. 2.2 Definition Of Factors
fig. 1 Chme / complexity

There are many dimensions of change or complexity believed to affect the
skill level decision. The specific variables mentioned in Appendix B that relate to
Change or complexity are: complexity of products, uncertainty of tasks and the
general level Of uncertainty in the environment. These variables can be further
divided into product / process change variables and environmental change
variables.

Such a division is warranted based on numerous findings from the
organizational structure ltheory literature. Lawrence and Lorsch (1967, 1986)
found that functional areas within the same company will deal with environments
Of varying complexity. This conclusion is buttressed by the finding that the
technical environment (manufacturing) is generally the least volatile in the firm,
hence production has to deal with the least change. Burns and Stalker (1961)
found that lower levels of change lead to the more mechanistic structure

generally associated with greater subdivision Of tasks and simpler jobs

91

Duncan (1972) notes that different parts of the firm deal with different
environmental factors and that there are two large sets of environmental factors,
internal and external( Jelinek 1977, Dess and Beard 1984, Kimberly and
Rottman 1987, and Jarvenpaa and Ives 1993). Not only are there internal and
external factors in the environment but a recent study by Wood and Albanese
(1995) indicates that variations in the use of policies linked to commitment
strategies (Often seen as a variation Of skill upgrading (Arthur 1992, 1994)) are
driven only by the internal environment, not the external environment.

This study proposes that measures Of change or uncertainty should be
split between internal and external environments of AMT for two reasons. The
first reason deals with the differences between the environment Of the AMT and
the entire firm. A firm may have a fairly turbulent environment, characterized by
a large and dynamic set of customers, competitors and government regulation,
yet may have a fairly stable production environment. The notion Of different
levels of change internal to the AMT as compared to the firm also makes intuitive
sense. Customers and regulation may change frequently but the shop floor
technology as a capital investment changes at a far slower pace. Secondly, the
external environment may not impact the internal environment.

A maker Of a mature commodity may have very different internal and
external environments. The high volume product that the firm sells does not
change much (if at all). However, the external environment may be full of other
firms making the same product, a number of different customer groups buying

the product purely on price, and possibly competition for raw materials. The

92

shop floor is very stable, yet the environments faced by marketing and
purchasing are very dynamic. The opposite can also occur. Many firms in
defense industries have a single customer (the US government), yet shop floors
of incredible complexity.

A final dimension Of change or complexity was defined by Duncan (1972)
who noted that environments can also be classified as static and dynamic.
Dynamism, or turbulence (Bourgeois 1980) is an important attribute of change.
The question must be expanded to include not only how much Change is
measured but also how Often and how easily it can be predicted. The installation
Of an AMT results in a great deal of change but happens infrequently in an
organization. Changes in product designs can occur every time a product is
made in an ETO environment, or almost never for a mature MTO commodity.
;2_.2.1.1 Pmduat/ process chapgg

Many authors (Kelly 1990, Zicklen 1987, Hazelhurst, Bradbury and Corlett
1969, Jackson and Wall 1989) have concluded that the level and predictability of
change on the production floor has a major impact on the skill level required by
the workforce. Dynamism resulting from smaller batches (Kelly 1990, Zicklen
1987), frequent new product introductions, high levels of customization, large
numbers of engineering changes and unpredictable shifts in demand and/or
product mix leads to a need for higher skills. Variety and predictability are
defined as important attributes of change by authors examining task complexity

or uncertainty (Hunter, Schmitd and Judiesch 1990, Dean and Snell 1991, and

Reshef 1993). The product/process change construct captures the amount Of

93

change on the shop floor (be it from process or product) as well as the
predictability Of that change.
3. 2. 2. 1.2 Environmenfirlficomplexity:

The external environment that a firm competes in has been linked to firm
structure (Burns and Stalker 1961) and resulting skill levels. In general the more
complex a given environment the more flexible a firm needs to be. Flexible
employees are able to do a broader range Of tasks (higher skills). Child (1985)
notes that the frequency of change, the degree of change and the predictability
of change are key determinants of the complexly Of the environment, hence the
structure Of the organization and design ijobs. This construct will capture the
amount of change in the external environment as well as the predictability of that
change.

3. 2.2.3 Managerial discretion

Braverman’s (1974) central argument was that management deskilled the
workplace to gain control Of the shop floor. This theme is echoed by Nobel
(1984) who notes that “Workers controlled the machines and through their
unions had real authority over the division of labor and job content”.
Management then uses simple jobs where the worker is easily replaced to avoid
the union having important knowledge about the control Of production.

If this premise is correct then the existence Of a union should in and Of
itself be a strong predictor of a low skilled workforce. There are some authors

who have found evidence of this ( Kelly 1990). Many more authors (to be

94

discussed below) have found mixed evidence as tO the effect Of unions, or more
factors than just the union that are needed before skill levels can be predicted.

Guest (1987) notes that there has been a reduction in trade union
pressure on management which should reduce the need for control. However he
also notes that some union practices such as specific job demarcations do not fit
with a flexible workforce ( a theme echoed by Osterman 1987). He notes that
when there are multiple unions at a site it may be difficult to increase the
flexibility of the workforce. For example in an FMS installation with separate
electrical and operating employees’ unions, Operators may not be allowed to
perform their own maintenance because they will be crossing job demarcations.

Gupta (1989) echoes the theme of trouble with multiple union sites, but
also cautions that being non-union is not a predictor of a strategic HRM focus
(usually associated with higher skill levels). Chaykowski and Slotsve also worry
that although a company may wish tO update work practices to reflect a
commitment model, the existence Of a union may make the changes
uneconomical.

Finally, Wood and Albanese (1995) found that the existence of a union
has no linkage to control or commitment policies. This recent study seems to
suggest that the existence of a union is not a good predictor of workforce skills.

One possible explanation is a seniority system. Seniority systems are a
method Of insuring that the highest skilled employees are the ones getting paid
the most. An important assumption is that the skills the workers have developed

over their tenure are related to the work they are presently doing. But AMT

95

installations Often require a new skill set, making senior employees a potential
liability, rather than an asset. Kelly (1990) found that seniority systems had a
strong link with lower worker skills. And the combination Of seniority systems
and a union made the likelihood of low skills even higher. Osterman (1987)
notes that seniority systems are not the domain Of just union firms but rather are
equally common in union and on-union settings. These findings point to an
industrial relations construct that is not as simple as the existence of a union(s)
and or a seniority system.

An additional problem may be that the union is related tO other factors that
impact the skill level of the workforce. Kochan, McKersie, and Cappelli (1984)
note that unions tend to be in older plants with far older workforces. Gupta
(1989) found that the age of the workforce was a predictor of skill levels with
older workers being less likely to want their skills upgraded. In a union
environment an older workforce with a good deal Of seniority might block
attempts to change tasks or expand the knowledge base needed for work. Older
plants also pose a problem because as Schroeder, Congden, and Gopinath
(1995) found, the longer a plant goes between technological changes the more
difficulty the plant has with the change. Although there is no direct link with old
plants and Old technology, an older plant with an Older workforce is potentially a
place where change is not wanted, and difficult to achieve. Combined with a
union environment the only way to install technology may be with jobs that

maintain the status quo, even if they are not appropriate.

96

The conclusion that the age Of a plant impacts work practices (in the guise
Of strategic human resource policies) is not consistent with the finding Of
Osterman (1994) who found that the age of an establishment had no impact on
the level of flexible work practices (his construct for strategic human resources).
Additionally, Osterman found that the existence Of a union and or seniority
systems did not impact the level Of flexible work practices.

A final issue in the debate about the impact Of a union is the type Of
relations that exist between a union and a firm. There is a growing body Of work
that looks not just at the existence Of a union but at the type Of
labor/management relations. As Guest noted unions are no longer threatening
to many managers, and so may not need to be as tightly controlled. Additionally,
Smith (1992) found that unions can have an impact on how technology is used
through the negotiation process. Thomas (1991) echoes this finding, noting that
the type of relationship between the union and management as well as the
political process impacted job design. In an FMS installation, operator jobs were
designed to remove operator control due to poor relations that convinced the
engineers that the Operators were a “ problem to be solved”. Horvitz (1994) adds
a cautionary note to the debate about the quality Of labor relations by Observing
that although there is evidence that unions and management can collaborate,
collaboration is Often driven by a crisis, with the collaboration ending when the
crisis ends.

Additionally, Jackson et. al. (1989) found that unions were related to many

practices such as bonuses for company performance that have traditionally been

97

associated with non-union environments. They took this finding as evidence of
unions and management developing win/win relationships. Finally, Arthur (1992)
notes that a union is not a predictor Of the type Of industrial relations system
management chooses.

These results are far from conclusive, but they indicate that a simple
dichotomous union variable is not appropriate. Rather, the hypothesized
interaction impact Managerial Discretion will have to be captured by a more
complete dichotomous variable.

This study will examine the impact that product/process change,
environmental complexity and Managerial Discretion have on the choices firms
make regarding the skill level of the operational workforce for AMT. The
literature review and the pre-dissertation case studies suggest that these three
factors are the best predictors of workforce skill levels.

3.3 Hypotheses

The literature review and pre-dissertation research were used tO build a
model of AMT adoption success based on the choices that management makes
in regard to the skill level Of the operators of the AMT. This study proposes to
bridge the gap between the large pool of research that indicates that a high
skilled workforce is a prerequisite for adoption success and the equally large
body Of research that shows that firms use AMT with a wide variety of operator
skill levels. The following Hypothesis were adressed in the research.

H1a: The skill level of the workforce is positively related to the success of
the adoption Of CNC and/or FMS.

98

H1b: Firms who choose appropriate levels of skill for the operational
employees Of FMS and CNC will be more successful adopters than firms
who choose inappropriate skill levels.

Hypotheses H1a and H1b are competing hypotheses. H1a tests the
proposition that high skilled workers are a prerequisite'for adoption success.

H1b tests the strategic choice framework.

Strategic choice posits that firms make Choices as tO structure, job design
and the resulting skill levels. This is in contrast to the high skilled workforce
suggested for all AMT installations by many authors (Walton and Sussman 1987,
Saraph and Sebastion 1992). The literature review provides some support for
choice. Plants have successfully adopted various forms of AMT with low skills
(i.e. Smith et. al. 1992). Plants where management chooses lower levels Of skills
(such as Adler’s 1989 Neotrad) should be just as successful with their adoptions
as a firm with high skilled workforces if the skill choice is appropriate.
Installations where the skill level is higher or lower than needed will not be
successful.

Blacker and Brown (1987) hypothesize that low skill workers or high skill
workers will be equally successful in low complexity environs, but they ignore
three key issues Of low complexity and high skills. The first is worker boredom.
High skilled workers who do not need to use most of their skills will probably not
be motivated and may actually perform worse than a motivated worker with lower
skills. Additionally, the costs Of a high skill workforce are not justified if these
skills are not used. Finally Karuppan and Schniederjans (1995) have shown that

a key source Of stress in automated plants is skill under-utilization. High skilled

99

workers doing work that does not require their skills will suffer a higher level Of
stress than low skilled workers in the same plant. In summary workers who are
not appropriate for the work they will do will either be unable to do the job
required, or overqualified, overpriced, and overly stresSed.

H2: The level and unpredictability of product/process change is positively

related to the skill level management chooses for the operators of CNC

and FMS.

The pre-dissertation case studies provide preliminary support for this
hypothesis. Boxes and Cars had simple shop floors and low levels of skills.
Composites had a complex shop floor environment and highly skilled workers.
The literature review also supports this hypothesis. Authors such as Kelly (1990)
and Milkman and Pullman (1991) note that the level Of change or uncertainty on
the shop floor is one Of the drivers of skill levels for Operating employees.

Higher skilled workers are needed to deal with the uncertainty created in
complicated internal environments (Child 1985). In simple or static environments
lower skilled workers will be adequate (Blacker and Brown 1987). Low skilled
jobs in complex environments will not be designed to deal with the variety of
tasks required. Jobs with high levels of preparation will be an unnecessary cost
in low complexity environments and may lead to stress due to skill
underutilization (Karuppan and Schniederjans 1995).

H3: The level and unpredictability of environmental uncertainty is

positively related to the skill level management chooses for the operators

Of CNC and FMS.

Based on the seminal work Of Burns and Stalker (1961), Lawrence and

Lorsch (1967), and Duncan (1972), as well as a number Of more recent works

100

such as Swamidass and Newell (1987), Kotha and Orne (1989), and Covin and
Slevin (1989); the more volatile and unpredictable the external environment that
the firm competes in the more flexible and skilled the workforce should be.

H4a: Managerial discretion does not moderate the relationship between
product/process change and skills.

H4b: Managerial discretion does not moderate the relationship between
environmental uncertainty and skills.

The pre-dissertation case studies showed a much more complex industrial
relations variable than a simple union / non-union dichotomy. Boxes was non-
union and had the lowest skills Of the group. Cars was union and had low skill
employees. Finally, Composites had the highest skilled workers and a union.
The distinguishing factor between Composites and Cars was the type Of union
relationships. Composites had generally good relations and was able to work
with the union, while Cars had a long history of animosity.

Although some studies found the existence of a union lowered overall
skills (Kelly 1990) the existence Of a union is not posited to be a driver Of skill
levels in and Of itself. Union management relationships have begun to change
(Thomas 1991, Jackson et. al. 1989) and studies such as Muller 1994 show that
unionized firms in the same industry and external environment may chose
different workforce approaches.

Union members are often older than non-union employees (Kochen,
McKersie, and Cappelli 1984) and Older employees are less likely to want tO
change (Gupta 1989). This unwillingness to change, when combined with a

seniority system for the AMT, will mean that the workers least likely to want to

101

change their habits will be working on the new technology. In firms where
relations are poor, nothing can be done about this labor assignment. However,
for firms with good relations with the union there will be an opportunity to use
workers who are willing to change / learn. The interactiOn Of Managerial
Discretion with the two Change variables is depicted in figures 3-4 and 3-5.

3.4 Detail of measures to be used

This study addresses the gap between the literature that states that a high
skill workforce is a necessity for AMT success and the literature that shows that
firms use workers with a wide variety of skills in AMT installations using the
theory Of strategic choice. It is posited that managerial Choice is driven by the
degree of product/process change, the firms external environment and the level
Of Managerial Discretion. Although these constructs have been examined in the
literature previously, not all Of the measures are well developed. (The method Of
data collection (discussed in detail in section 3.5.1) was chosen partially to
validate and if necessary refine the measures that are proposed in the following
sections.

The following sections define the measures of skill, product/process
change, environmental complexity, Managerial Discretion, appropriateness Of
choice, and installation success that will be used in this study. A complete
structured interview format can be found in Appendix C.

3. 4. 1 SkLII level
There are a number Of different ways to measure the skill level of the

workforce (Spenner 1983, Form 1987, Snell and Dean 1991, Hazelhurst et. al.

102

Figure 3-4 Interaction of managerial discretion and product process change

High

Skill Level of
Workforce

Medium

Low

 

    

High Managerial Discretion

   

Discretion

Low Managerial

 

Low

Medium

Product Process Change

High

103

Figure 3-5 Interaction of managerial discretion and environmental complexity

       
 

H' h
1g High Managerial Discretion
Skill Level of
Workforce
Medium Low .Managerial
Dlscretlon
Low

 

 

Low High

Environmental Complexity

104

1969, Braverman 1974). At one end of the spectrum are abstract concepts
such as Braverman’s craftsman Of the 19th century who Braverman compares
tO the shop floor worker of the 20th century. At the other end Of the spectrum
are Milkman and Pullman (1991) who measure multiple site specific tasks
performed before and after a technological change in a single auto factory.

Skills, as Spenner (1983) notes are multi-dimensional, with the two major
dimensions being substantive complexity and autonomy. Complexity is the core
of skills and may not vary greatly across people holding similar jobs in different
firms. However, autonomy varies not only from job to job but from company to
company.

Spenner’s assertion that complexity is the core of jOb skills is echoed by
Form (1987). Form notes that although the two dimensions (complexity and
autonomy) are different they tend to correlate highly (.5 - .7) Not only do they
correlate but in factor analysis reliable indicators of each factor appear in the
structure Of the other. Spenner (1980) found strong correlations between
complexity and supervision, task variety, control and repetition, making
complexity in and of itself a good measure Of the skills required by a job.

Although complicated measures of complexity exist, (i.e. Van de Ven,
Delbecq and Koening 1976 ) Form suggests that the total preparation time
required for a job is the best available measure of worker skills when doing work
that cannot use the dictionary of occupational titles (DOT). Total preparation

time includes general education, vocational education, and on the job training.

105

A measure Of preparation time is useful for a number Of reasons. First, it
captures the notion Of complexity well, in that the more time spent preparing for a
job the more likely the job is to require either a larger mix of tasks and or very
complex tasks. Secondly the measure avoids the pitfall of just counting the
number of tasks performed by an operator. Cross training may be a good
indicator Of flexibility but an operator who performs a large number Of simple
tasks may be lower skilled (and need less preparation) than an operator with a
single complex task. Third , the measure is more useful than simpler measures
Of educational attainment which may only be an indication Of credentialism
(Braverman 1974), rather than actual skills. Finally , while a fairly simple
measure to calculate, this measure is far more useful than some other simple
measure such as locus of programming control.

Kelly (1990) as well as Gupta and Yakimchuk (1988) and Shaiken,
Herzenberg, and Kuhn (1986) use the locus of programming control as a proxy
for skill level in AMT environs. While this measure seems to capture both
complexity (the more programming an operator does the more understanding Of
the process and products they need) and autonomy (operators who write their
own programs are essentially controlling their own equipment) it is flawed.
Companies with CAD [CAM systems may have direct links from design to
machine programming. This link would imply low skills because the Operator is
not doing the programming. But as early site visits showed, workers (see

composites) can have a great deal Of skill without doing their own programming.

106

The measure Of skill level was a continuos variable Operationalized as the
total preparation time needed for an average employee to perform the job at
average / acceptable levels Of performance (from Form 1987). This measure is
not without flaws, the largest being individual differences. The measure attempts
to differentiate between differences in people by looking specifically at average
holders of the job. For example Hamper (1992) describes training for an auto-
assembly line. New job entrants are given three days to learn their jobs from the
incumbent. In some cases the employee needs far less time but the training
period for all jobs and all employees is three days. In this example the formal
training period would be three days, even though some individuals may master
jobs in less time. Total preparation time included general education, specific
education, experience, on the job training, and any type of continuing education.

Three respondents at each firm were interviewed: the manager directly
responsible for the operation of the AMT, the person in change of human
resources for the AMT and an actual AMT operator when possible. Sampling all
three individuals allowed for triangulation as well as a fuller understanding Of all
the various types Of preparation an individual will need. Each individual has a
unique perspective as to what exactly the operator needs to know: the
operations manager will probably be most familiar with formal on the job training,
the HR person will be aware of educational requirements, while the Operator may
get considerable mentoring or other forms of informal training from other

Operators.

107

The specific questions to address this construct are explained in detail in
Appendix 2, but are summarized here. All Of the responses were recorded in
weeks. General education was considered a zero for high school or less. The
total number of weeks for each type Of training were added together to form a
single value for total preparation time.

1. What is the average level of general education reguired (in years) to perform
operational tasks at an average (acceptable) level ?

2. How much vocational training (in months) does an average Operator require
to perform Operational tasks at an average level ?

Vocational training is defined as: time spent in a specialized school or a union
sponsored apprenticeship program before an operator is hired.

3. How much formal on the job training (in hours ) does an average Operator
receive to perform at average (acceptable) levels ?

Formal training is defined as: time specifically set aside for the Operator to
learn their task. This can include class room time, time spent following
another operator, and or time spent in an apprenticeship program. Formal
training differs from vocational training in that the Operator is an employee Of
the firm.

4. How much experience (in months) are average Operators required to have to
be hired/assigned to the job ?

5. How much continuing education (hours per year) does the company
request/require of the Operators?

Continuing Education is defined as: time spent in a classroom or other
learning environment, after the operator has reached proficiency. This could
include classes on statistical process control, other quality tools, machine
maintenance or any other job related tasks.

The case studies were used to determine the importance Of each of these

elements of preparation time. Respondents were asked to address both the

amount Of each type of preparation as well as the importance. Respondents

108

were also asked to address the issue of substitution of different types Of training,
as well as for a formal job description. The final skill level construct used equal
weights because there was no clear pattern Of favoring one type of training over
another.

3. 4.2 Change /complexitv

3. 4. 2.1 The internal environment: product/process change

Change in the internal environment in the guise of product and process
Characteristics has been found by a number Of authors to impact the skill level of
the workforce. Product characteristics such as batch Size have been found by
various authors (Hazlehurst, Bradbury and Corlett (1969), Kelly (1990), and
Zicklen (1987)) to be related to skill levels with smaller batch sizes being equated
with higher skill levels. Milkman and Pullman (1991) found that the overall
complexity of the product (number Of parts) impacted the skill level of jobs.
Finally Smith (1992) found that the point in the product life cycle impacts skill
levels, with introductory and growth products being linked to higher skills, while
mature and declining products were linked with lower skills.

Process characteristics have also been found to impact skill level choices.
Jackson and Wall (1991) found that in high uncertainty environments Operator
control (in the guise of troubleshooting their own equipment) lead to lower
downtime because Operators prevented breakdowns. In low uncertainty
environments the performance was not significantly different with or without
Operator control which is a form of upgrading or increasing autonomy as well as

complexity of skills. Brass (1985) notes that when technological uncertainty is

109

high it is difficult to have pre-specified tasks for the employee to perform.
Technological uncertainty flows from a conversion process that is not well
understood (or changing).

Duncan (1972) lamented the general paucity of definitions and measures
for environments. Cooper, Sinha, and Sullivan (1992) specifically mention
product and process complexity but the constructs are never formally defined.
Kim and Lee (1993) have a measure Of technical complexity that includes level
of mechanization, level of predictability in production system and level Of
synchronization. Kotha and Orne (1989) explicitly define both process and
product complexity but do not test the constructs.

The three sets Of measures proposed by the above authors do have some
useful features. Two authors note that there are product and process
dimensions while the third notes that predictability (or dynamism) is an important
attribute of a measure of Change. The product/process measure will build Off
these works and will include those factors which have been empirically linked to
the skill level decision.

The specific questions (explained in detail in Appendix C) that were asked
to address the product / process change construct were:

1. How many distinct part/product families are made on the AMT?
A part/product family is defined as a group Of parts or products that are
similar in shape, size, material, and tolerances. A family is made using
similar set-ups (programs), jigs( if required) and tooling.

2. What is the average batch size for the AMT in actual parts produced and
hours Of supply?

110

Batch size is defined as the number of parts/products that are made on
the equipment without having to perform a set-up (or change program).

3. How many new parts/products (as a percent Of existing part/product mix)
are introduced on the AMT in a year?

New parts/products are defined as parts/products that can not be run on
the equipment using existing programs. If a new program does not have
to be written for the equipment then as far as the Operator is concerned
changes are minor because they are running a part/product in a manner
that is similar if not the same as, a part/product that already exists.

4. How many parts/products (as a percent Of existing part/product mix) are
retired from the AMT in a year?

A part/product is retired from the AMT if it will never be made on the
equipment again.

5. Percent deviation from master production schedule (MPS) in volume per
week.

6. What is the total number Of product and or process Changes in an
average month?

7. How predictable are the product and process changes?

The responses will be standardized as number of standard deviations
from the mean and then summed to form a single product / process value.

Question 1, number of part/product families, is derived from the pre-
dissertation case studies. The number of distinct products made by an AMT
installation is an indication Of the number of set-ups that must be performed,
the number of tool arrangements that are possible, the variety Of sizes,
shapes and tolerances an operator deals with and the potential for mistakes.
However, Boxes (with their infinite variety of end products) showed that the

real issue is product/part families. If set-ups, tooling, tolerances, shapes,

111

sizes and the like are all similar the level of product process change will not
be high, even if the size Of the product family is large.

Question two, batch size, is based on empirical findings that smaller
batches lead to higher skills (Kelly 1990). Questions three and four address
the amount Of churn in products. The more products that are introduced
and/or retired in a year, the more changes that the Operator will have to
make. This is true even if the operators are only material loaders. New
products may require: changes in procedure, new tOOls, new methods of
control, new programs, locating parts in new locations, using new suppliers
and a host Of other changes that may require some learning. The retirement
Of a product/part may mean that a familiar process is no longer performed.
New products require new learning, while retirements mean that
products/parts that have traveled far down the learning curve are no longer
produced.

Question five, percent deviation in MP8, comes from Handfield
(1992). This measure captures the amount of uncertainty on the shop floor.
Plants that can adhere to the MPS will have a higher degree of certainty
about what they will be doing, hence easier planning. In installations where
the MP8 is changing all the time it will be difficult to plan tasks, levels Of
material, tooling, capacity and staffing. This increased uncertainty will
require workers who are able to adapt to the change.

The final two questions are managerial perceptions about the amount

Of change, in general, on the shop floor as well as the predictability Of the

112

Change. This information may be useful because each installation may be
faced with specific types Of change (or predictability problems) that are
unique.

Because the product/process change construct has not been used
previously a construct validity study was conducted along with the
dissertation. Approximately 600 manufactures were sent a one page survey
that contained the items intended tO address this construct. This study was
carried out during the initial phase Of the dissertation in order to leave time to
refine measures. Perceptual information (in addition to questions 6 & 7) was
also be collected, as was information about engineering changes,
percentage Of standard parts, product complexity as a function Of total parts
and process reliability and dependability.

3. 4.2.2 ExtemaLenvironment: environmentaL complexity

The external environment has been studied extensively. Lawrence and
Lorsch (1967, 1986) looked at how well job requirements were known, the
difficulty in developing, manufacturing and selling a product and the amount Of
time between performing a task and receiving feedback on the outcomes.
Duncan (1972), who was careful to distinguish people’s perceptions of the
environment from the actual environment, looked at the amount of information,
knowledge Of outcomes, and the ability to assign probabilities to events. Kotha
and Orne (1989) looked at what they deemed the organizations scope which

included the geographic scope, the market and customers. Wernerfelt, Birger,

113

and Karani defined uncertainty as having 4 dimensions (in order Of importance);
demand, supply, competitive and external.

Duncan (1972), Bourgeois (1980), and Swamidass and Newall (1987)
note that perceptions Of the environment are more important than the actual
environment. If a manager perceives the environment as complex they will make
decisions that are designed for a complex environment. This conclusion has
positive and negative implications for the design of measures. On one hand
managerial perceptions may be a better explanation Of the choice of skill level
than the actual. On the other hand managers who have faulty perceptions Of
their environment may make a skill level decision that is not appropriate. \NIth
this in mind questions that can be externally verified will be included along with
perceptual information.

The following environmental complexity measure (explained in detail
in Appendix C which is a complete sample questionnaire with definitions) are
a subset Of the measures used by Swamidass and Newell (1987) that were
taken directly from the seminal measures suggested by Duncan (1972). All
items were scored on a seven point scale that ranges from always
predictable to never predictable.

Actual users Of your products.
Competitors for your supply Of raw materials.
Competitors for your customers.

Government regulation controlling your industry.

9:59.”?

The publics Political views and attitudes towards your industry.

114

The structured questions were augmented by questions regarding the
number Of customer groups and competitors as well as an Open ended
discussion Of the firms external environment, in terms Of management’s
perceptions. The environmental complexity construct 'was a summation Of the 5
items.

3. 4.3 Managerial discretion

The existence of a union is Often linked tO the skill level Of the workforce
(Kelly 1990, Chaykowski and Slotsve 1992, Wood and Albanese 1995). But a
number of changes have occurred and continue to occur in the relationship
between unions and management (Kochan, McKersie, and Cappelli 1984). The
impact Of these changes may be a more collaborative relationship (Thomas
1991) that might indicate a willingness Of the two sides to work together,
removing management’s need to lower skill levels to keep control of the
workforce. Therefore the Managerial Discretion construct will capture not only
the existence of a union but the type of relationship that the union and
management have. A union in and Of itself should not impact the level of skills of
the workforce, but poor labor relations would lower the skill level choice. This
construct interacts with the two change constructs. Firms with low levels Of
Managerial Discretion will choose a lower skill level than firms who have similar
environments but higher levels Of Managerial Discretion. Therefore, this
construct will be operationalized to interact in the following manner.

When there is a union with a seniority system (for the AMT) and/or

industrial relations are perceived by management as poor or combative lower

115

skill levels will be chosen. Union environments without seniority systems for the

AMT that have good relationships with management are not hypothesized as

changing the skill level choice.

The questions (explained in detail in Appendix C -sample questionnaire)
used to address this construct will be:

1. Are the Operators of the AMT unionized?

2. Are jobs for the AMT assigned based on seniority?

3. How would you describe the relationship between management and the
Operating workforce, or their union (ranging from collaborative at one end of
the scale to combative at the other)?

Managerial perceptions are most important regarding the quality of labor
relations, but is possible that a manger would give a response that is socially
desirable (cite) rather than admit to a distrust Of the union. The pre-dissertation
case studies also raised a question about the balance of power in the
relationship. The management at Composites has a good relationship with their
union but they are not hesitant to exercise their power to fire employees or
Change work rules. However, Cars has a history Of capitulating to union
demands in order to keep production flowing. Cars may have chosen lower
skilled jobs because they fear the power of the union, even if they have a good
relationship. The first stage of data collection (see section 3.5.1) will be used to
further develop the Managerial Discretion construct.

The following questions may also be indications of the quality Of labor

relations and or the power balance between the firm and the workforce:

116

1. How many grievances for the work area where the AMT is installed were
there the year the AMT was installed? In the previous year? In the present
year?

This question addresses the workers reaction to the technology. A large
number Of grievances, during and after installation would indicate that the
workforce either does not trust management, or the technology is being used
in a manner that employees view as detrimental to their future. A persistent
pattern of grievances could indicate poor relations between the union and
management.

2. Was there been a strike or slowdown over the installation of the AMT? Over
work practices for the AMT after installation?

This question, which deals only with work stoppages or slowdowns that are
related to the AMT, also assesses the operators reaction to the technology.
Strikes and slowdowns may lead managers to want more control to avoid lost
production. Strikes also indicate an inability to informally solve problems.

A final area that is not directly addressed in this study is the number Of
different unions that either operate or maintain the equipment. Multiple union
sites may have less flexibility and be more likely to choose low skills, than single
union sites. Because of the ease of collecting this information it will be gathered
for future analysis.

3. 4.4 Appmpriate skill level

The skill level Of the workforce is a result of the managerial decision as tO
how to respond to the level of product/process change, the amount Of
environmental complexity and the type of industrial relations the firm has. An
appropriate skill level is one where employees have all the necessary tools to do
theirjobs but where skills are not wasted. Overly skilled employees become

bored and lose motivation to use the skills they do need. Additionally, there is a

high cost to maintaining a high skilled workforce. Highly skilled workers expect

117

high wages, may require ongoing training and will be attractive to other firms.
Conversely, workers who do not possess the necessary skills will be incapable of
doing their jobs.

Duncan (1972) and Jelenek (1977) both point Out that there are different
levels of complexity for different parts Of a firm. They also note that
manufacturing may be better buffered from the external environment of the firm
than other functions. The findings of Muller (1994) and Wood and
Albanese(1995) support the hypothesis that for manufacturing jobs the external
environment may not have the impact of the internal environment. This also
makes intuitive sense. A firm who has a large number Of competitors and
customers may make a single product or product line. From the standpoint Of
the manufacturing function there is a good deal of certainty of what they are and
will continue to do, while there is little certainty as to who the firm will sell tO and
how much. Makers of mature commodities best fit this scenario.

The Opposite is also possible. A firm may be one of a few companies (or
perhaps the only company) in an industry with a stable customer base, and
products that change infrequently, yet have a complex manufacturing
environment due to the nature of the products or the sheer number Of products
made. Many firms in the defense industry (until recently) faced this situation.
Products were extremely complex, yet they were sold to a single customer.
Once a bid was accepted by the government work might continue for years. The

external environment was almost fixed for extended periods of time. However

118

the work on the shop floor was extremely complex due to the nature of the
equipment.

Tool shops for large manufacturing concerns also face certain external
environments and uncertain internal ones. The tOOl rOom makes an endless
variety Of one Of parts to repair equipment and or improve equipment. What is
needed and when is Often impossible to predict However, the external
environment is characterized by a single customer group that never changes.

The findings Of Muller (1994) and Wood and Albanese(1995) might
indicate that an external measure of complexity is not needed, but the sheer
volume Of work that notes that environmental complexity drives structure (i.e.
Burns and Stalker 1961, Lawrence and Lorsch 1967, 1986) as well as the
findings Of, Kimberly and Rottman (1987), Wall and Davids (1992) and
Jarvenpaa and Ives (1993) which support the notion of environmental
complexity impacting managerial choices indicates that external factors should
have some impact on the skill level Of the workforce used for an AMT.

Table 3-2 shows the appropriate skill level for the operators of an AMT
given the level of product/process change, the level Of environmental complexity
and the level Of Managerial Discretion . The low, medium, high trichotomy Of
product/process change mirrors similar works (i.e. Van De Ven and Delbecq
1974) that attempt to allow for sufficient variation and ease of analysis. The
trichotomy is far more appropriate than a dichotomy of skill level because as
authors like Adler (1990) note there are firms not following the extreme

strategies. Environmental complexity is dichotomized because Of the findings

   

119

that show that internal change has a larger impact on the manufacturing
function. When environmental uncertainty is high, skill levels for low and
medium levels Of product process change will increase. When it is low, the

product process measure will drive the skill level choiCe.

Table 3-2 Appropriate skill level

 

  
   

Product /
Process Change

High level Of Discretion Low Level of Discretion

 

 

High skills High skills Medium skills Medium skills

 

 

 

 

 

   

Medium skills High skills Low skills Medium skills
Low skills Medium skills Low skills Low skills
Environmental L H L H
Uncertainty

 

 

 

 

 

Table 3-2 also considers the effect of Managerial Discretion. When a
company has both a union and a seniority system for the AMT, it will have limited
discretion as to who works on the AMT. The employees with the appropriate
seniority level may be unwilling or unable to learn new skills even if the
environment seems to dictate this change. Successful companies in this
situation will use lower skill levels (from high to medium and medium to low) and
give more responsibility to management or other union groups (such as skilled
trades) in an attempt to deal with employees who can not or will not work at
higher skill levels.

Companies with poor relations, as well as a union and/or a seniority
system, will also opt for lower skills if they are to be successful. These firms not

only have limited control over who works in what job, they have limited ability to

 

120

work with the employees. High skilled jobs are generally complicated and have
at least some degree of autonomy. When the company cannot pick who will
work on a job and does not fully trust the workforce, they want jobs that are easy
to control; low skilled jobs. This may be expensive in Complex environs.
However, poor relations with little or no flexibility about who runs the system and
under what conditions will put a company in the position where they must choose
between managerial control (and the associated higher overhead) or the
potential for even higher costs from workers who do not have the same interests.

Appropriate skill level was a dichotomous fit measure. For data analysis
purposes the construct was operationalized as a 1 when there is fit and a zero
when there was not fit.

3.4. 5 Successful adoption

Measuring the level Of success Of an AMT installation must be done at the
level Of the AMT not at the firm level. Firm level measures will be difficult, if not
impossible to use for comparisons for a number Of reasons.

First, as Vickery (1992) notes when choosing a measure Of success in the
furniture industry, different industries use different financial measures to indicate
success. This makes it difficult to say that specific levels of ROI or ROS are
indicative Of success or failure across industries.

An equally large problem involves the placement Of the AMT within the
firm. For a small company with a single production process, firm level analysis

may indeed be appropriate. But for large companies where the AMT has a small

121

impact on overall financial measures it will be difficult tO determine what impact
the AMT really had.

Non-financial measures at the firm level are just as problematic. Arthur
(1994) used scrap rate and labor efficiency as indicatOrs of performance. These
measures are flawed because a company making smaller batches may have
lower labor efficiency than a firm with long runs, yet be more successful in
satisfying customers. Similar problems arise from using scrap rate, which may
be contaminated by the complexity of the product, the quality levels required, the
quality of suppliers and a host Of other variables. An additional problem with
measures of this nature is that success has two dimensions (Zairi 1992) ,
technical and commercial success, while most of these measures account only
for technical (scrap rate) or commercial (ROI) factors but not both.

Financial measures at the AMT level are just as problematic as at the firm
level. Comparisons will still be difficult, and there is the additional problem that
many firms install AMT for non-financial reasons (Boer, Hill, and Krabbendam
1990). Specific cell level measures are suggested by Pricket (1994) who uses a
production index ((sales -costs) / sales) as a gauge Of success. This measure,
like many other productivity measures (such as total factor productivity ) suffers
from the same problem as Authur’s scrap rate and labor efficiency measures,
namely comparisons are difficult if not impossible across firms due to differences
in batch sizes, product complexity, process complexity, number Of new product
introductions and a number of other possible factors. Kelly (1995) adjusts for the

product complexity in a measure that looked at average machine time for

122

average product adjusted for product complexity. This measure addresses some
Of the shortcomings of other methods but is still mainly a technical measure with
no notion of commercial success.

Zairi(1992) defines success as occurring when the set Objectives and
predictions carried out by the adoption strategy are fully realized. The closer
these predictive plans are to the achieved targets, the higher is the measure of
success” and then warns that there is no one right measure to determine AMT
success because each installation is different. Therefore, success was
measured as goal achievement, rather than as a specific efficiency or financial
measure. This approach has been used by a number Of authors (Boer, Hill and
Krabbendam 1990, Boer and Krabbendam 1992, Beatty 1993) when dealing with
AMT in general and FMS specifically.

This approach is problematic from the standpoint that the goals may not
exploit the relative advantage of the technology (Venkestann 1987 ) or that the
original goals may be forgotten or changed by the time research is carried out.
These, and other problems of perception (perceived goal achievement), can be
troublesome but a goal achievement measure makes it possible to define the
success] failure of an installation without having to make comparisons to other
organizations. It is also possible for a manager to determine if the AMT does
what was intended, while it may be difficult or impossible to determine the real
costs of Operating the system or the true efficiency Of a system. Finally, goal

achievement allows for both technical and commercial goals to be examined.

123

The plant manager and the person in charge of Human Resources were
surveyed as to the achievement of AMT goals. In order to standardize
responses, a list of goals derived from a list Of critical competitive capabilities
from the operations management literature (Miller and Roth 1994), will be
presented to the managers, as well as an option for other goals. In addition to
being asked about the success at meeting standardized goals, respondents
were also questioned about the actual results achieved be they intended or
unintended. Finally, open ended questions about the overall success of the
installation along with explanations (especially if there is no agreement with
previous responses about specific goals) explored the importance of intangible
factors such as learning or increased integration. The following questions were
used to address success:

1. Please rank the following goals on importance to the AMT installation (at the
time of installation). '

quality

cost/efficiency
delivery/responsiveness
flexibility

innovafion

other

2. For each goal, rank the level Of achievement from 1-7 (explanations will be
asked for).

3. Did the installation have consequences that were not originally intended?
what? Why?

4. How successful do you consider the installation? Why?
The first two questions were used to formulate an overall success

measure. Success was determined by multiplying each factor’s importance by

124

the level of achievement. The impact of unintended consequences can be
added to the measure the same way. The Open ended portion of the unintended
results as well as the final question address issues that have not been captured.
A firm may view a failed installation as a success because they have learned that
they do not want another AMT, or perhaps the lessons learned were useful in
later installations.

Finally, some financial and customer service data was collected.

Although this information is difficult to make any type of comparison with, it can
be used as a method of triangulation to be sure that a firm is not trying to portray
their AMT performance as significantly better or worse than it actually is.
Information of ROI, sales, market share, and the like will be asked for (with a
promise of anonymity). Additionally, when possible, information about the
satisfaction of the direct customers Of the AMT (be they internal or external) wa
solicited.

3.5 Methodology:

3. 5.1 Data collection:

The constructs of interest are well developed in the literature but not well
tested. There are a number of studies that conclude managerial choice drives
the skill level decision for the operators of AMT but far fewer studies that set out
to test this relationship explicitly. Hence, few firm conclusions as to what drives
choice and how to measure it have evolved. There is evidence that
environmental factors impact skill levels of work, as well as evidence that unions

(especially with poor labor relations) lower the skill level Of Operators. However,

125

the interaction Of these variables has not been well examined, nor has the
impact of choice on success or failure of an AMT.

Exploring the proposed relationships, as well as allowing for deeper
investigation Of the constructs and relationships propOsed was done using
structured interviews in a field setting (which are a form of case study according
to Yin 1994). This allowed for the exploitation Of the strengths of both case
studies and surveys while eliminating many of the problems of both. Case
studies are useful for building novel and valid theories and getting to the heart of
relationships (Eisenhardt 1989, Kerlinger 1986, Yin 1994). But case studies are
Often difficult to generalize from (Kerlinger 1986). Surveys are useful because
standardized measures (that are a necessity for making comparisons (Fowler
and Mangione 1990)) can be used across a broad population in order to make
generalizable conclusions. However, surveys are not always good at getting to
the deeper depths of many relationships (Brewer and Hunter 1989).

Structured interviews in a field setting allow for the strengths Of both
approaches to be exploited. The interview with its standardized questions allows
for comparisons across multiple firms. At the same time, a field setting allows for
an examination of actual practices as well as open ended discussions, both of
which may lead to interesting or important conclusions about the relationship
between the skill level Of the workforce as it relates tO success or failure of AMT.
This design also allows for the collection of quantitative data as well as
qualitative data. Quantitative data, such as the proposed measure of worker

skills, can be compared with what actually happened on the shop floor, or how

126

management describes the workforce. It is possible that workers will receive a
good deal of informal on the job training that is not captured by the quantitative
measure. A survey would not pick up on this discrepancy, while the proposed
method would.

The data gathered in the field was used in an interative fashion to improve
the model. Huberman and Miles (1994) note that the process Of qualitative
research should be an interative one where data is collected, reduced, displayed
and interpreted throughout the study. This process Of refining theory from the
data gathered in the field is similar to grounded theory development (Straus and
Corin 1994), controlled opportunism (Eisenhardt 1989), or the suggestion of
using replication across multiple cases to refine theory (Yin 1994).

Respondents were the manager most responsible for the daily operations
Of the AMT, as well as the person who is in charge of the personnel practices for
the AMT. For site visits an Operator Of the equipment was interviewed if
management would allow it. Multiple interviews allow for triangulation, as well as
the Opportunity to learn how various functions view the installation. The interview
with an Operator provides a unique opportunity to see how someone outside Of
management views both the AMT, and the way it is used.

3. 5. g Sample selection:

The most representative samples are those that are randomly drawn from
the target population (Cook and Campbell 1979). But the random sampling for
representatives model (Cook and Campbell 1979) requires resources, both time

and money, well beyond the scope Of this dissertation. Additionally, the

127

information as to what company uses what type Of technology is not generally
well known, which makes it difficult if not impossible to truly define the population
of users. Finally, the small number of FMS installations would make a
randomized design problematical In terms of getting access to a truly
representative sample.

The sample in this study was a convenience sample Of firms using FMS
and CNC equipment. The paucity of FMS installations make it difficult to ensure
a wide range of variance across all of the variables of interest. However, recent
work by Kaku (1994) shows that even in a small sample Of FMS firms there is a
great deal of variety in the ways in which the equipment is used. The firms
studied by Kaku was the starting point for building a sample base.

The inclusion Of two types of technology is similar to deliberate sampling
for heterogeneity (Cook and Campbell 1979) where a wide range Of instances
from each group are represented in the sample. Cook and Campbell also note
that this design is especially useful for “avoiding the pitfalls of restricted
influence that results from failure to consider sampling questions about
secondary influences”. The CNC firms will provide a check for the impact of
technology (as proposed by much of the Operations management literature) on
the skill level of the workforce.

The population Of interest is plants using either FMS or CNC equipment.
The Number of FMS installations in the country made it difficult to control the
number Of installations that would agree to take part in the study, or the

industries that the companies would be involved with. The small number of FMS

128

installations lead to a pure convenience sample of FMS located in the Mid-
western portion of the United States, but the CNC population was restricted in
the following ways:

1. CNC firms (who are numerous) will be chosen to be close matches to FMS
firms in terms Of industry based on SIC codes and indexes Of “local”
manufactures (i.e. Dunn’s Industrial Guide, Michigan Business Directory)

2. Multiple plants from the same company were studied if the plant management
has control of workforce practices within the plant (no company wide training
or compensation programs that limit managerial discretion).

The sample was drawn from a number of sources. FMS locations were
located through the suppliers of the technology such as Cincinnati Milacron and
lngersol Milling (who have a vested interest in increasing diffusion so hopefully a
vested interest in providing names Of customers), and academic experts on FMS
such as Katherine Stecke at University of Michigan, lnjazz Chen at Cleveland
State University, and BK. Kaku at University of Maryland. Users Of CNC
equipment in the relevant industries were be identified using the same sources
as for FMS; as well as local sources of information such as the Society of
Manufacturing Engineers, the users Of FMS, experts such as Dr. Steven Melnyk
located on campus and a telephone survey of companies in the same SIC code
industries as the FMS firms.

3. 5.3 Limitations of research

There are four potential limitations of the research: selection bias, the

small sample size, non-respondent bias, and the potential for respondents to

have a preconceived notion that a high skilled workforce is more desirable.

129

The sample is clearly a convenience sample and not a random
representation of the larger population of AMT users. Despite this limitation, the
generalizability Of the study may not be seriously threatened. Sampling from
users of two distinct production processes allows fOr testing Of impacts that are
purely driven by the technology. Additionally, firms in a variety of industries can
be chosen to limit industry impacts. Finally, the method of data collection may
mitigate some generalizabilty concerns by having the benefits of the valid theory
developed through a case study and the ability to make comparisons through the
structured interview questions.

A second concern is the sample size. The small sample size may make
the use of many statistical tests questionable. But it is felt that an in depth
treatment of a small number Of installations will be much more enlightening than
a broader study with less depth. The depth is seen as vital based on the findings
from the pre-dissertation research which showed that many existing measures
(such as locus of programming as a proxy for skill level) are flawed. The in
depth nature of the study will allow for discussions not possible on a with a mail
survey. The design also allows for refinement of the proposed measures, many
of which have not been used before, or have not been used for studies Of this
nature.

A third concern is the possibility Of respondent bias. Although no data will
be collected from non-respondents it will be possible to make some comparisons

between sampled firms and general populations. General performance data,

130

such ROI, ROA, customer satisfaction, scrap rates and the like will be gathered
and compared to industry averages.

A final concern is that many firms may view specific workforce practices
as either cutting edge, more professional or perhaps. socially desirable. Scott
(1987) notes that firms may adopt an innovation because they feel it is the most
professional or somehow the best choice. This may also extend to saying that
they have adopted practices when in fact they have not, or have only made
minor steps toward the innovation. This problem may appear in discussions of
skill levels. Many recent works (Walton 1985, Saraph and Sebastion 1992,
Aurthor 1992) make a compelling case for using a commitment strategy for
human resources in all settings. Respondents may believe that this is the
response that the researcher is looking for and hence give socially desirable
answers that do not reflect actual practices. This may be especially true of
Human Resource managers who wish to give the impression of being true
“professionals”. This problem is mitigated by talking to someone on the shop
floor and actually seeing the work (when possible). Operational employees and
plant tours will provide a clearer picture Of what is happening and may point to
respondents whose answers are biased.

3. 5.4. 1 Statistical analysis

Statistical analysis was performed in three steps. The first step was to
test hypothesis H1a to determine if skill level is related to the success or failure
of an installation. The second step was to test the strategic choice framework

and the variables hypothesized to drive choice (hypotheses 2-4). The third step

131

was to test hypothesis H1b: the appropriateness of choice is related to adoption
success or failure .

Step one tested hypothesis H1a: skill level has no impact on adoption
success. This hypothesis was tested by correlating Success with skill level (total
preparation time). If the strategic choice framework is correct there should be no
relationship (no covariation) between success and skill level. The performance
Of this test was greatly aided by the use Of visual aides. It is possible that only a
few of the successful firms will be using a specific strategy vis-a-vis workforce
skills. If this is the case, there may be a fairly strong correlation between skills
and success which is due to the shape Of the distribution rather than a single
right skill level. A visual aid, such as a scatterplot, would show which firms were
successful, yet outliers. These firms, as well as unsuccessful high skill firms
would be explored in greater detail.

Step 2, testing hypotheses h2-h4 regarding the drivers of choice was
accomplished by regressing complexity/change and managerial discretion
variables on the skill level (total preparation time) Of the workforce. This test was
be performed using the moderated regression approach suggested by Cohen
and Cohen (1983) and Stone and Hollenbeck (1989. This hierarchical
regression method worked as follows:

1. The main effects Of product/process Change, environmental complexity, and
managerial discretion would be entered into the model.

2. The interaction effects of managerial discretion and the two change]
complexity variables would be entered in the model.

132

If managerial discretion has an interaction effect, the slopes Of the regression
lines where the managerial discretion variable is hypothesized to have an impact
would be different from the slope of the regression line for firms where
managerial discretion in hypothesized to have no impact.

The third step, testing the overall model, linked the two halves. The
proposed model suggests that successful adoption is related to the appropriate
Choice Of worker skills (using a trichotomy Of skill levels to form the zero ] one fit
measure). This relationship would be tested using the correlation between
appropriateness and success. The correlation between appropriateness and
success should be significant, though it may not be all that large with a small
sample size and the number of unmeasured impacts on success.

A stronger relationship may be the one between failure and inappropriate
choices. Appropriateness of choices is one of many factors that drive success.
Making choices that are appropriate for the given environment may be a
necessary but insufficient driver of success. However, failure, or low levels Of
goal achievement, may have a strong correlation with inappropriate choices. A
workforce that is not able or willing to perform the necessary tasks, or that is
underutilized (with attendant costs), may cause the entire system to malfunction.

In addition to testing the overall model, the sample has been designed to
allow for a number Of comparisons that resemble quasi-experiments (Cook and
Campbell 1979). Comparisons (t-tests) can be made between the users Of FMS,
and the users of CNC tO see if there are technology impacts on skill level choices

(as predicted by much of the Operations management literature). Firms in similar

133

industries but using different technologies can also be grouped to see what
impact industry has on Skill level choices. Union and nonunion firms could also
be compared. Other experiments could be conducted tO test a number of the
contingency variables such as size, frequency Of purchase, and type Of markets
(Hambreik and Lei 1984, Bluedron et. al. 1994) that have been found to have
impacts on firms structure and strategy.

Finally, the methodology allowed a more in depth examination Of
interesting or anomalous cases. Visual displays (discussed in detail below) were
useful for identifying outliers and or firms who are defying either the model or the
dominant logic. These outlier cases may be used to buttress some of the
statistical conclusions, to build new models if the proposed model does not stand
up to statistical testing, or to show weaknesses in the statistical procedure.

3. 5.4.2 Data analysis: visual displays

Many authors note that visual displays Of data are very useful when
dealing with quantitative data (Eisenhart 1989, Harlwig and Dearing 1979
Huberman and Miles 1994, Miles and Huberman 1994, Yin 1994). Visual
displays were useful for grouping the low, medium and high levels of skill, to
study the interactions between managerial discretion and the complexity
constructs, as other relationships. Visual displays were also be useful for
identifying firms who are doing things other than what is predicted. Additionally,
Yin (1994) notes that pattern matching (between cases or comparing to an

empirical model) is one of the dominant modes Of analysis when doing case

134

study research. Finally, it has been noted that summery statistics (Harlwig and
Dearing 1979) are not always useful, and may hide important information.
3.6 Validity issues

Cook and Campbell (1979) note that there are four types Of validity which
all research should judged on: statistical conclusion validity, construct validity,
internal validity and external validity. Statistical conclusion validity is a large
potential problem for this study, since the power Of a study with such a small
sample is likely tO be very low. Cook and Campbell note that when the sample
size is small it is dangerous to rely solely on statistical significance, and this
study addresses concerns with statistical conclusion validity by using a host Of
approaches such as visual displays (Miles and Huberman 1994), replication and
pattern matching (Yin 1994), and explorations Of outliers, along with statistical
tests.

Construct validity, or confounding, is a major problem for case studies
(Yin 1994). This study used statistical procedures as well as the research design
itself to limit construct validity problems. The measure of product/process
change is the only measure to be used that has not been used previously.
Therefore this construct was tested using a separate construct validity study.

Approximately 600 firms were sent a one page survey with the product
process items and the external environment items. The firms will be selected
from three sources: a list of companies who participate in the Materials and
Logistics Management Program at Michigan State University; Dunn’s Industrial

Guide and the Michigan Business Directory. The first source should improve

135

response rates, while the other two were chosen to increase the variety of firms
sampled. This sample will be large enough (Scwab 1980) to perform factor
analysis to determine if all items load on the same underlying constructs.
However, it should be noted that these constructs are being measured with
indexes rather than scales and it is likely that individual items will not load on the
two factors hypothesized. If this is the case the construct validation study will still
be useful to determine the relationship between internal and external
environments, as well as questions that may be vague, misleading, and or
redundant.

The other constructs are to be measured using existing scales which
should have some degree of validity. The design Of the study helps to deal with
construct validity by: using structured questions, having pre-specified
relationships based on an exhaustive search of the literature, _pre-testing and
refining of the measures, collecting information from multiple sources and
planing for respondents to review field notes after they have been typed and
organized. This combination of strategies should mitigate some construct
validity problems (Yin 1994).

lntemal validity concerns deal with establishing causal relationships
between variables and in non-randomized designs Cook and Campbell suggest
that the researcher must explicitly rule out all possible threats one by one. Many
Of the possible threats are not Of concern in this study because the design is not
interested in the changes wrought by the introduction of AMT ( if it were the pre-

test condition would be non-computer controlled automation and the post test

136

would be AMT) but rather, what is done with the AMT. Therefore, problems such
as history, maturation, instrumentation and statistical regression to the mean
may not be a problem. The main internal validity issue will be judging the
causality of relationships.

This study assumes that product/process change, environmental
complexity, and managerial discretion are the drivers of the skill level decision. It
is also possible that the skill level decision drives the decision as to the types Of
products a firm makes, the environments it is willing to compete in and the type
Of relations that exist between the workforce and management. In order to
address this issue Yin (1994) suggests the use of time series analysis, where the
order Of event occurrence is taken into consideration. In order to determine the
temporal order of decisions as to worker skills, markets, and the like questions
are included in the survey instrument to try and determine when (or if) changes
in practices and markets have occurred. For instance, Composites installed
CNC equipment in an attempt to enter new markers. This lead to greater
complexity on the shop floor, and the need for a more flexible workforce.

The final validity issue is external validity; the ability to generalize beyond
the present study. Multiple case studies with a convenience sample will make
broad generalization to all AMT installations difficult, but the study is designed to
be generalizable to multiple technologies as well as industries. Two distinct
technologies allow for more robust conclusions as to the impact of technology
verse the impact of choice. Matching firms with different technologies in similar

industries also allows for control of industry effects. Other firm characteristics

137

that have been shown to be important contingency variables (Hambreik and Lei
1984) will also be measured their effect tested (through corelations and or t-
tests). The end result should be generalizable across a number of technologies
and industries which should make up a majority of the AMT population.

3.7 Reliability issues

The issue Of reliability is always troubling in case study research (Yin
1994). This dissertation will mitigate the reliability problems associated with case
studies in the following manner. First, the installations that were studied using
the case study format followed a strict protocol. The use Of a strict protocol may
not allow for complete replication, but it does give a clear guide to what was
done (Yin 1994). The multi-method, multi-respondent nature Of the case studies
will also increase reliability. The use of qualitative and quantitative questions,
multiple respondents and additional information gathered by actually Observing
the installation should allow for triangulation Of methods and reduce concerns
about reliability (Brewer and Hunter 1989).

The case studies also allow for refinement of measures. Respondents
were asked to define many of the constructs themselves, in an effort to assure
the clarity of survey instruments. Questions that were unclear, or that measure
constructs other than those of interest were changed, or adjusted, reducing

ambiguity and increasing reliability (Kerlinger 1986).

Chapter 4 .

ANALYSIS
4.1 Overview and chapter contents

Chapter 4 describes the analysis performed on both the construct

validation and primary sample data. Section 4.2 discusses the construct
validation study including the sample, analysis, a brief description Of the results,
and a discussion Of how the construct validation results were used to analyze the
data for the primary sample. The results Of this analysis are provided in Chapter
4 rather than Chapter 5 (discussion) as the analysis focuses on a set of
environmental indexes used to test the relationship between Strategic choices of
operator skills and AMT adoption success. Section 4.3 discusses the sample for
the primary data collection, similarities and differences between the primary data
and the construct validation data, and the distributions of the data in the main
sample. This analysis provides some insights into whether the sample used in
the primary data collection effort is representative of the broader cross section Of
manufacturers in the construct validation sample. Similarities and differences
between the primary sample and the construct validation data may help tO
explain some of the outcomes, or at least describe the findings within the context

of manufacturing plants in general. Section 4.4 describes the analysis performed

138

139

on the quantitative data from the primary sample used to test the hypotheses
presented in Chapter 1. Finally, Section 4.5 describes the supplementary
analysis done using some Of the qualitative data collected in the primary data
collection effort that may provide additional insights. into the proposed
hypotheses.
4.2 - Preliminary data analysis

The construct validation study was a mail survey performed at the plant
level (for a discussion on the appropriateness Of plant level data collection see
Flynn et. al.1995). This part of the analysis addressed two specific issues: the
measurement of product/process change, and the relationship between internal
and external plant environments. The first question addressed was: do the
proposed variables that make up the product/process change construct load on
a single factor or on multiple factors ? The latter case would suggest the use Of a
shop floor environment index as an appropriate measure. The variables that
make up the proposed product/process measure were selected for two reasons:
1) previous research and the pre-dissertation case studies suggested that these
shop floor variables were potential predictors of Operational skill levels, and 2)
these variables had high face validity in terms of their individual and collective
ability to describe the level of uncertainty in the manufacturing environment.

It was hoped that the variables would load on a single factor Of
product/process change. However, there was no a-priori reason to assume that
this would occur because the variables of interest need not covary. For instance

a plant may only have a single part family that does not change over time (no

140

churn), but production is in small batches to allow some level of bolt on
customization. In this example batch size indicates a more complicated
environment while families and churn do not. If the variables did not form a
single factor the intention was to use a summed index Of the standardized
product/process variables.

The second issue is the relationship between internal and external
environments. Because of the relatively small number of cases in the primary
sample, there was not enough statistical power to test the relationship between
internal and external environments conclusively. The primary sample is also
limited in terms Of the process technologies and industries represented and any
conclusions reached using only the primary data would be difficult to generalize
across a broader populations of manufacturing firms. However, the construct
validation sample covers a large number of industries and process technologies,
and the results from tests of this data are may be more generalizable. Testing
the relationship between internal and external environments is important
because many organizational researchers (i.e. Burnes and Stalker 1961; Covin
and Slevin 1989) argue that the external environment determines structure
(lntemal environment) and hence the resulting Operational jobs and skill
requirements. If it is indeed the case that measures of the external environment
are statistically correlated to measures Of the internal environment, then there
may be no need for the product/process measure. In other words, if prior
findings that external environmental factors drive choices in the internal

environment are supported, than the external environment measure can be used

141

to test the drivers Of skill choices without introducing a new, un-validated
measure. However, if there is no relationship between internal and external and
internal environments (or if the relationship is weak) than there would be a need
for both the product] process complexity and external complexity variables to be
included in the analysis of the drivers of skill choices.
4.2.1.1 - Construct validation 3am

The construct validation instrument was a two page mail survey (see
Appendix D) sent to 600 randomly selected plants in the United States. Plants
were randomly selected from 2 sources; Dunn’s Industrial guide (1994-1995) and
Michigan Manufacturing Directory (1995). Of the 600 plants approximately 150
were from Michigan with the remaining 450 randomly distributed throughout the
remaining 48 continental United States. Surveys were sent to plant mangers
and almost all Of the responses were filled out by a the top manager at the plant
who was usually either a president (single plant companies) or a plant manger.

Two weeks after the initial mailing a second reminder mailing was sent to
non-respondents. The total response rate from the two mailings was 120 out Of
600 (20%). Approximately 40 surveys were returned either because the plant’s
address had changed, the plant was no longer in use or the facility did not
manufacture anything. Of the 120 responses, 91 were usable. Responses were
not used if they were incomplete or if the company did not manufacture any
products at the facility. The overall response rate is 120/560 or 21% and the
useable response rate is 91/560 or 16.3%. This response rate is fairly low, but

not uncommon for unsolicited mail surveys (Fink 1995).

142

Responses were received from plants in 23 Of the 48 continental United
States. The average plant employed 203 people while the median plant
employed 129. Plants ranged in size from 8 employees to over 4000 employees.
The plants represented 50 different 4 digit SIC codes. Approximately 38% Of the
plants were unionized. Overall the sample represents a diverse group Of
manufacturing firms who vary greatly in terms Of size, location, industry and
workforce practices.

The sample may suffer from two sources of bias. The first potential bias
is that the average size Of the respondent plants was larger than the average for
all manufacturing plants in the United States. The average sized plant in the
United States employs 52 people (Manufacturing USA 1994) while the mean for
this sample was approximately 260 employees. This bias in the respondents
was probably due to a few very large plants (there are 4 in the sample over 1000
employees) which skew the results slightly. When the 3 largest plants are
removed from the sample the average plant size drops to 200, which is closer to
the overall average for all manufacturing facilities in the United States. The
second issue is the use of a single respondent for each company, especially for
perceptual data such as the external environment. However, Miller and Roth
(1994) note that when respondents are placed higher up in a company’s
managerial ranks the impact of single respondent bias is reduced. Because the
vast majority of respondents were the senior person at the location the biases

introduced through the use of a single respondent are possibly minimized.

143

4.2.1.2 - Data preparation - outliers and missing values

Prior to analysis Of the construct validation data, the sample was screened
for outliers and missing data points. The sample was relatively small and every
attempt to preserve data points was made. NO response was used if more than
two variables within the same hypothesized index were missing. Listwise
deletion Of cases with missing data was not performed, in order to preserve
power, so a different approach was needed.

Roth (1994) notes that there are a number Of methods for dealing with
missing data. In situations where less than 10% of the data is missing (as it was
in this sample) Roth notes that Monte Carlo studies indicate that there is little
difference in parameter estimates no matter what missing data technique is
used. However, the same author notes that the use of imputation strategies
such as regressing the missing values from other values is not acceptable when
doing factor analysis, because this method effectively “stacks the deck”.
Therefore, replacing the missing value with the mean Of that variable was chosen
as the simplest manner in which to replace missing data that would not unduly
effect the subsequent analysis. Post hoc analysis suggests that correlations
between variables before and after the missing values were replaced did not
change by more than .05, and no relationship went from significant to non-
significant after the missing data was replaced.

In screening for outliers, every effort was made to maintain sample size.
However, 9 plants were removed because they had one or more variables which

were significant outliers. For instance the largest plant had 2600 more

144

employees than the next largest plant and was greater than eight standard
deviations from the mean plant size. If a plant was an extreme outlier on one or
more variables the entire data point was eliminated. The final sample size for
analysis was 82 plants.

4.2.1.3 Afilysis —Conelations

The construct validation study was intended to address two issues. First,
do the proposed product/process variables load on a single factor or will an index
Of internal complexity be needed? Second, what is the relationship between the
lntemal and external environments ?

It was assumed a-priori that there would be no relationship between the
variables that made up the product/process construct because there are many
possible scenarios where some Of the variables indicate high levels Of change
while others do not. For instance a company may produce relatively few part
families in large batch but replace a significant portion Of their parts in any year.
Batch size and families would indicate low levels Of product/process change
while churn would indicate high levels Of product/process change. Therefore
factor analysis was not expected to result in a single factor or factors. Instead it
was assumed that an index (Nobel 1995) of internal complexity which is referred
to as “product/process Change” would be required. Since this was indeed the
case, analysis was performed by examining the correlations between variables in
the product/process construct as well as looking at correlations between

product/process change and environmental uncertainty.

145

Exploratory factor analysis indicated that there was not a single factor or
even a pair Of factors underlying the variables in the product/process change
measure. These variables included batch size, number of part families, master
production schedule deviation and product churn. Additionally the variables that
made up product/process complexity did not load on the overall external
complexity index nor on any of the variables contained within it (see Appendix E
for factor loadings). Therefore the analysis was carried out using the
correlations.
4.2.1.3. 1 Product/process chapgp

The product/process change index measures the predictability of the
internal manufacturing environment. Table 4-1 shows the relationships between
the variables that make up the index (for a complete correlation table for the
construct validation study see Appendix F). None of the relationships between
the variables is significant using an alpha Of .10. The variables were

standardized and then summed to form the index.

Table 4-1 Correlation Matrix for Product/Process Variables

Batch Size Families MPS deviation Churn
Batch size 1.0
Families -.075 1.0
MP8 -.028 .062 1.0
Churn -.143 .110 -.147 1.0

 

In developing this index the following assumptions were made. First it

was assumed that all things being equal a smaller batch size reflects more

146

frequent product or part Changes (e.g. a more complex environment). Therefore
batch size was coded in a manner such that smaller batch sizes were indicative
Of greater complexity. Part families are inversely related to batch sizes. Thus as
the number of part families increases, the number Of unique items produced in
the plant increases. The increased demands on both structure and infrastructure
increase the complexity of an environment characterized by large numbers of
part families. Master production schedule deviation is a measure Of the ability to
predict shop floor activity (Handfield 1990). Stable master production schedules
allow easier planning, leading to reduced complexity. Finally product churn
measures the percentage Of parts that are introduced and retired each year and
is a good indication of the amount Of change that is occurring in what the plant
makes.
4.2.1.3.2 Product/process chwnd environmentaL uncerm

Table 4-2 shows the relationship between the individual product/process
variables and the environmental uncertainty index which measures the external
environment. Environmental uncertainty is measured using an existing measure
(Duncan 1972) used in prior research projects (i.e. Swamidas and Newel 1987).
The index is a summation Of 5 perceptual question related to the predictability of
the agents Of a plant’s external environment: competitors, suppliers, customers,
government and the public. In addition the table shows the relationship between
product/process complexity (the summed index) and external complexity. This

table addresses the issue of the relationship between product/process change or

147

the internal environment and the environmental uncertainty. None of the

relationships were significant at a = .10.

Table 4-2 Correlations Between EU and Product/Process Variables

 

Batch Churn Families MPS Product]
g deviation process

Environmental -.153 -.134 -.007 .045 .049
Uncertainty (p=.171) (p=.229) (p=.948) (p=.690) (p=.665)

 

 

 

 

 

4.2.1.4 Discussion Of results Of construct valrpation study

The construct validation study addressed two questions. First the premise
that the product/process variables would need to be combined to form an index
was supported. There was no factor or set of factors that underlies the variables
used to address the product/process or internal complexity measure.

This result is not surprising. The individual product/process measures
have been addressed both in prior research and in the pre-diSsertation case
studies as possible drivers of skill choices. However, there was no indication in
the literature that these variables would co-vary. For instance, a company may
have a large number of product families but never make changes in the product
(little churn). This stability leads to larger batch sizes in an effort to gain
efficiencies and to reduce set-ups costs (and additionally to simplify shop floor
control). In the aforementioned example one product/process variable, families,
indicates a complex environment, while the others do not. Another company
may have only one or two families, but constant changes in these products. For

instance an injection mold maker may make only 1 or 2 types of molds (few

148

families), however each mold is unique and is produced in batches of one - using
a MP8 that is always changing because of large numbers of engineering
changes. In this example the number of families is fairly low, however churn
within the family is high (each part needs to be designed for the customer), batch
sizes are low and shop floor control is complex.

Because the product/process variables do not co-vary, the internal
environment was measured using a summed index Of the standardized
measures. This index is a composite Of variables that taken together are an
indication of the various elements that help to contribute to a complicated and/or
unpredictable shop floor.

The second area of interest was the relationship between the internal and
external environments. The data indicates that there is no relationship between
the complexity Of the external environment and the complexity of the internal
environment. This conclusion adds validity to the notion of using two separate
measures of the environment rather than a single external measure. This result
contradicts much of the recent organizational work (i.e. Covin and Slevin 1989)
which posits that structure is driven by the external environment. Instead, the
findings seem to indicate that manufacturing structure and infrastructure are a
result of the distinct manufacturing environment rather than the external
environment where a plant’s products and services compete.

The lack Of a relationship between the internal and external environments
is a key finding for the construct validation stage Of the dissertation. If existing

research which posits that the internal and external environments were related is

149

true, than most manufacturing decisions should be in response to the external
environment. A primary assumption underlying this belief is that the external
environment drives internal choices. However, the analysis indicates that
manufacturing decisions should be the result Of the manufacturing environment,
rather than the external environment. This finding suggests that manufacturing
decisions predicated on external conditions rather than on the needs of the
internal environment could lead to lower performance. This implication does not
suggest that manufacturing strategy need not support business strategy, but
rather that manufacturing structure and infrastructure decisions should be based
on the key manufacturing task(s) which support the business strategies (Skinner
1969,1996)

4.2.1.5 Possible sources of measurement error in construct validation data

Although every effort has been made to ensure that the construct
validation sample is representative Of the population Of manufacturing firms in
the continental United States, there are a few potential sources Of error that may
render some Of the conclusions suspect.

First, plant size and the use of a single respondent may bias results. In
addition to these two potential sources of error there is a potential measurement
problem. The internal environment was addressed by summing the standardized
responses to a number of objective questions, while the environmental
uncertainty index was composed of a number Of perceptual items that are rated
using Likert scales. It is possible that the lack of a relationship between the

lntemal and external environments is a statistical artifact related to how the

150

indexes were addressed rather than an actual difference. TO address this
possibility the structured interview format used for the primary data collection
contained Objective measures of the external environment as well as the
proposed perceptual measure.
4.2L Construct validation study and remaining analysis

The construct validation study results were used in the subsequent
analysis in a number of ways. First the results support the use Of two separate
indices to measure internal and external environments, rather than a single
measure and or the use Of specific factors.

Second, the study is a cross section of a vast variety Of plants. Therefore
the product/process variables from the main sample were standardized using the
data collected from the construct validation study. The primary data collection
effort is purposely limited by process technology and by industry. In limiting the
sample to users of FMS (who are relatively rare (Handfield and Pagell 1996)),
the population is already small. Every effort was made to ensure that multiple
representatives of an industry were in the sample to allow comparisons between
similar plants. Therefore, standardizing the primary sample with the means and
standard deviations generated by their sample may cause serious problems of
generalizability.

The sample used for the primary data collection effort is not generalizable
tO the realm of manufacturing in general and the product/process measures that
would be generated using the means and standard deviations from this sample

would not be comparable to product/process measures generated by other

151

studies. However, the use of the distributions from the construct validation study
allowed the primary sample variables tO be placed along a distribution that was
more likely to represent all manufacturing plants. Hopefully, this approach
allowed for easier comparisons to future research that may have different
constraints in terms Of industry and or process technology.

The final set of results surround a specific product/process variable,
master production schedule deviation. This was the only variable in the study
that had a significant (p = .07) relationship with a perceptual performance
variable that was also collected. This finding may indicate that some or all of the
product/process variables are important in their own right and therefore they
were tested individually as well as collectively in the product/process index.
Because of the low power of the small sample such tests were not conclusive,
yet they provided interesting insights for future research.

4.3 Primary sample: preliva analys_i§

The primary data collection was done using structured interviews in a field
setting. Thirty plants in 8 industries (see Table 4-3) were visited between March
1996 and January 1997. All but two Of the industries had multiple
representatives and there was at least one FMS user in 5 Of the 8 industries.
The following sections describe how the sample was selected, potential biases
in the sample due to the selection process, comparisons between the primary

sample and the construct validation sample, and outliers.

152

Table 4-3 Sample Breakout by Industry

Auto Parts

Diesel Engines
Office Furniture
Stamping Dies
Molds

Machine TOOls
Consumer Goods

Construction Equipment
—_-—

4. 3. 1 Sample selection

 

 

 

 

 

 

AONOO—tmat

 

A—sooooormcnor

 

 

 

The population of interest for the study was North American plants using
FMS and/or CNC equipment. Although a random sample drawn from the
population Of all North American CNC and FMS users would have been the most
representative sample possible (Cook and Campbell 1979) this was not feasible
for a number of reasons. 5

First, a truly random sample would have required resources far beyond
the scope Of this dissertation. Second, determining the technologies used by
specific companies is difficult. Third, even if all Of the users Of FMS and CNC
could have been identified, the number Of FMS users is very small (Handfield
and Pagell 1995). Thus Obtaining a truly representative sample of this select
group Of plants would have been difficult. For these reasons a pure convenience
sample was used. Initially, local users Of FMS were compiled from a number Of
sources including recent research (i.e. Kaku 1994) and industry contacts Of the

various faculty at Michigan State University and other institutions.

153

Once an initial list Of FMS users was identified, these companies as well
as others in their industry were contacted. If the company was using either FMS
or CNC technology and was willing to participate in the study, they became a
participant. In the end some of the companies that Were identified as FMS users
did not participate in the research or turned out tO be defining FMS in a different
manner than this research (i.e. one company was using what they called FMS-
Flexible Machine Centers - which were CNC machines with a carousel for
multiple pallets). Therefore some industries are not represented by FMS users.

Each company is represented by a single plant (with a few exceptions). In
some cases, multiple plants from a company were included in the sample if they
differed on one or more Of the key criteria in the study. For example there are
two plants from one of the diesel engine manufacturers, but one uses FMS to
make parts for high volume engines while the other uses CNC equipment to
make similar parts for lower volume engines. In another instance two plants
from different business units Of the same company are represented. In this case,
both plants employ FMS to produce different products, but one is unionized and
the other is non-union. In the sample of 30 installations (one installation per
plant), 22 different companies are represented.

The plants were located in 4 mid-western states: Michigan, Ohio, Indiana
and Illinois.

4.3.2 Potential biases in the primary sample
The use of a convenience sample has several potential sources Of bias,

including limited industry coverage, limited geographic coverage, and selection

154

bias. These biases may make it difficult to generalize beyond the sample. There
are only 8 industries covered in the sample. The companies were classified by
industry based on their primary products and how they defined their production
outputs.

This limited industry coverage makes it impossible to generalize to all
users Of AMT, especially companies who are using AMTs in industries such as
electronics or aviation which are not represented. However, for users of
manufacturing based AMTS (see Boyer, Ward and Leong 1996 for a compete
discussion Of all types Of AMT) the sample provides broad coverage in terms Of
the ranges of volumes, types Of parts (size and use), materials (not all
companies are machining metal) and order winners (Hill 1994). Additionally the
only major user of FMS not represented is the aviation industry (i.e. Kaku 1994),
so the sample of FMS users may indeed by a good representation of the
population Of FMS users.

The second source of bias in the sample is the geographical limitations
imposed by the resources of the researcher. All Of the plants that were visited
were within 400 miles of Lansing, Michigan. Therefore, it may be difficult to
generalize the results (especially when it comes to managerial discretion) to
other regions.

The final source of bias is selection bias. The literature indicates that
AMT failure rates are very high (i.e. Snell and Dean 1991). However, all of the
plants in the primary sample describe their AMT as at least somewhat

successful. And the majority considered the technology very successful. There

155

are many possible explanations for this outcome. First, companies who failed or
were in the process of failing may have been more likely to say no to
participating in the study. Second, failures may be much more likely earlier in
the life cycle of a technology. In other words the large number of failures
reported in earlier literature were due to limited travel down the learning curve by
both users and producers of the technology. As CNC and FMS mature failure
may become much less likely. Third it is possible that earlier research
overstated the number of failures. However, Snell and Dean (1991) report failure
rates Of over 40%, suggesting that even if the researchers were wildly inaccurate
there should be a few failures in the primary sample of 30. Since these failures
are not in evidence it is likely that the primary sample suffers from a restricted
range, which will influence the interpretation and generalizability Of the results.
The three main sources of bias that are introduced by the selection
methOd are the limited coverage of industries, limited geographic scope, and
selection bias. The first two issues are important and may limit findings.
However, these effects may be mitigated by the large amount of variance in
terms of products and processes represented, and the fairly good coverage of
FMS users by industry. The aforementioned limitations are due to the design Of
the research and were expected. The very good possibility of selection bias was
not part of the research design and will have to be addressed in both the
analysis and the discussion Of results. In addition to these three sources of bias
the primary sample may differ in other unexpected ways from the general sample

of manufactures as represented by the construct validation sample. In order to

156

determine other sample related limitations, the primary sample was compared to
the construct validation sample. These analysis are discussed in the following
section.

4.3.3 Comparison of Primary sample to Construct Validation Sample.

In order to test for differences between the primary sample and the
construct validation sample, two types of tests were performed. For variables
that were normally, or nearly normally distributed t-tests were carried out (See
Appendix G) for means, standard deviations, kurtosis and skewness Of the
construct validation sample). These tests sought to identify statistical differences
between the means Of the primary sample and the construct validation sample
for each specific variable. Non- normally distributed variables were tested using
a non-parametric test (Kruskal Wallis) and a t-test was performed to identify
differences between the means of the primary sample and the construct
validation sample. These tests were performed to determine if and or how the
primary sample differed from the construct validation sample ( with the
assumption that the construct validation sample is a better representation Of
manufacturing firms in general). Such differences might indicate other limitations
to the research, yet may also be useful for explaining other results Of the
research project.
4.3.3.1 Product/process change vafiaflap

Much Of the research on manufacturing flexibility (i.e. Gerwin 1993) posits
that firms attempt to become more flexible in response to market needs.

Increased flexibility can mean the ability to make more unique items with no

157

change in cost structure, to change production volumes without incurring major
costs, to easily change the mix of parts produced and to make other changes in
the system without incurring major costs. Because AMTS are one Of the major
ways that manufacturing firms can become more fleXible, it stands to reason that
a sample Of companies using CNC and or FMS should exhibit some evidence of
increased flexibility. Indicators might be increased part families, more product
churn (ease of new product introduction) and smaller batch sizes. Therefore, we
expected some differences between the construct validation sample of
manufacturers in general and the primary sample (which is limited to companies
that have invested in a technology that, in theory, gives them increased
flexibility).
4.3.3.1.1 Batch size

The distribution of batch size (like all Of the individual variables that make
up the product/process index) is highly skewed, with a very long right tail.
Because the variable does not exhibit signs of normality the differences between
the construct validation and primary samples was tested two ways; using t-tests
(that assume normality) and a non-parametric test, Kruskal-Wallis (KW test).
The KW test is similar to the t-test without the assumption Of normality (WIIkinson
1990 (Systat manual)).

The results for the two tests are reported in Table 4.-4. The t-test
indicates that there is no significant difference between the means of the two
samples. However, the KW test indicates that the two groups are different. In

examining the batch size means, it is evident that the mean batch size of the

158

primary sample is much higher than the mean batch size of the construct
validation sample. This may be related to the fact that the primary sample is
comprised mainly Of plants who are either in non-repetitive industries (die and
mold makers, machine tool makers) or companies Who are making very large
batches (i.e. auto parts). Therefore the distribution of batch size is bi-modal with
most installations at one extreme or the other. Because Of the t-tests
assumption of normality, the bi-modal distribution is ignored by the test.
However, the distribution free KW test finds a significant difference between the

primary and construct validation samples.

Table 4-4 Comparison of Batch Sizes

 

T-Test Primary Sample Mean: 19,023 Construct validation Mean:
10,31 1

 

T statistic = -.864 Probability = .393

U statistic = 848.5 Probability = .012

 

 

 

The results indicate that the primary sample is probably not representative
Of manufacturers in general, on this specific variable (batch size). The flexibility
literature, as well as the literature on FMS (i.e. Nemetz and Fry 1988) would
suggest that the primary sample should have much lower batch sizes, when in
aggregate this is clearly not the case. This may mean that the primary sample
differs both from the general population of manufactures as well as users Of AMT
on this variable. However it is more likely that the shape Of the distribution has

lead to the counter-intuitive results. Half Of the sample has an average batch

 

159

size under 10, however there are also a number Of companies who Operate their
AMT in a near continuos mode and have batch sizes in the hundreds of
thousands. The average batch size suggest that all of the companies in the
sample have very large batch sizes, when in reality a small (but significant)
number Of installations have seriously skewed the distribution. The end result,
on this variable, is that it is likely that the companies in this sample generally
represent the extremes in that they are either producing in batches Of less than
ten or more than a thousand, with few companies in the middle of the
distribution.
4.3.3.1.2 Chum

Churn is the combination Of new product introduction and product
retirements (in percentages) and measures the amount Of change in products
produced in a year. The flexibility literature suggests that being able to introduce
new products on existing systems is a primary benefit Of being flexible (i.e.
GeMin 1993). Therefore users of AMT in general and FMS specifically should
be better suited to large numbers Of product ] part introductions and or the
running Of parts that will never be run again.

Churn does not exhibit the properties of a normally distributed variable so
once more both t-tests and the KW test were performed. The results (Table 4-5)
Of the t-test indicate a significant difference between the two samples, with the
primary data sample having significantly more churn than the general sample of
manufacturing firms. However, the KW test indicates that there are not

significant differences between the primary and construct validation samples.

160

Table 4-5 Comparison of Churn

 

l T-Test Primary Sample Mean: 56.2 Construct validation Mean: 33.6

 

T statistic = -1.657 Probability = .105

KW U statistic = 1404 Probability = .253

The results, while not conclusive, seem to indicate that the primary

 

 

 

sample is composed Of companies who, generally introduce and retire more
parts] products in a given year than the sample of manufactures in general.
However, the results Of the KW test (which may be more appropriate) indicate
that the two samples are not statistically different. The end result is that
although the difference is in the direction that the flexibility and AMT literature
would predict, there is not ovenrvhelming evidence that the firms in the primary
sample truly differ from those in the construct validation sample, on this variable.
4.3.3.1.3 Families

The number Of part families indicates the number of unique groups of
parts] products run on the manufacturing system. The variable “Families” does
not exhibit the properties of a normally distributed variable and is tested using
both t-tests and the KW test. An additional complication for this variable is
introduced by a number of companies in the sample who claimed that every part]
product that they produced was truly unique and not part of a family. These
companies were generally in the die or mold making business. If after the

structured interview the plant manager was adamant that they had no families,

161

and that any learning in the system was process based not product based, this

variable was coded as the number Of unique parts ] products the facility made in

a year.

The results for the Families variable are very Similar to the results for

Churn (Table 4-6). The t-test indicates that the primary sample companies do

indeed make many more part families than the sample of manufactures in

general. However, the results of the KW test indicate that the samples are

statistically not different.

Table 4-6 Comparison of Families

 

T-Test

Primary Sample Mean: 1108 Construct validation Mean: 21.5

 

 

T statistic = -2.95 Probability = .006

 

 

U statistic = 1376 Probability = .334

 

The results, while not conclusive, seem to indicate that the primary

sample is composed of companies who, in general have more part / product

families than the sample Of manufactures in general. However, the results of the

KW test indicate that the two samples are not statistically different. The end

result is that although the difference is in the direction that the flexibility and AMT

literature predicts, there is not overwhelming evidence that the firms in the

primary sample truly differ from those in the construct validation sample, for this

variable.

162

4. 3. 3. 1.4 Master production schedule deviation

Master production schedule deviation is used to measure the amount Of
unplanned change on the shop floor. AMTS, because Of their computer
controlled systems can be linked to planning and control systems (to form a CIM)
system. However, there is no guarantee that a company that has invested in a
manufacturing based AMT will also make investments in Administrative AMT
such as Material Resource Planning ( Boyer et al. 1996). Therefore there is no
suggestion in the literature that users of AMT will necessarily have better
scheduling systems. In fact, a system that is designed to allow increases in
parts with decreases in batch sizes may actually be harder to predict than a less
flexible system.

MPS deviation is not normally distributed and the differences between the
primary sample and the construct validation sample were conducted using both
the t-test and the KW test (Table 4-7). The results of both tests indicate that the
primary sample and the construct validation sample do not differ statistically on

this variable.

Table 4-7 Comparison of MP8 Deviation

 

T-Test Primary Sample Mean: 11.9 Construct validation Mean: 14.4

 

T statistic = 1.09 Probability = .279

 

 

 

 

163

4.3.3.1.5 Product/process champ

Product/process change is the index used to assess manufacturing
environments. This index is comprised of the standardized values for batch size,
families, churn and MP8 deviation. The literature in flexibility and AMT suggests
that the use Of these technologies allows a company to produce more parts in
smaller batches with no cost penalty (i.e. GenIvin 1982; Nemetz and Fry 1988).
However, there is also literature (i.e. Jaikumar 1986) that suggests that many
AMT users do not take advantage of these capabilities. The index does exhibit
normality and was only tested using a t-test.

The results Of the t-test (Table 4-8) indicate that there is no significant
difference between the primary sample and the construct validation sample.
Therefore installations in the primary sample are statistically similar to the
sample Of manufactures in general, when it comes tO the level of predictability on

the shop floor.

Table 4-8 Comparison of Productlprocess Change

 

T-Test Primary Sample Mean: .467 Construct validation Mean: 0.00

 

T statistic = -.678 Probability = .502

 

 

 

 

4. 3. 3. 1.5 Environmental uncertaiaty
Environmental uncertainty is an index that addresses the level Of
predictability in the plant’s external environment which is composed of the plant’s

customers, competitors, regulatory bodies, the public and the other plants which

164

compete for the same raw materials. The literature on AMTS suggests that a
primary reason that companies adopt these technologies is to respond to the
uncertainty created by fragmented, global markets populated by customers who
expect products to be tailored to their specific needs (i.e. Handfield and Pagell
1995). This implies that users of AMT would most likely have more uncertainty
in their markets, not less.

All Of the perceptual items that made up the external complexity index are
normally distributed, as is the overall index. In addition t-tests on the individual
items all have the same result; namely that on the individual items and the
overall index the primary sample respondents perceived their environment to be
less uncertain than the construct validation sample respondents(T able 4-9).
Therefore the managers in the primary sample perceive their environments to be
more certain than the general sample of manufactures represented by the
construct validation sample.

4. 3.4 Discussion Of similaritiesgnd differences between the primary sample and
the construct validation sample.

The comparisons between the primary sample and the construct
validation sample were performed to determine if there were significant
differences between the two samples that might limit generalizability of the
results and or explain results that can not be explained using existing theory.
There are two key conclusions from these comparisons. First, the internal
environments of the primary sample and the construct validation sample are not

statistically different. Secondly, the external environments of the primary

165

Table 4-9 External Complexity: Individual Items and Summed Index

How predictable are the
actual users of your
products?

How predictable are the
competitors for your
supply Of raw materials?

How predictable are the
competitors for your
customers?

How predictable is
government regulation
controlling your industry?

How predictable are the
publics political views
and attitudes towards

your industry?

External Complexity
(summed index)

 

Primary Sample Mean:
2.67

Construct validation
Mean: 3.27

 

T statistic = 1.91

Probability = .061

 

Primary Sample Mean:
2.38

Construct validation
Mean: 2.84

 

T statistic = 1.7

Probability = .095

 

Primary Sample Mean:
2.58

Construct validation
Mean: 3.38

 

T statistic = 2.39

Probability = .021

 

Primary Sample Mean:
2.73

Construct validation
Mean: 3.38

 

T statistic = 1.82

Probability = .077

 

Primary Sample Mean:
1.73

Construct validation
Mean: 3.28

 

T statistic = 6.92

Probability = .000

 

Primary Sample Mean:
12.1

Construct validation
Mean: 16.13

 

 

T statistic = 5.066

 

Probability = .000

 

sample respondents differ statistically from the external environments Of the

construct validation respondents.

The first conclusion indicates that in general the installations investigated

for this study have overall levels of uncertainty on the shop floor that are not

statistically different from the levels of uncertainty on the floors Of manufactures

in general. However, some Of these conclusions are based on the results of a

non-parametric test which differs from those Of the t-test. Because t-tests are

robust to violations of the assumptions that underlie them (Kerlinger 1986),

 

166

discounting the results of the test outright may be erroneous. Therefore on
some Of the variables that make up the product/process index there are some
potential differences between the primary sample and the construct validation
sample.

The larger batch sizes exhibited by the primary sample are believed to be
a result Of the bi-modal distribution, in a sample that is limited by industry. The
differences in the number Of families and product Churn are in the expected
directions. In other words, the literature on AMT in general and FMS specifically
suggests that users of these technologies have the ability to not only run a
greater number of part families but to change the parts within the families more
frequently. However, some researchers (i.e. Jaikumar 1986) have noted that
users of AMT do not necessarily take advantage Of all of the capabilities Of the
equipment . Therefore one could conclude that the lack of conclusive statistical
evidence is a result of companies not taking advantage Of the equipment’s
abilities rather than a result Of industry biases. Our conclusion is also that on the
overall measure of the manufacturing environment (product/process change) the
two samples are not statistically different and that results from the study that
relate to the level of product/process change are generalizable to manufacturers
in general.

The second major conclusion from these comparisons is that the external
environment of AMT users is more certain than the environment for
manufactures at large. This result is counter-intuitive. In general it is assumed

that the increased flexibility of AMT is installed at least somewhat as a response

167

tO increased uncertainty in the external environment (i.e. Gerwin 1982). This is
especially true for users of FMS (i.e. Mansfield 1993a; 1993b). However, the
reverse seems to be true; namely that users of AMT perceive their environments
as more certain. This result could Obviously be a result Of industry bias, or other
possible reasons explained below.

The first possible explanation for this result is that some Of the plants in
the sample experienced limited external environments. Some Of the plants were
supplying a majority of their output (or all Of their output in a few cases) tO a
single customer, who was Often a division of the same company. Additionally,
many of the plants were in industries where there was a limited number of
competitors for customers. However, the index is composed Of items that
assess more than the number of customers served or the number of competitors
in the market. It is also important to note that the differences between the
samples were statistically significant for all of the individual items as well as the
overall index.

A second explanation may be a temporal one. According to the literature
(i.e. GenIvin 1993) companies adopt AMT to deal with environmental uncertainty.
However, once the technology is installed, the increased manufacturing flexibility
enables the plant to manage their uncertainty - and perhaps even reduce
uncertainty. This may be especially true for companies who were earlier
adopters.

The perceptual nature of the measure is the third possible explanation for

this counter-intuitive result. Management who has installed a technology that

168

could reduce uncertainty, may change their perceptions of the environment,
regardless of the actual Change in uncertainty.

Any or all of these explanations are possible. However it is difficult to
come to any firm conclusions because of the number of confounds for each
explanation and or combination of explanations. However, this result is
important for two reasons. First is the issue Of generalizability. Results that deal
with the external environment will not be generalizable to the population Of
manufacturers at large but will instead be limited to companies who view their
external environments as more predictable. Secondly, if the result are not due to
sampling bias, industry differences, or a temporal explanation, an important
question must be raised: why did companies with lower levels of uncertainty than
the population of manufacturers at large adopt a technology that should help
them deal with uncertainty? This question leads to an important second
question: does the proposed index of environmental uncertainty measure what it
is intended to? If the measure accurately assess the perceptions of the external
environment, should it be used to predict operator skill levels? Because of the
uncertainty raised by this issue the external environment will also be examined
using some of the qualitative data that was gathered in the study.

4.3.5 Detection of outliers in the primapy sample.

Identifying outliers in a sample Of 30 creates a dilemma. Any data that is
removed will lower the power tO detect effects, when power is already at a
premium. However, data that is truly extreme also has the potential to seriously

bias results. Every attempt was made to maintain the sample, and data points

169

were only removed if they were significantly outside the ranges that the construct
validation sample suggested were applicable to manufacturers in general.

Detection Of outliers was accomplished by comparing the ranges Of the
construct validation and primary samples. For the primary indices (product/
process and environmental uncertainty) the ranges are very similar for both
samples. The ranges are wider for the construct validation sample for both
indices (see Appendixes 7 and 8). All Of the primary data fits within the range for
environmental uncertainty generated by the construct validation effort.

For the product/process variables there are a few data points that are
below the minimum generated from the construct validation sample, but the
differences are minor (minimum product/process index for primary sample is -
7.95 while the minimum for the construct validation effort was -7.5). One major
difference between the two samples is the range of product families, which is
much larger for the primary sample (for reasons discussed in section 4.3.3.1.3).

The only index or variable that can not be compared to the construct
validation data is total preparation time (Skill). The responses for this variable
ranged from 1 week to 342 weeks. The mean response was 120 weeks with a
standard deviation Of 105.9. All responses are within 2.5 standard deviations Of
the mean for this variable, which displays normal properties (see appendix H).
NO responses on this variable were considered outliers.

Based on comparisons to the construct validation data and the distribution
Of skill there appear to be no outliers in the primary data set. Therefore no data

points were removed before the subsequent analysis.

170

4.4 Analysis of primary data set:

The following section provides details on the various statistical and
qualitative tools used tO test the 4 hypotheses proposed in chapter 3. Section
4.4.1 discusses the correlation matrix for all the items and indexes used in the
analysis. Section 4.4.2 presents correlations, visual tools and regression
analysis used in testing H1A (the relationship between skill level and adoption
success). Section 4.4.3 presents the correlations and visual tools used to test
H1 8 (the relationship between appropriate choice and adoption success).
Section 4.4.4 discusses the regression analysis and visual tools used to address
hypotheses H2-H4 regarding the drivers of skill choice.

4.4.1 Correlation matrix primrv 3%

Appendix I is a complete correlation matrix for all variables used to test
H1-H4. The appendix has the correlations and probabilities for all of the indices
as well as the items that make up the indices. While all Of the relationships are
potentially important, two sets of correlations are of primary interest. First, the
relationships between the variables entered into the regression model used to
test hypotheses H2-H4 are examined for evidence Of multi-collinearity. The
second area Of interest are the significant relationships between different
variables and measures Of skills and or success.

Afifi and Clark (1990) note that when the independent variables being
used in regression are highly intercorrelated, the regression coefficients are
unstable and interpretation is therefore tenuous. A simple test of multi-

collinearity involves examining the correlations between the various independent

171

variables. Table 4-10 shows the relationships between the variables that are
hypothesized to drive the choice Of skills (H2-H4). None Of the correlations have
an absolute value greater than .2 and none of the relationships are significant.
Therefore multi-collinearity should not be an issue When performing subsequent

analysis.

Table 4-10 Relationships Between Variables Hypothesized to Drive Choice

 

 

 

Productlprocess External Managerial

Complexity Discretion
Productlprocess I 1.0 |
External | .065 ( p = .732) 1.0 |
Managerial I -.207 ( p = .273) .091 ( p = .631) 1.0 j

 

 

 

The second area of interest is the relationships between individual items
and measures of success and skills. These relationships may suggest issues
that may require further investigation either in this work or subsequent research.
Among the product/process variables (see Appendix I) churn has a significant
and positive relationship with skills. In addition both batch size and product
families have fairly strong relationships with skills. These results are not
surprising considering the proposed relationship between the overall index and
skills. However, the results may lend support to a model where the individual
product/process items drive skills. The only other item that has a relationship
with skills is Q16 (predictability of customers) from the environmental uncertainty
index. Because the rest Of the items are not Significantly related to skills there is

little support for using the various external complexity items individually in a

172

regression. However, this result may suggest substituting a more qualitative
measure of external environment in subsequent analysis. This measure could
be derived from the portionof the structured interview format that dealt with
number Of customers, number of markets served and an Open ended discussion
Of the plant’s overall external environment.

The other set of relationships that were Of interest were those of individual
items and indexes with adoption success. The literature review notes there are
any number Of issues that relate tO adoption success, beyond the skill level of
the workforce. Some of the relationships with success may reinforce earlier
findings and or point out other areas in the adoption of technology literature that
need to be explored in the future.

Success (measured as goal achievement) was Significantly related to the
environmental uncertainty (p = .035) index. The relationship is negative
indicating that as complexity increases success with advanced technologies
decreases. In addition two Of the items that make up the index Q16
(predictability of customers) and Q17 (government regulation) were also
significantly related to success. Because environmental uncertainty is also
related to skills there may be a more complicated relationship between
environmental uncertainty, skills and success than has been posited in this
dissertation, which can be explored in future research.

The only other item that is significantly correlated with success is the
amount of churn in products, which has a positive relationship with success.

This provides some counter-intuitive evidence that the users Of AMT are more

173

successful when they have more new product introductions and or product
retirements. This finding may reinforce the notion that the individual items that
make up the product/process index have some validity being tested on their own.
In addition it is possible that the relationship between skill level Of the workforce,
product/process complexity and the success Of the installation is not as simple
as proposed in this work.
Key findings from the correlations Of the variables are:
no multi-collinearity among the variables to be used in testing H2-H4
the possibility that the items being used to assess product/process
change may need to be tested individually as well as collectively
c that the relationship between environmental uncertainty, skills and
success may not be the simple one proposed by this work.
The product/process variables will therefore be entered into the regression on
the drivers of skill individually as a secondary test (see Section 4.5). Note that
with a sample Of 30 there is not enough statistical power to do more than
compare multiple R squares between the regression with the index and the
regression with the individual variables. The potentially more complex
relationship between the drivers of skills and adoption success are beyond the
scope of this dissertation but may be addressed as a future research project.
4. 4. L Testing H 1A: the relationship between skills and adoption success.
A central question driving this research is the relationship between skill
level and adoption success. The literature review provided numerous examples
(i.e. Nemetz and Fry 1988, Snell and Dean 1992) of research that concluded that

a highly skilled workforce was required for adoption success. However, there

was an equally large body Of work that showed that not all companies chose

174

high skills (i.e. Spenner 1988) and that high skills are not necessary for adoption
success (i.e. Adler 1991). Before tests Of the drivers of skill Choices and more
importantly the appropriateness of those choices can be made, we must first
assess the relationship between skills and adoption Success. H1A states: The
skill level Of the workforce is positively related tO the success Of the adoption of
CNC and or FMS. This hypothesis was addressed in three ways: correlation Of
skill level and success, a scatterplot of the same relationship, and finally
regression tO address the possibility of a curvilinear relationship between skill
level and adoption success that was suggested by the scatterplot.
4.4.2. 1 Cone/ation Of skill level and adoption success

The correlation between skill level and adoption success is -.077 which
has a probability of p = .685. The relationship between the total preparation
time of an average worker to perform theirjob at an acceptable level is unrelated
to the success of the AMT adoption, in this sample Of AMT adopters. However,
the restricted range of the goal achievement measure (the absence Of failed
installations in the sample) indicates that the correlation is likely to be attenuated
(Cohen and Cohen 1989) and hence understate the relationship between skills
and adoption success. Because this relationship is the key to any future
conclusions it is imperative that other tests besides correlation be used,
especially with a small sample and restricted range in the DV.
4.4.2.2 Scattegplot Of skill level and adoption success

Many authors (i.e. Miles and Hubbennan 1994; Hartwig and Dearing

1979) suggest that summary statistics do not tell a complete story. Therefore

175

researchers should also examine pictures of the data. An especially useful
picture to examine the relationship between two variables is the scatterplot.
Figure 4-1 is a scatterplot of the relationship between the skill level Of the

workforce and adoption success measured as goal achievement.

 

 

 

 

 

Figure 4-1
Skill Level and Adoption Success
1
0.9 O
0 8 1. .9 . ‘0 o. 9
' ,g’ ‘ . O .0
.. 0.7 -
E ¢ 9 o. .
0.6 ~
g 9
IE 0.5 »
0
f 0.4
I!
O
0 0.3
0.2 _
0.1
0 ‘ :
0 50 100 150 200 250 300 350
Skills - Total Preparation Time In Weeks

 

 

 

The scatterplot adds credence to the conclusion that there is no evidence
of a linear relationship between skill level and goal achievement. A number of
low skilled firms have achieved high levels of goal achievement. In fact, the
highest achieving companies were those with some of the lowest skill levels! The
scatterplot suggests that both high and low skill installations are successful, but it
does not rule out a relationship between skill level and success. The scatterplot
suggests that skill level may have a curvilinear relationship to success, where

companies who are at either end of the distribution Of skills are successful, while

176

mid-levels of skill are the least successful. Regression with a polynomial term
was used to address the possibility of a curvilinear relationship.
4.4.2.3 Regression Of skill level on success using a polynomial term

Affi and Clark (1990) note that when a curvilinear relationship is indicated

the appropriate regression equation is:

2

Y=A+BlX+BZX

For this specific relationship the X term is skill level. The non-significant results

Of the regression are reported in Table 4-11.

Table 4-11 Polynomial Regression

M-

 

 

Constant .783 30.074 .000

Total Preparation .000 .228 .821
Time

Total Preparation -.000 -.357 ~ .724
Time Squared

 

 

 

 

 

The multiple R squared for this regression is .011. The F statistic is .145
which has a probability of .866. The regression with a polynomial term does not
Offer any support for a curvilinear relationship between skill level and adoption
success. Once more this conclusion must be tempered by noting the attenuation
effects Of the restricted range of goal achievement.
4.4.2.4 Summary of tests Of H 1A

The relationship between skills and adoption success was tested in three

ways. The correlation results indicate that there is no relationship between skill

177

level and adoption success, although these results may be partially due to the
restricted range of the DV. The scatterplot reinforces that there is no linear
relationship, but leaves open the possibility of a curvilinear relationship between
skills and adoption success. The regression provides no evidence of a
curvilinear relationship. In sum, the tests, along with the successful low skill
firms, suggest that companies can indeed be successful with varying levels Of
skills. These findings do not rule out the possibility Of a relationship between skill
level and adoption success, but they do provide evidence that firms can be
successful with low and high skills, providing enough evidence to falsify H1A to
proceed with testing H1 B.

4.4.3 Testing H1B: the relationship between apprOpriate choice apd adoption
success

A second key hypothesis in this dissertation is that successful adopters Of
AMTchoose the level Of skills for their operational employees. H1 B states:
Firms who Choose appropriate levels Of skill for the operational employees of
FMS and CNC will be more successful adopters than firms who choose
inappropriate skill levels. This hypothesis was addressed using correlations and
visual tools. The following sections detail the methods used to Operationalize to
the fit measure, the correlation between fit and success, alternative
operationalizations Of fit, and visual representations of the data.

4.4.3. 1 Operationalization of fit
Fit was assessed using the measure of appropriate skill level proposed in

Chapter 3. The measure assumes that installations where the skill level Of

178

operational employees matches an environment described by managerial

discretion, external complexity and product/process change will be more

successful than installations where skills are either higher or lower than the

environment demands. Table 3-2 (reproduced belOW) shows how fit was

assessed, given a specific environment.

Table 3-2 Appropriate Skill Level

   

 

 

 

 

   

 

 

 

 

Product ] High level Of Discretion Low Level of Discretion

Process

Change
H High skills High skills Medium skills Medium skills
M Medium skills High skills Low skills Medium skills
L Low skills Medium skills Low skills Low skills

Environmental L H L H
Uncertainty

 

 

 

The construct validation data was used to determine the categories for

prodUct/process Change and environmental uncertainty, while the primary

sample was used to determine the categories for skill. Both of the indexes were

fairly normally distributed and did not have any Obvious gaps that could indicate

different groups. Therefore, the environmental uncertainty variable was

dichotomized by finding the median value Of the index and dividing the sample at

this value (see table 4-12). The same process was followed to trichotomized the

product/process variable.

The total preparation time variable is trichotomized based on the primary

sample data. This variable did have two obvious gaps, and therefore categories

were selected based on these gaps. The end result is a nearly equal distribution

 

179

(see Table 4-12), although there are a few more high skill installations than low
or medium skill installations.

Once the categories were determined the primary sample data was
categorized and then fit was assessed. The distribUtion Of the primary sample
data is not even across the categories because the categories for environmental
uncertainty and product/process change were generated from the construct
validation data, which differs in some ways from the primary sample. The
product/process variables have fairly equal distributions across categories (see
Table 4-13) but the majority of the installations have external environments that
are low in complexity. The uneven distribution Of environmental uncertainty was
expected based on the results Of the tests of differences between the two

samples.

Table 4-12 Categorization of Variables

 

 

 

 

 

 

 

 

 

 

 

 

Number of
respondent
s in cell
Low -7.502 - -.7027

| Medium - .5996 - .3068 27 I
| High . 3809 - 8.876 27 |
External complexity | Low 6 - 16 41 I
High 17 - 25 40 |
Skills I Low 1 - 8 8 I
| Medium 32 - 112 9 I
| High 192 - 342 13 I

 

180

Table 4-13 Distribution of Variables by Category

Category m-

 

 

 

 

 

 

Productlprocess Change Low 11
I Medium 7

| High ‘ 12

External uncertainty | Low 26

I High 4

 

 

The final step of the process was to assess the actual choice made at
each installation (see table 4-14 for the number Of respondents in each cell). If
the skill level was in the cell suggested by table 3-2, the installations was coded
to indicate that management had made the hypothesized appropriate choice. If
the skill level was in any other cell, management made an “inappropriate” choice.
Appropriate choices were coded as a one and inappropriate Choices were coded

as a zero (see Appendix J)

Table 4-14 Respondents Per Cell

     

 

 

   

 

 

 

     

Product] High level Of Discretion Low Level of Discretion
Process
Change
9 2 1 0
2 0 3 2
6 1 4 0
Environmental L H L H
Uncertainty

 

 

 

 

 

 

4.4.3.2 Correlations between appmpriate skill level and success
The relationship between the fit measure and success was first addressed

through the use Of correlations. The correlation between FIT1 (the initial

181

formulation of the fit measure) and success is .311 which has a probability of
.095. This relationship is significant. More importantly, the size of the effect
(.311) is fairly large given the number Of factors that are hypothesized to impact
adoption success (see Figure 1-1). However, the restricted range Of goal
achievement must be noted in these tests as well. According to Cohen and
Cohen (1989) restricted range will attenuate the correlation indicating that the
effect for the entire population of AMT users could be even larger. However, it is
also possible that there are not large differences in installation success ,
suggesting that the correlation is statistically questionable. Therefore there is
some evidence supporting the hypothesis that installations with skill choices
matched to the three environmental attributes proposed had higher levels of
success than those installations where skills where poorly aligned. However,
this evidence must be carefully interpreted in light Of the small sample that is
limited to successful users Of the technology.

In addition to testing the original formulation Of the fit measure a few
secondary measures of fit were addressed. FIT2 tests fit based on
product/process change and managerial discretion only. This test seeks to
replicate the findings of Wood and Albanese (1995) who discovered that
choices regarding HR systems are not related to factors in the external
environment. The correlation of FIT2 and success is .383 ( p =.037), which is a
larger effect than the effect for FIT1. This result supports the findings of Wood
and Albanese. In addition, the stronger relationship between FIT2 and success

may be related to the lack of a relationship between environmental uncertainty

182

and product/process change. In cases where environmental uncertainty drives
skill choice, the installation may not be as well matched to their manufacturing
needs as installations where skill choice is based only on internal factors.

FIT3 takes the concept Of internal factors one step further and uses only
product/process change to determine fit. The relationship between FIT3 and
success is .353 (p = .056). Once again this finding seems to Offer some support
for the supposition that decisions based only on internal environments are better
than ones based on internal and external environments. The stronger
relationship between FIT2 and success suggests that controlling for the
manufacturing environment and managerial discretion provides the best results.

FIT4 addressed the issue raised by Blacker and Brown (1987) that skill
levels which are tOO low may negatively affect success while excess skills would
not. Companies who had skills at least as high as they needed, were coded as
having made an “appropriate choice”. When operationalized, this fit measure
produced 26 companies in the “appropriate” category and 4 coded as
“inappropriate”. This measure was not subsequently tested because it was not
viewed as being able to discriminate. This is evidenced by the fact that 8 Of the
12 companies who were categorized as inappropriate by FIT1 had chosen skill
levels that were too high, not too low.

The final measure, FIT5, addressed the effect of managerial discretion on
success. Categorization was done using product/process change and
environmental uncertainty. Based on the findings from the previous tests, which

indicate that the internal environment is a better predictor Of success than the

183

external environment, this fit measure was not expected to perform well, and it
did not. The correlation between FIT5 and success is .131, which is not
significant. Once more this test suggests that decisions that consider only the
external environment will not lead to as much success as those that look only at

the internal needs.

Table 4-14 Correlations for FIT and success

 

 

FIT1 FIT2 FIT3 F|T5
Correlation I .311 .383 .353 .131
Probability I .095 .037 .056 .492

 

 

 

 

4.4.3.3 Categorization Of success

The effect sizes for most Of the fit variables are fairly large given the
nature of adoption success. However, the fit measures are all binary and it is
difficult tO determine exactly how well the concepts of fit and success are
addressed using a simple correlation. Therefore a simple visual categorization
tool suggested by Kerlinger (1986) is used. The proposed relationship to be
tested is that companies who match their skill choice to their environment will
have higher levels Of success than companies who do not match skills to the
environment. If the data for success and fit are arranged in two columns, sorted
by success, the fit column should have a number of zeros (equal to the number
Of installations where the choice did not fit the environment) followed by a
number Of ones (the number Of installations where choice did match the

environment), assuming the fit measure was a perfect predictor. Because the fit

184

measures are not perfect predictors, one way to assess how well they do predict
success is to compare their predictions with what a perfect measure would have
predicted.

FIT1, the original formulation, determines that 18 companies have made
appropriate choices, while 12 have not. Therefore, when success is sorted in
ascending order a perfect measure would have 12 zeros (not appropriate)
followed by 18 ones (appropriate) in the FIT1 column. The FIT1 measure
correctly Classified 20 out of the 30 installations (see appendix K for sorted data)

based on relative levels of success.

Table 4-14 FlT1’s ability to correctly categorize success

 

 

 

 

Not Appropriate Appropriate Total
Number in Sample 12 18
Number correctly 7 13
predicted
Percentage 58.3 72.2

 

 

 

 

The same procedure was followed for the remaining three fit measures
(see appendix12 for sorted data). FIT2 classified 22 out of the 30 installations in

the correct category, for a 73% accuracy rate.

Table 4-15 FlT2’s ability to correctly categorize success

 

 

Not Appropriate Appropriate Total
Number in Sample 13 17 30 |
Number correctly 9 13 22
predicted

 

 

 

 

Percentage 69 76 I 73.33 I

 

185

F |T3 Classified 18 Of the 30 installations in the correct category for a 60%

accuracy rate.

Table 4-16 FIT3’s ability to correctly categorize success

 

 

Not Appropriate Appropriate Total
Number in Sample 14 16 30 I
Number correctly 8 10 18
predicted

 

 

 

 

Percentage 57 63

 

F IT5 classified 18 of the 30 installations in the correct category, for a 60%

accuracy rate.

Table 4-17 FlT5’s ability to correctly categorize success

 

 

Not Appropriate Appropriate Total f
Number in Sample 10 20 30 |
Number correctly 4 14 18
predicted ‘

 

 

 

 

Percentage 40 70

 

This test is rather trivial, in that it just determines what percentage of
respondents were correctly categorized, based on the relative levels Of success.
It does not discern if there is any covariance between the variables. However, if
the fit variables were unable to place an installation into the correct category
then they would be of questionable value. This categorization ability, when
combined with the fairly strong (and significant) effects for 3 of the 4 fit measures
suggests that not only is there a relationship between appropriateness of choice

and success, but that the proposed categorization variables work fairly well for

186

this sample. The FIT2 measure is both the best predictor of success, and is
most likely to correctly classify an installation.
4.4.4 Drivers of skill choices

Hypotheses 2-4 deal with the drivers of skill Choices. H2 posits that
higher levels of product/process change lead to higher levels of skill. H3 posits
that higher levels of environmental uncertainty lead to higher levels of skill. H4
posits that installations characterized by a low level of managerial discretion will
have lower skills than installations with high levels of discretion and Similar
internal and external environments. These hypotheses were tested using
moderated regression, as well as a number Of visual tools.
4.4.4. 1 Moderated regression

Managerial discretion is hypothesized to have a moderating effect. When
discretion is high, decisions regarding skill will be based on product/process
change and environmental uncertainty. However when discretion is low the
choice Of skills will be lower to account for reduced managerial flexibility in
regards to worker tasks. Hierarchical (Cohen and Cohen 1983) or moderated
regression (Stone and Hollenbeck 1989) was used to address this relationship.

The procedure was carried out in two steps. First, the main effects of
product/process change, environmental uncertainty and managerial discretion
were entered into a regression model. In the second step the interaction Of
managerial discretion with both product/process change and environmental

uncertainty was entered into the model. If a moderation effect is present the

   

187

interactions should increase the explanatory power of the model as well as
change the slopes of the regression lines (Cohen and Cohen 1983).

The results Of the first step are displayed in table 4-18. The overall model
was a very good predictor of the skill level chosen for the workforce, and was
statistically significant. In addition all three Of the hypothesized drivers Of skill
choice had statistically significant beta coefficients. This supports hypothesis

H2-H4 that all three elements have direct effects on skill choices.

Table 4-18 Step 1 of Moderated Regression

    

 

 

 

Variable Coefficient P(2 tail)
(standard
coefficient)
Constant 10.954 (.000) .238 .814
Product/process 16.687 (.549) 4.401 .000
Change
Environmental 10.421 (.348) 2.845 .009
Uncertainty .
Managerial -74.515 (-.337) -2.7 .012
Discretion

 

 

 

 

 

The R squared for the model is .616 (adjusted R squared was .572), with
an F statistic Of 13.909, which has a probability of p = .000. In addition the
relationships were all in the predicted directions. Product/process change and
environmental uncertainty were positively associated with increased skill levels,
while decreases in discretion led to lower skills. Note that managerial discretion
is a binary variable, and was coded as 1 = low discretion. Therefore decreases

in managerial flexibility were noted with a higher value for this variable. SO as

188

discretion decreases (coded as a one) skills decrease, which is what the
negative coefficient indicates.

The second step of the procedure was to enter the interactions of
managerial discretion with product/process change and environmental
uncertainty into the model. The new model has almost the same predictive
ability as the original model, however the adjusted R squared is lower due to an
increase in the number of independent variables. Table 4-19 shows the results

Of this analysis.

Table 4-19 Step 2 of Moderated Regression

 

 

 

 

 

 

Variable Coefficient P(2 tail)
(standard
coefficient)

Constant 1 1.725 (.000) .204 .840

Product/process 17.078 (.562) 3.564 .002
Change

Environmental 10.324 (.345) 2.192 * .038
Uncertainty

Managerial -74.686 (-.338) -.714 .482
Discretion

Managerial -1.272 (-.023) -.146 .885
discretion by PP
change

Managerial .012 (.001) .001 .999
Discretion by EC

 

 

 

 

 

The R squared for the model is .616 (with an adjusted R squared of
.537), which has a F statistic of 7.7.15 ( p <.000.) This model predicts skills
almost as well as the original model. However, there is nO evidence Of a
moderating effect Of managerial discretion on skills because the explanatory

abilities Of the model do not increase with the addition of the interaction terms.

189

This finding suggests that there is a direct effect for managerial discretion, but
not an interaction between managerial discretion and the other two
environmental variables. However, it is also possible that the hypothesized
relationship exists but that the small sample size creates a bias. A secondary
test Of the main effects was used to increase the validity of the results.
4.4.4.2 Visual representations of the drivers of skill choices.

Scatterplots were used to address the direct and moderating effects that
were hypothesized for the drivers of skills. The scatterplot Of product/process
change against skill level (Figure 4-2) indicates that for most installations as
product/process change increases (i.e. the manufacturing environment becomes

less predictable), skills should also increase. This plot reinforces the

 

 

 

 

 

 

Figure 4-2
Product Process Change and Skill Level

8 .

6 L ‘ O
s * o
E 4 a O .
o 2 .. . _
g 0 v O O l w 3 O
2 i? o Q) 109 150 200 250, 300 3 ,0
o. -2 . .
E
u -4
2 ‘ '

_8 . , '

Skill Level - Preparation Time in Weeks

 

relationship between skill level and product/process change that is detected

using the statistical tools. The plot is also useful for identifying installations that

 

190

do not fit the predicted and or prevalent pattern. For instance the circled
installation with a total preparation time of 110 weeks and product/process
change of -5.5 is one of a few installations that differ greatly from the predicted
pattern. This and other non-conforming points may be useful in explaining
conditions for which the relationship may not occur, and will be explored in more
detail in chapter five.

The next relationship of interest is the effect of environmental uncertainty
on skill level. The scatterplot of this relationship (figure 4-3) is not as clear. The
regression suggests that environmental uncertainty has a strong, positive
relationship with skill level. However the scatterplot does not reveal a clear

relationship between the two variables.

 

 

 

 

 

 

Figure 4-3
Environmental Uncertainty and Skill Level
20 c
18 v e
e
g 16
E 14 . °
0
g 12 _ . . e e e
a 10 L ‘. . -
g 8 3. ° 0 e e e
= 6 ._
2 e
'; 4 .
B
I“ 2 .
0 ' . .
. 0 50 100 150 .200 _ 250 300 350
i Skill Level - total Preparation Time in Weeks

 

L_ __ i- ,#__ % -¥ 4

The next 2 scatterplots address the issue of managerial discretion, which

was hypothesized to have a moderating effect on skill level. The regression

191

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 44
Skills to Product Process Change: High Discretion
8
e
6 . e
s . -
5 4 g",
5 Q ' A
3 2 .
0
§ 0 0 . , 9 ‘
E If e no 100 150. 200 250 °
8 -2 .
1: e
2
n. .4 k
-5 O 0
Skills - Total Preparation Time in Weeks
Figure 4-5
Skills to Product Process Change: Low Discretion
8
o 6 .
D
g 4 _
S
0 2 _
8
° 0 e j . . ..
o. '5 .-\ > ' ‘_ . A . ,7 '_ .
(j Skills - Total Preparation Time in Weeks

 

 

 

suggests that the relationship between managerial discretion and skill level is a
direct effect. Figures 4-4 and 4-5 are respectively the plot of product/process

change against skills for the high discretion installations, and the plot of

192

product/process change against skills for the low discretion firms. The plots
suggest that for high discretion firms a pattern similar to that for the overall a
sample exists: namely that in general as product/process change increases skills
increase. The plot for the low discretion installations shows very little variance in
the level of product/process ( with 2 extreme exceptions which are circled)
although skills do vary. These plots show some indication of different slopes for
the relationship providing some support for the proposed moderating effect.
Figures 4-6 and 4-7 are respectively the plot of environmental uncertainty

against skills for the high discretion installations. and the plot of environmental

 

 

 

 

 

Figure 4-6
Environmental Uncertainty to Skills- High Discretion
350 .
300 .
250 . e .0
o .
p 200 .- . f . . o
5 150 .
100 . 9 ex
..
o e . g
0 5 10 15 20
Environmental Uncertainty

 

 

 

uncertainty against skills for the low discretion firms. The plots do not seem to
reveal the proposed moderating relationship. Note that in these plots the x and y

axis have been switched from figure 4-3. The rotated axis seem to show the

193

 

 

 

 

 

Figure 4-7
Environmental Uncertainty to Skills - Low Discretion
200
180 . °
160
140 .
120 .
a
E 100
.I
(0 80 . e
60 .
4o . 0 e
20
0 I O 9 A t I
0 5 10 15 20
Environmental Uncertainty

 

 

 

relationship between environmental uncertainty and skills in more detail.
Although the distribution of environmental uncertainty is fairly wide. the overall
relationship does seem to be the positive one as established in the regression.

The final technique used to analyze the relationship between the various
variables in the regression was to study descriptive statistics. The means for the
high and low discretion groups are compared across skills, product/process
change and environmental uncertainty (see Table 4-20). The statistics indicate
that on environmental uncertainty the two groups are basically the same. The
low discretion installations have lower product/process change, and much lower
skills. These results provide a additional evidence that a moderating effect exists
uniquely for the product/process change variable.

To summarize the scatterplots and descriptive statistics reveal two major

results. First, they show the specific attributes of the relationships tested using

194

Table 4-20 Descriptive Statistics by Discretion

Level of discretion Environmental Product/process
Uncertainty Change

 

High 11.875 .9676 150.85

 

 

 

Low I 12.55 -.5335 58.32

 

regression. The direct effect of product/process Change on skill level is easily
discerned from the scatterplot. The relationship between environmental
uncertainty and skills is not as Obvious but when the axis are rotated (Figures 4-6
and 4-7) there does seem to be an upward sloping (i.e. positive) relationship.
The second interesting result is that there is only limited evidence of a
moderating effect for managerial discretion, and this occurs only for the
product/process change variable.
4. 4.5 Conclusions of primary data analysis

The analysis Of the data, based on the methods proposed in Chapter 3
lends at least partial support for all of the proposed hypotheses. Not only is
there evidence that skills levels are not robust predictors Of adoption success,
but there is also evidence that installations with skill levels that match their
environment have higher levels Of adoption success. Finally, the proposed
drivers Of skill all have direct effects, and predict a large amount Of variance,
especially in light Of the small sample and number Of other factors that could
drive skill choices. However, the sample is small and many Of the measures are
untested. In addition, the qualitative data collected during the structured

interviews has yet to be addressed. The following section explores some Of the

195

relationships that were tested in section 4.4 using alternative measures that are
derived from the additional data that was collected as part of the study.
4.5 Additional Analysis

Section 4.5 addresses the additional analySis that was suggested in the
previous sections. Section 4.5.1 deals with additional quantitative analysis.
Section 4.5.2 deals with comparisons between users of FMS and CNC. Section
4.5.3 addresses the individual industries for evidence of industry effects.
4.5. 1 Additional quantitative analysis

The following sections detail additional quantitative analysis that was
suggested by: the previous analysis, the literature and/or Observations made
during the collection Of the primary data. Section 4.5.1.1 addresses the issue Of
using the individual product/process change variables in the regression, that was
raised in section 4.4.1. Section 4.5.1.2 uses a secondary measure Of success
that may address some shortcomings in the goal achievement measure. Section
4.5.1.3 addresses the lack Of a relationship between internal and external
environments by substituting more objective measures for the external
environment into the analysis. Section 4.5.1.4 addresses an issue raised in the
literature review; namely how does the age Of an installation effect skills and
success? Section 4.5.1.5 addresses the issue Of plant size and skills that is also
raised in the literature review.
4.5.1.1 Using the ind_ividual product/process change items rather than the index

The correlations between some of the individual product/process change

variables and other variables Of interest that are presented in section 4.4.1

196

indicate that the product/process change items may be better predictors of skill
when used individually rather than as a single index. This issue was addressed
using a number of regression models.

First the individual variables that make up the index were regressed on
total preparation time (Skill). The results are presented in Table 4-21. The overall
effect is fairly large (.466) and the model is significant. However, only the
relationship between churn and skills is significant. Additionally the R squared for
the overall index and skills is .411 which is a similar sized effect. Finally, when
adjusted for shrinkage the R squared for the index is higher (.390) than the R

squared for the individual items (.381).

Table 4-21 Regression of Product/Process Variables on Skills

 

 

 

 

 

Variable Coefficient T P(2 tail)
(standard
coefficient)
Constant 79.402 (0.0) 2.789 .010
Batch Size 0.00 (-1.50) -.996 .329
Churn .882 (.570) 2.956 .007
Families .005 (.104) .552 .586
MP8 Deviation -.778 (-.069) -.464 .646

 

 

 

 

 

The individual items and the index seem to have similar predictive
abilities, although using the index rather than the individual items results in less

shrinkage in the R squared. Because the effect size of the index is nearly the

197

same as the effect size Of the individual items, to parsimony (and to conserve
power) the index will be used. However, for future studies there is still some
evidence that with a larger sample size the product/process change items could
be addressed individually as well as collectively.

4.5.1.2 Self assessment Of success

The measure Of goal achievement was chosen because it is site specific
and has fewer confounds than financial data. However, the primary data
collection effort raised some questions about the validity of goal achievement in
all installations. For instance in Older installations three things could have
occurred between adoption and the site visit: 1) the management has changed
and the original goals have changed or been forgotten, 2) the original goals may
have Changed with changes in manufacturing strategy, 3) the original goals may
have been unrealistic (such as a few companies desire to run untended) hence
changed to more achievable goals.

An additional problem with the goal achievement measure was trying to
quantify the overall difficulty of an installation’s goals. Easily achieved goals
could be reflected in higher achievement than difficult to achieve goals, even if
the performance of the systems was relatively similar. Therefore a secondary
measure of success was collected.

This measure, deemed self assessment, is derived in the following way.
After the discussion of overall goal achievement, managers were asked to
assess the overall performance Of the system, using their own words. At the end

Of this conversation they were asked to rate the system in terms Of not meeting,

198

meeting, or exceeding expectations. The conversation addressed any
inconsistencies between their rating and previous comments about goal
achievement. Finally the comments were rated from 1-7. A “1” indicated that the
installation meet none Of the managers expectations. A response Of “3” indicated
that the majority Of goals were not met. A response Of “5” indicated that the
majority Of goals were met. Installations were coded a “6” if all expectations were
met but none were exceeded And finally, a “7” indicated that all expectations
were met and some expectations were exceeded. This measure was used to
address two issues: skill level and fit.

Section 4.4.2 addressed the relationship between goal achievement and
success and found no relationship using correlations, regression with a
polynomial term and a scatterplot. Similar activities were repeated substituting
self assessment for goal achievement. The correlation between self assessment
and skill level Is .073 which has a probability Of p = .701. This relationship is
almost identical to the non-significant relationship between goal achievement
and skills (correlation = .077). The lack Of a relationship was also addressed
using a scatterplot to determine if there was a non-linear relationship underlying.
Figure 4-8 shows the relationship between self assessment and skill.

The secondary analysis reconfirrns the previous findings. The scatterplot
does not indicate a relationship between skill level and the respondent’s
assessment of the success Of the AMT. In addition, all but one installation meet

at least some Of their goals, and nearly half meet or exceeded all of their goals.

199

 

 

 

 

 

Figure 4-8
Skill Level and Self Assessment of Adoption Success
7 c: s c e s
6 o e e e e o e
E 5 I} 00 e e e
4
g 0
g 3
<
3.; 2 .
en
1 .
0 - . . I . ,
0 50 100 1 50 200 250 300 350
Skill Level - Total Preparation Time In Weeks

 

 

 

This is in contrast to earlier studies, such as Snell and Dean (1992), who suggest
that over half Of all installations fail. Obviously the convenience sample used for
this study makes any generalization impossible. However, for this group of 30
AMT users half Of the installations were doing as well as expected or better,
while the other half were meeting a majority of the managers goals and/or
expectations. Finally, none of the installations was characterized as a failure by
the respondents.
4.5.1.3 Additional measures of the extemal environment

The literature review suggests that there is a relationship between
environmental uncertainty and/organizational structure. The early works of
Burnes and Stalker (1961) and Lawrence and Lorsh (1967) have been
developed into a school of thought deemed environmental contingency theory

(see Bluedorn 1993 for a thorough review). The results indicate that internal and

200

external environments are not related (for both the construct validation sample
and the primary sample). These results seem to be in conflict with this well
developed stream of research. One possible explanation for the lack of a
relationship between internal and external environments is the perceptual
measure used to address the relationship, even though Bluedorn et. al (1994)
note that in general Duncan’s (1972) measure for environmental uncertainty (the
one used in this work) is the most prevalent in research used to address
environmental contingency theory. Therefore, a few other items that were part of
the structured interview format were used as substitutes for the environmental
uncertainty index.

Many authors have noted that competition is an element of the
environment that may also predict and/or affect strategy and/or structure choices
(i.e. Dean and Snell 1996). Therefore the respondents were asked to assess
how many direct competitors they had in their main market. In addition the
respondents were asked how many market segments they served from the plant.
The first item assess the number of companies the plant competes with for their
main business, while the second addresses how many potentially different
customer segments have to be served from the same facility. Table 4-22 shows
the correlations between environmental uncertainty, direct competitors, market
segments, product/process change and total preparation time.

The results displayed in Table 4-21 indicate a number of things. First, the

relationships between environmental uncertainty, segments and competitors are

201

Table 4-22 Correlations For Environmental Measures

 

 

 

 

 

Environment Custome Direct Produc Total
al r Competitor tlproce Preparation
Uncertainty Segment 3 ss Time
s , Chang
e
Environment 1 .0
al
Uncertainty
Segments -.244 (.244) 1.0
Competitors .079 (.684) .080 1.0
(.681)
Product .063 (.745) .062 .487 (.007) 1.0
Process (.748)
Total Prep. .352 (.061) .000 .372 (.047) .649 1.0
(.999) (.00)

 

 

 

 

 

 

 

all non-significant. Second, the number of segments served is not statistically
related to any of the other variables and provides less insight than the original
environmental uncertainty formulation. Third, number of direct competitors is
significantly related to both skills and product/process change.

This last set of relationships is the most interesting. The correlations may
indicate that there is a relationship between internal and external environments,
but that the relationship is driven by level of competition rather than perceived
uncertainty. The environmental uncertainty measure that was used in this
research was selected because of past use and because authors such as
Swamidass and Newell (1987) note that perceptions are more important when
dealing with choices. The number of competitors item is still somewhat
subjective (perceptual) because it requires the respondent to define both their

market and whom they see as their competitors. Therefore, there may be some

202

validity in using this measure as a substitute for environmental uncertainty. In
addition this finding provides the impetuous to test some of the objective
measure (i.e. Keats and Hitt 1984) of the firm’s environment hypothesized to
have a relationship with AMT in future research.

Because of the relationships between direct competitors and both
product/process change and skills, hypotheses H2-H4 were re-tested
substituting direct competitors for environmental uncertainty. Tables 4-23 and 4-
24 show the results of two regression models. The first model test the direct
effects of product/process change, direct competitors and managerial discretion
on total preparation time. The second model adds the interaction of product!
process change and direct competitors that is suggested by the significant

relationship between these variables.

Table 4-23 Dependent Variable: Total Preparation Time '

 

 

 

 

 

 

Variable Coefficient T P(2 tail)
(standard
coefficient)
Constant 120.705 (.000) 5.592 .000
Direct Competitors 1.371 (.162) 1.078 .291
Product/process 15.709 (.516) 3.424 .002
Change
Managerial -62.094 (-.281) -2.006 .055
discretion

 

 

Both models are good predictors of total preparation time. The first model
has an R squared of .518 (adjusted .463) and is significant at p < .000. The

second model has an R squared of .560 (adjusted .489) and is also significant at

Table 4-24 Total Preparation Time

  
 

   
 
 

  

203

 

  
 

 

 

 

 

 

 

 

Variable Coefficient P(2 tail)
(standard
coefficient)
Constant 1 16.159 (000) 5.465 .000
Direct Competitors 3.098 (.366) 1.849 .076
Prod uctlprocess 19.450 (.639) 3.817 .001
Change
Managerial -63.822 (-.289) -2.1 13 .045
discretion
Product/process * -.582 (-.346) -1.533 .138
Direct Competitors

 

 

 

p < .000. However neither model predicts skills as well as the original model
which had a R squared of .616 (adjusted .572). The results indicate that
environmental uncertainty is a better predictor of skill choices, but that the
relationship between the internal and external environments may be captured
better using the level of competition rather than the level of uncertainty. Because
this study is interested in skill choices and their drivers, environmental
uncertainty, as originally formulated, is still viewed as the proper index for
discussion of results.
4.5.1.4 Installation aqe. skill level and adoption success

The age of the installation may effect both the skill level choice and the
success of the adoption. The literature presents a mixed view of how the age of
an installation might effect skills. Zicklen (1987) posits that as a technology
matures skill level requirements will drop. However, both Kelly (1990) and
Meredith (1987) suggest that later adopters of AMT tend to use higher skills. Age

of the installation may also have an effect on goal achievement. Intuition

204

suggests that the longer a technology has been in place the more time a
company has had to learn the limits and to improve their ability to use the AMT.
The data collection effort provided at least a few examples that support the
intuition - a few of the recent adopters were the loWest performers. Because of
the potential for age of installation to effect both skills and success the
relationships were tested using correlations. The result of the analysis are

presented in Table 4-25

Table 4-25 Correlations With Age of Installation

Goal Achievement Self Assessment Skills
Age of Installation I .263 (.161) -.075 (.694) .051 (.791) |

 

 

 

 

The analysis does not indicate any relationship between age and either
skills or adoption success. The relationship between age and adoption success
is not significant and the effect size is minuscule. Therefore there is no evidence
that age leads to higher or lower skills. The relationship between age and self
assessment is also insignificant, suggesting that the differences in performance
between installations is not due to age. Finally the relationship between age of
installation and goal achievement is non-significant. However, the effect size for
this relationship between age and what is effectively performance is reasonably
large. This outcome may be related to the aforementioned problems with the
goal achievement measure. Plants that have had the technology in place for a
long period of time may have rationalized outcomes, while recent adopters may

be expecting gains that either have not accrued or will not accrue. The findings

205

do not support controlling for the age of an installation when looking at goal
achievement and/or skills, but age of the installation should not be ignored in
future research.
4.5.1.5 Size of plant

Employment or company size is often a key variable in the organizational
literature (i.e. Bluedorn 1993). However, there is no consensus as to how firm
size effects skill choices for AMT. Some authors (i.e. Kelly 1990) linked lower
skills to larger firms, while others (i.e. Dean and Snell 1991) concluded that large
firms may have higher skills. The level of analysis of the dissertation makes
comparisons with other research difficult, but it is possible to test for the impacts
of plant size on the outcomes. Employment is correlated to skills at -.308 which
has a probability of p = .098. Therefore there is evidence that larger plants tend
to have lower skills for AMT installations. However the larger plants in the
sample tend to be unionized (correlation of employment to union is .404, p =
.027), which is a major potential confound. In addition adding the employment
variable to the original model to predict skills (skills are driven by product/process
change, managerial discretion and environmental uncertainty), adds no
predictive power to the model. The R squared value does not change and the
employment variable is the only item which does not have a significant beta. It is
likely that the variance caused by employment is captured by the managerial
discretion variable. In other words the large plants in the study tend to be the
union plants which are much more likely to have low levels of discretion. The

small plants in the sample tend to be non-union and have high levels of

206

discretion. For this sample there is an effect for size, but it is not possible to say
if the effect is driven by size, the existence of the union, or the interaction of
both. The only possible conclusion is that in general large union pants in this
sample had lower skills than small non-union plants. However the large non-
union plants and the small unionized plants are not explained.

4. 5.2 Comparisons of CNC and FMS

The following sections detail comparisons between the users of CNC and
FMS. An FMS is a collection of CNC machines that are linked by automated
material handling. The linkage of multiple machines requires a much higher
degree of sophistication than is required with a stand alone machine. The
integration of an FMS is usually seen as a driver for increased skills. In addition,
FMS is generally presented as a way for companies to increase their level of
flexibility, often as a response to increased environmental uncertainty. The
combination of integration and increased levels of flexibility (which implies more
part families as well as increased levels of new product introductions) suggests
that the users of FMS should have more complicated internal and external
environments and hence higher skilled operational employees.

The sub-samples are small and the distribution of FMS and CNC is not
equal (8 FMS and 22 CNC installations). In addition some of the variables in the
sample are not normally distributed. The combination of small, unbalanced sub-
sets and non-normality means that any statistical comparisons of the sub-groups
that were performed would be highly suspect. Therefore, the comparisons

between the sub-samples were made by examining means and ranges. In

207

addition the relationships that were tested in H1 - H4 were examined using a
variety of visual tools.
4. 5.2.1 Comparisons of CNC a_nd FMS - key variables a_nd indices

Appendix L is a complete table that has the minimum, maximum and
mean for each variable, across both types of technology as well as the overall
sample. The table is used to address a few key areas that the literature
suggests make FMS different from other forms of AMT. For instance FMS and
CIM are often posited as being technological responses to environmental
uncertainty (i.e., Nemetz and Fry 1988), because these integrated technologies
give a company a way to compete on economies of scope (Goldhar and Lei
1983) or mass customization (Pine 1993). In addition, the integration of FMS
has been posited to be a driver for higher skills (i.e. Walton and Sussman 1987).
These relationships will be examined in the following sections. Section 4.5.2.1.1
examines differences in the product/process variables as well as the overall
index. Section 4.5.2.1. 2 examines the external environment. Section 4.5.2.1.3
examines skill level. Section 4.5.2.1.4 examines the size of the installations.
Finally sections 4.5.2.1.5 - 7 addresses H1-H4 for the individual technologies.

4. 5. g. 1. 1 Comparing product/process @bles for CNC and FMS.

AMT in general is posited to give a company the ability to introduce more
products, more frequently because of the negligible set-up times and ability of
the equipment to process many different types of work. The difference between
an FMS and stand alone CNC equipment is the ability to move parts from

machine to machine without human intervention. Multiple linked machines

208

theoretically mean that more processes can be performed on a part between
human interventions. In addition multiple linked machines, even if they are
identical, give a company routing options that are also accomplished without
human intervention. The reduced set-up and move times as well as multiple
routings could make it easier to introduce more parts, more frequently in an FMS
environment. However, FMS might be harder to schedule because multiple
machines need to be controlled for simultaneously. Therefore, FMS installations
could have higher product/process scores because more products could be
made in a more difficult to predict environment.

The data in Appendix L does not indicate that FMS installations have
more complicated internal environments. Instead the overall product/process
change index for FMS seems to be restricted to a fairly narrow band. For the
overall sample, product/process change scores range from 57.95 to 6.53.
However, the FMS installations range from -1.01 to .954. In other words FMS is
not being used in the most certain, or most uncertain of manufacturing
environments. In addition, the mean for FMS is -.335 which is in the middle of
the medium range of product/process change (see Table 4.4.3.1). In contrast
the CNC installations cover the entire range of product/process change values
and have a mean that is in the high category.

Examining the individual variables that make up the index helps to explain
this unexpected result. The FMS users tend to have lower batch sizes.
However, there are no FMS installations with an average batch size of one

(minimum FMS batch size is 2.8) while there are a number of CNC users who do

209

have average batch sizes of 1. CNC users are also more likely to be operating
in a continuous or near continuous mode, where a machine is set up for a single
part that is run in batches of hundreds of thousands (common for some auto
parts). Batch size mirrors the overall index with CNC users having more
vanance.

FMS installations have lower amounts of churn. Additionally, a number of
the CNC users are producing parts in lot sizes of one that will never be produced
again, giving them churn of 200 (churn is the percentage of new parts introduced
in a given year added to the percentage of parts retired- so churn of 200 means
each part is unique and not repeated), while the maximum churn for an FMS
installation is 51. Therefore the FMS users in the sample are not using the FMS
to introduce a higher percentage of new and/or custom products than the users
of stand alone CNC in the sample.

FMS and CNC users are similar for the number of part families produced
on the equipment. There are users of each type of equipment who run a single
part family (or in some cases a single part) while others run 5000 unique part
types each year. MPS deviation is also similar, indicating that for the
installations in the sample, predicting what is going to be done on the shop floor
is not impacted by the type of technology.

The key differences between FMS users and CNC users seem to be the
limited range of batch sizes, as well as the reduced product churn. These

elements combine with families and product/process change to give FMS users

210

in this study much more limited lntemal environments then the users of stand
alone CNC equipment.
4. 5. 2. 1.2 Comparing extema_I environments for CNand FMS

AMT in general and FMS in particular (Nemetz and Fry 1988) has been
posited as a response to environmental uncertainty. Section 4.3.3.1.5
discussed the finding that in general the users of AMT in the primary sample had
more certain environments than the sample of manufactures in general.
Therefore, there is some evidence that the adopters of AMT in the primary
sample were not responding to environmental uncertainty when they made the
decision to adopt CNC and/or FMS. In addition, there were no real differences in
environmental uncertainty for the users of FMS and CNC in the primary sample.
However, for the secondary measure of the environment, direct competitors, the
CNC installations on average have more competitors. Additionally the range of
competitors for the CNC users is much wider. The maximum number of
competitors for the FMS installations was 12, while the CNC installations had up
to 50. As Dean and Snell (1996) note, competition is one of the key dimensions
of the environment. Therefore, FMS users have less complex environments than
the CNC users in the sample, at least on the environmental dimension of level of
competition. In sum there is little evidence that for this sample of AMT users that
environmental uncertainty drove the choice of AMT, especially when it comes to

the adopters of FMS.

211

4._5.;1.3 Comparison of total preparation time for CNC and FMS

Authors such as Walton and Sussman (1987) have posited that FMS
requires highly skilled employees to be successful. Their argument (repeated by
many others - see Chapter 2) is that the integratiOn of FMS (multiple machines,
large tool stores, automated material handling and the like) requires a larger skill
set than simple stand alone equipment. However there is no evidence of FMS
users having higher skills. Instead, the users of FMS have much lower average
preparation times (98 weeks as compared to 128 for CNC). Additionally the
range of skills for FMS users is not as wide as the range for CNC users. Finally
there are companies using both technologies with minimal training (1 week).
4. 5. 2. 1.4 Comparisons of levels of success for FMS and CNC users

The increased integration of FMS may also make adoption more difficult
than adopting stand alone equipment. By nature FMS requires multiple
elements to function together, which often involves software and infrastructure
that is not required for a stand alone machine. Therefore it would not be
surprising if FMS users had more difficulty achieving their goals. However, there
is only a minuscule difference between average goal achievement for FMS users
and CNC users. Additionally, the ranges of values are very similar, although
both the lowest and highest achieving installations were CNC installations.

Self assessment of success also shows no difference between CNC and
FMS installations. Not only are the users of both AMTs generally successful

(with the average installations meeting the majority of their goals and/or

212

expectations) there are no differences between outcomes for the two
technologies in the sample.
4. 5. 2. 1.5 Comparison of the size of FMS a_nd CNC installations

One of the key organizational variables is size (i.e. Bluedorn 1993).
Previous analysis (section 4.5.1.5) noted that although there was a negative
relationship between size and skills, that the relationship was confounded by the
relationship between unionization and size. In other words for the overall sample
many of the large firms were also the firms with low levels of managerial
discretion. The relationship between size and discretion may be an important
determinant of the choice of skills. However, size in and of itself may be
important for helping to explain the adoption of FMS. Mansfield (19933,b) notes
that the cost of FMS may be one of the reasons that FMS has not diffused as
quickly as other innovations such as industrial robots. The cost of FMS might
prohibit small organizations from adopting it, even if an FMS might be
appropriate for their environment. The sample shows evidence that FMS
installations tend to be in much larger plants than CNC installations. The
smallest CNC installation in the sample had 10 employees, and there were 3
other CNC plants with less than 100 employees. This is in contrast to the
smallest FMS installation which is in a plant of 450 employees. In addition the
average FMS installation employed 1128 people while the average CNC
installation employed 437 people (average for the entire sample was 621). FMS

is, on average, installed in larger plants than CNC for this sample of AMT users.

213

4. 5. 2. 1.5 H 1A: Relationslyp between skills a_nd adaption success for CNC a_rlg
FMS

H1A states that there is no relationship between skills and adoption
success. The analysis for the overall sample used three tools to address this
hypothesis and found no evidence of a relationship between skills and success.
Equally important from the standpoint of addressing previous literature, there
was no evidence that higher skills were linked to adoption success. This section
addresses the same hypothesis, but for each technology individually. Because
there are only eight FMS installations in the sample all analysis for the FMS will
be based on qualitative tools. However, the sample of 22 CNC users is large
enough to use simple statistical tools, along with the visual tools.

Figure 4-9 is the scatterplot of skill level to adoption success for FMS

users. The figure does not indicate any type of relationship between these two

 

 

 

 

 

Figure 4-9
Skill Level and Adoption Success - FMS
0.9
E 0.85 e
E 0.8 e
g 0.75 .
:‘E, 0.7 a- . Q .
< 0.65 -_
E 005: _-
0 ' i
0.5 5 l l l 7*
0 50 100 150 200 250 300
Skill - Total Preperation Time in Weeks

 

 

 

214

variables, for this technology. The figure is augmented by Table 4-26 which
shows the sorted (ascending order) levels of success and the corresponding skill
level at each installation. Once more there is no evidence of a relationship
between skills and success. Equally important is that the two most successful

installations had 32 and 2 weeks of training respectively.

Table 4-26 Skills to Success for FMS

 

Skills-total preparation 1 2 32 50 81 105 252 265
time

 

Success-goal .757 .831 .857 .743 .695 .694 .714 .792
achievement

 

 

 

 

 

 

 

 

 

 

The relationship between skills and adoption success for CNC users was
addressed using correlation and visual tools. The correlation between goal
achievement and adoption success for the CNC users was -.053 (p = 814). The
relationship is portrayed visually in Figure 4-10. Neither the correlation nor the
scatterplot provide evidence that higher skills lead to higher levels of success.

The results for the sub-samples on H1A are the same as the results for
the overall sample. There is no evidence that higher skills lead to higher levels
of success for either CNC or FMS installations. In addition, the most successful

4. 5. 2. 1.6 H18 - The relationship between appropriate choice a_nd adaption
success for FMS a_nd CNC

H1 B states that installations where appropriate levels of skills are selected
will be more successful than installations where inappropriate levels of skill are

selected. For the overall sample 4 fit measures were tested; the original

215

 

 

 

 

 

Figure 4-10
Skill Level and Adoption Success - CNC
0.9
6 0 ° v.’
0.8 .. . O. Q .
‘E 0 e
g 0.7 . .
g P . ..
.2 0.6 . .
f, e
S 0.5 .
G
O
‘9 0.4 .
0‘3 * . : : : r '
0 50 100 150 200 250 300 350
Skill - Total Preparation Time In Weeks

 

 

 

formulation as well as three other operationalizations based on the literature.

For the overall sample analysis was performed using correlations as well as
calculating the percentage of installations that were properly categorized by each
measure. For the FMS installations correlation would be inappropriate because
of the small sample size. Therefore just the ability to categorize will be
assessed. However, the CNC sample is large enough to use both the statistical
and the categorical tests.

Table 4-27 shows the levels of success and the 4 proposed fit variables
for the 8 FMS installations. FIT1, FIT2 and FIT5 classify 50% of the installations
correctly while FIT3 only classifies 25% of the installations correctly. Because
none of the measures are capable of classifying more than half of the
installations correctly (what random assignment should accomplish) there is no

evidence that for this small group of FMS users appropriate skills leads to

216

adoption success. However, it should be noted that changing one prediction
would change the number properly categorized to over 62%. Therefore, no
strong conclusions can be made about FMS adoption and appropriate skill

choices.

Table 4-27 FIT and Adoption Success For FMS

 

 

 

 

 

 

 

 

Success FIT1 (original FIT2 (Just FIT3 (Just FIT5 (No
Formulation) internal) Productlproces Discretion)
8)
.694 yes no no yes
.695 no no yes no
.714 yes yes yes yes
.743 no no no no
.757 yes yes yes yes
.792 no no no no
.831 yes yes no yes
.857 no no no no

 

 

 

 

 

 

 

The CNC sub-sample was analyzed using correlations and ability to
categorize. Table 4-28 shows the correlations between the proposed fit
variables and success. The results are similar, although stronger, than the
results for the overall sample (see Table 4-14). The original fit formulation as
well as F IT2 and FIT3 have strong correlations with success. The large and
significant effect sizes are supported by the ability to categorize. Table 4-29
shows the ability of the individual fit measures to place an installation in the
correct category. For FIT1 and FIT2 82% of the installations were correctly
categorized, while for FIT3 and FIT5 73% of the installations were properly

categorized.

217

The results from these tests are mixed. There is no evidence that
appropriate choices lead to adoption success for FMS installations. However, for

the CNC installations the majority of the correlations are very large considering

Table 4-28 Correlations Between Fit and Success - CNC

 

 

FIT1 FIT2 FIT3 FIT5
Correlation | .488 .488 .461 .159
Probability I .021 .021 .031 .478

 

 

 

 

Table 4-29 Ability of various FIT variables to Categorize for CNC

 

 

 

Variable Number Correct Number Wrong Percentage
Correct
F1 18 4 82
F2 18 4 82
F3 16 6 73
F5 16 6 73

 

 

 

 

 

that this is a relationship with a performance variable. Additionally the
relationships for FIT1, FIT2, and FIT3 are significant. Both F IT1 and FIT2
perform equally well for this sample because they happen to have categorized all
installations the same way. Based on previous results as well as these, any of
the fit measures (with the exception of FIT5) is probably appropriate, although
FIT1 and FIT2 find the most support in the literature. The lack of findings for
FMS may be due to the sample or because the majority of the literature used to
build the theory for this dissertation was based on CNC installations which may

differ from FMS technologically and in ways that have yet to be uncovered.

218

4. 5. 2. 1. 7 H2 - H4: Drivers of choice

Choice is hypothesized to be driven by the level of product/process
change, the level of environmental uncertainty and the level of managerial
discretion. The analysis of the overall sample (section 4.4.4.1) found evidence of
direct effects for all three indices, but only limited evidence of the hypothesized
moderating effect of managerial discretion. The FMS sub-sample is examined
using a matrix and scatterplots, while the larger CNC sub-sample is examined

using regression and scatterplots

Table 4-30 Key FMS Variables Sorted by Skills

 
 

 
   

    

  

 

 

 

 

 

 

 

 

 

 

Skills i Productlprocess Environmental Managerial
(preparation time . Change Uncertainty Discretion
i in weeks) l
1 (low) -1.01 (low) 10.5(low) high
2 (low) II -.616 (medium) 10 (low) low
32 (medium) ll .46 (high) 12.5 (low) ligh
50 (medium) H -.6307 (medium) 8.5 (low) low
81 (medium) -.4368 (medium) 17 (high) low
105 (medium) JI -.975 (low) 19 (high) high
252 (high) [I .954 (high) 13 (low) high
265 (high) | -.425 (medium) 13.5 (low) high

 

The variance of product/process change for the FMS sub-sample is fairly

limited. Therefore the degree to which the level of product/process change could

effect skill level is also limited. Table 4-30 shows limited evidence that skills

increase with product/process change, while Figure 4-11 does exhibit some

evidence of the predicted upward slopping relationship. However, the limited

 

219

variance on this variable makes it difficult to come to any strong conclusions as

to the relationship.

Figure 4-11

 

Product Process Change and Skill Level (FMS)

 

0.8 .
0.6
0.4 . 9
0.2 .

 

-0.2ll so 100 150 200 250 _ 300
-o.4 . - .
-0.6 . ,
-o.a .
-1 ,, e
-1.2

Product Process Change

 

 

 

Skill Level - Total Preparation Time in Weeks

 

 

Figure 4-12 examines the relationship between environmental uncertainty
and skill choice for the FMS sub-sample. The scatterplots suggest that in
general as uncertainty increases so do skills. A similar relationship is evidenced
in the table, where the higher preparation time installations have the higher
environmental uncertainty scores. The relationship between uncertainty and
skills is very similar to the relationship for the overall sample; skills increase with
uncertainty but the distribution is fairly wide.

The third hypothesized driver of skill choices is managerial discretion. For

the overall sample the level of managerial discretion did have a significant effect

 

220

on the level of skill chosen. The FMS sub-sample exhibits a similar

characteristic. The 3 highest skilled FMS installations all have high

Figure 4-12

 

Environmental Uncertainty and Skill Level (FMS)

300

 

250 .- O

200.

100. 0’

50. Q

Skill Level- Total Preparation
Time In Weeks
8

 

 

0 - e e

0 5 10 15 20

 

Eivironmental Lhcertainty

 

 

 

levels of discretion, while the lower skill installations are mixed high and low
discretion. High discretion would be required to have high skills, while low and
medium levels of skill can occur with both high and low discretion depending on
the product/process and environmental uncertainty levels. The evidence
suggest that high discretion is indeed related to high skills for FMS, although
coming to conclusions about lower levels of skill are more difficult.

The evidence for the FMS sub-sample is fairly limited. However there is
at least some evidence that each of the three drivers of choice exhibits some of
the expected characteristics. When combined with the findings for H1 B

(appropriate choices) it is possible that management for FMS installations is

221

making skill choices based on some of the same drivers as the overall sample.
However, these drivers may not lead to appropriate choices for FMS based on
the lack of a relationship between any of the FIT variables and success.

The proposed relationship between the drivers of choice and skill level are
much stronger for the CNC sub-sample than for the FMS sub-sample. Table 4-
31 shows the results of the regression of the drivers of choice on skill level. The
R squared is .730 (adjusted .685) which is much higher than the R squared for
the overall sample which was .616 (adjusted .572). In addition all three of the

hypothesized drivers are significant in the predicted directions.

Table 4-31 Drivers of Choice Regressed on Skill Level for CNC

 

 

 

 

Variable Coefficient P(2 tail)
(Standard
Coefficient)
Constant -5.44 (000) -. 120 . .906
Product/process 15. 458 (.579) 4.550 .000
Change
Environmental 12.598 (.417) 3.286 .004
Uncertainty
Managerial -84.154 (-.373) -2.878 .010
Discretion

 

 

 

 

 

The larger R squared as well as the strongly significant results for the
individual items, even with the decrease in sample size, may indicate that the
overall model of the drivers of skill choices is a better predictor for CNC
installations than for AMT overall. Like the finding that the FIT variables tested
for hypothesis H1 B were more suited to CNC, the evidence seems to suggest

that the drivers of choice are at least somewhat different for CNC and FMS.

222

However, the limited variance on product/process change for the FMS
installations may also be a chief explanation for the differences. Although there
are FMS installations with all three levels of product/process change (low,
medium and high as described in Table 4-12) the reality is that they hover
around the middle of the range. Therefore with slightly different categorizations
the FMS installations might have fit much better. The end result is mixed
evidence on the appropriateness of the model for FMS, with strong statistical

evidence for the CNC installations.

 

 

 

 

 

 

 

Figure 4-13
Product Process Change and Skill Level (CNC)
8
2 3 2 - ’
n. 8 .
I! E 0 j 9 l .L l e? l :
r? g _2 _ 0m . 100 150 200 250 300 3150
E I: -4 _
3
E -6 P O
.I
“’ —8
Product Process Change

 

 

In addition to strong statistical support the scatterplots (Figures 4-13 to 4-
15) of the proposed relationships also show the hypothesized relationships.
Figure 4-13 clearly shows that as product/process change increases skills also

increase. Figures 4-14 and 4-15 are similar to the figures for the overall sample

 

223

Figure 4-14

 

Environmental Uncertainty and Skill Level (CNC)

 

 

 

 

20 c
z. 18 . °
=
“ 16 _
e
g 14 ° ° ,
g 13.? ’1 _ g . ’3:
g 8 p e . 3 e
: 6 ' .
.2 4
E 2 .
"“ o . . .» . . , ’ >
O 50 100 150 200 250 300 350

Skill Level - Total Preparation Time In Weeks

 

 

Figure 4-15

 

Environmental Uncertainty to Skills (CNC)

350
300 . ,
25o. » " .° 0 . .
e

200 . g . . 1.4»
150 . V . ‘
100 . O

50 _ . 0‘». _ .
0 2 ‘ 9 I ‘ 3 V‘:

 

 

 

Skills- Total Preparation Time

 

Environmental Uncertainty

 

 

in that the plot of environmental uncertainty to skills needs to be rotated so that

skills are on the Y axis in order to see the proposed relationship. Once rotated

 

224

(Figure 4-15) the figure indicates that as environmental uncertainty increases
skills generally increase as well (although, the distribution is fairly wide).
4.5.2.2 Conclusions of compag’sons of FMS and CNC

The analysis of the sub-samples resulted in findings similar to those for
the overall sample. However, FMS and CNC do differ to a degree on a number
of facets that are important to the study. First FMS installations have a much
more restricted range of product/process change responses. This finding may
be a result of the sample or it is possible that FMS is adopted in fewer different
environments than CNC. The internal environments for FMS may be limited but
there was no evidence that the users of FMS and CNC differed on their external
environments. Therefore differences between FMS and CNC outcomes are not
likely to be driven by the uncertainty of the external environment.

FMS users also use lower skill (on average) employees than CNC users.
In addition there are successful FMS installations with very low levels of skill.
Therefore there is no evidence that FMS requires higher skills than stand alone
equipment. Nor is there evidence that FMS requires high skills to be successful.

In addition to there being no evidence of a relationship between skills and
adoption success (H1A) the sub-sample data generally supports the remainder
of the hypotheses as well. H1 B, appropriate choices lead to success, is strongly
supported for the CNC users, although it is difficult to come to any conclusion for
the FMS users. Hypothesis H2 - H4 on the drivers of choice are also strongly
supported for the CNC users, with less support for the FMS installations.

Therefore it appears that FMS and CNC users make skill choices based on

225

similar variables, but that adoption success for FMS installations may be due to
other variables not examined in the dissertation.
4. 5. 3 Industry Effects

The previous analysis has shown various levels of support for the
proposed hypotheses for the overall sample, as well as each of the technologies.
One area that has not been examined is industry. Companies who are making
the same or similar products for sale to the same customers may have similar
order winners, losers and qualifiers. Therefore, they could have similar
manufacturing environments. Additionally they should face the same external
environment. If these assumptions are true than a significant portion of the
driver for skill choice, as well as the appropriateness of the choice could be the
industry that the plant’s products competed in.

However, authors such as Roth and Miller (1994) and Boyer et. al. (1996)
have shown that even when there are industry effects, companies still make
different choices in regards to manufacturing strategy and AMT investments.
Environmental uncertainty is a perceptual variable (as it should be to explain
choice) and will vary based on how management views the environment.
Manufacturing environments will be partially a result of manufacturing strategies
which do vary even within an industry (Roth and Miller 1994). Finally the
concept of an industry is also rather hard to define, especially due to increasingly
niche-type market segments.

For instance, there are 8 companies in the primary sample that are in SIC

code 3544 (specialty tools and dies - which also includes molds). However,

226

these eight companies make two very different products; dies and molds.
Additionally the die makers do not all compete with each other, rather they are in
two different strategic groups (Porter 1980). One strategic group that sells large
automotive stamping dies and another strategic group that makes much smaller
stamping dies for auto and other applications. The end result is that even within
a fairly tightly defined industry represented by one or two 4 digit SIC codes there
can be many different groups of companies who are dealing with different
customers who have different needs. The products the companies make may
share some similarities, which will lead to some homogeneity between
installations, but there is the potential for large differences as well.

The following sections examine the eight industries in the primary sample
that had multiple installations in an attempt to find evidence of similarities and
differences among members of the same industry. Industry was generally
defined using the four digit SIC code for the primary product made within the
plant. However, SIC 3544 was split between die makers and mold makers
because the respondents themselves considered these separate industries. In
contrast SIC 2521 and 2522 (wood and steel office furniture respectively) were
combined because the companies considered their market to cover both
materials. Similarly SIC 3541 (machine tools metal cutting) and 3542 (machine
tools metal firming) were combined. Finally SIC 3714 (auto parts) contains a few
firms who consider themselves screw machine parts makers (SIC 3451) but their

output is all for automotive applications.

4.5.3.1 Machine Tools

227

Companies in this industry make metal forming and cutting tools that are

sold around the world. These companies may be unique because they are

making AMT, as well as using AMT. A distinction which may give them an

advantage. Table 4-32 contains information about all 8 installations in this

industry, as well as summary information for the industry and the overall sample.

Table 4-32 Machine Tools

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable I - 1 l - 2 I - 3 I - 4 l - 5 I - 6 l - 7 l - 8 Industry Sample
Avg Ag
Batch Size 10 17.5 2.8 15 3.5 3.5 200 20 259 I 19023
0 10.33
Churn 35 15 51 140 15 0 6 100 45.25 56.15
Families 7 1 24 200 500 20 50 3 888.125 1108
0 0
MP8 35 10 12.5 25 5 20 20 10 17.18 11.9
Product , .28 -.44 .95 4.28 -.43 -.97 -.81 2.7 .70 .467
Process ‘ 8 4“
External 19 12 13 12 13.5 9 8.5 9 12 j] 12.1
[Skill 196 1 252 262 265 49 4.2 223 ll 156.5fll 120.01
Compet. 30 4 12 4 4 7 6 12 9.875 H 7.828
Size 130 300 170 120 550 140 147 150 404 621 .5
0
Discretion low low high high high high low hig
h
Technology CN CN FM CN FM CN CN ON
C C S C S C C C
Self 4 5 5 6 6 6 6 6 5.5 5.87
Assess
| Success |F57 .67 .71 .79 .79 .83 .84 .89 .76 .78

 

The table is arranged so that installation 1 (I -1 ) had the lowest level of goal

achievement while installation 8 (I -8) had the highest.

228

The table shows that even within the industry there are some significant
differences. First is the way that the AMT is used. Installation 7 (I7) is using
CNC equipment to make small parts in fairly high volumes, while the remainder
of the installations are making parts in much smaller batches. In addition, there is
a large amount of fluctuation in the range of product churn, with one installation
(I6) using CNC equipment to make small batches of parts that are not retire or
changed, while another (I4) is using FMS to make slightly larger batches of parts
that are often not run again. The end result is product/process change scores
that range from -.97 to 4.28, with two installations having low levels of
product/process change, three with medium levels of product/process change
and three with high levels of product/process change.

The variance in the internal environment is also evident in the external
environment where companies have scores that range from 8.5 to 19. The
variance in internal and external environments should, based on previous
analysis, lead to large variations in the skill level of the workforce, which it does.
There are installations with preparation times that range from 1 week to 265
weeks.

Not only does this industry display large variance in terms of internal and
external environments, there is also evidence that the proposed model of skills
choices and adoption success applies. For instance the lowest performing
installation has a preparation time of 196 weeks. I - 1 has an internal
environment that would be categorized as medium, an external environment that

is highly uncertain and low levels of discretion. However they have chosen high

229

skills. The end result is that this company which installed CNC equipment to
lower their costs was unable to do so. The union insisted on twice as many
highly skilled employees as the company thought they needed, seriously inflating
operational costs. This inappropriate choice of high skills leads to low levels of
success. Meanwhile l7 had a preparation time of 4.2 weeks, yet is the second
most successful installation in the industry. This is because I7 has matched low
skilled employees to an environment characterized by low levels of product/
process change, low levels of environmental uncertainty and low levels of
discretion.

Within this industry there is evidence of a great deal of variance on most
of the variables of interest. This variance may be due to different strategies
and/or different perceptions. For instance the majority of the companies are
using AMT to make parts in fairly small batches, however l7has chosen to make
parts in much larger batches. The other companies may outsource the parts
made in house at I7, or may have chosen to use another technology to make
these parts. Decisions about what to make and buy, as well as width of product
offerings seems to effect what the internal environment looks like, even for these
companies within the same industry.

4. 5.3.2 Stamping Dies

The companies in this industry make stamping dies that are used to form
sheet metal parts. In addition some of the companies also repair and/or make
engineering changes in existing dies. This group of companies is much more

homogeneous than the machine tool makers were, as is evidenced by Table 4-

230

33. The internal environments are very similar, as are the external

environments and the skill choices.

The product process scores for all 5 installations are high, with the

majority being at the extreme end of the distributiOn. All of the installations make

Table 4-33 Stamping Dies

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

Variable Installat Installat Installat Installat Installat Industry Sample
ion 1 ion 2 ion 3 ion 4 ion 5 Avg Avg

IBatch Size 1 1 1 1 1 1 19023

Churn 190 60 40 200 190 136 56.15

Families 5000 6 5000 5000 5000 4001 1 108

MP8 15 .1 0 10 15 8.02 II 11.9

Product 6.15 1.27 .522 6.53 6.15 4.12 H .467

Process

External 9.5 13 15 9 10.5 11.4 I 12.1

Skill 192 250 201 192 216 210.2 120.01

Compet. 7 50 8 50 15 26 7.828

Size 600 90 500 310 130 326 621.5

Discretion low high m high high

Technology CNC CNC CNC CNC CNC

Self 5 6 6 6 6 5.8 II 5.87

Assess

Success .781 .8 .81 .86 .88 .83 .78

    

 

 

 

 

 

or repair dies in batches of one. In addition four of the five installations have

levels of churn greater than the overall sample average, with three of the

companies doing little to no repair or rework so that each part is basically unique

and never repeated. The two installations where churn is much lower are owned

by companies who actually do metal stamping as well. These installations do

many repairs and will see a die a number of times over its life. Hence the

231

companies have lower churn compared to the companies who only make dies
and do not stamp sheet metal for production purposes.

There is a wide range in the external environment, as compared to the
internal environment for this sub-sample. The variance is more evident when
examine the number of competitors rather than environmental uncertainty. The
companies making the smaller dies compete with upwards of 50 other die shops
for work, while the companies making larger dies are competing with a smaller
number of other die shops.

Skills do not vary much for this group, which is not surprising given the
high level of product/process change. All but one of the installations use
journeymen machinists who have gone through an apprenticeship that lasts a
minimum of four years. The fifth (I3) uses an in-house program that is very
similar to the apprenticeship program.

Levels of success also do not vary significantly. The lowest performing
member of this group still performs at the average level for the overall sample.
High levels of success may be because the companies in the industry have
adopted a standard level of skill (journeyman status required) that is mated to
their manufacturing needs. It is also possible that while there was no evidence
of the age of an installation effecting overall levels of success in the compete
sample, that this industry has been using CNC and before that NC for about 20
years in most of the installations. Therefore the companies may be very good at

using CNC for their needs.

232

The stamping die industry does display homogeneity on a number of the

variables. This homogeneity is indicated in the very high levels of productl

process change and the correspondingly high levels of skill. Within this industry

high skills may be the appropriate choice for all inStallations.

4.5.3.3 Auto Parts

Installations were classified in this category if their primary products were

used in automotive applications, regardless of type of technology (hence the

inclusion of screw machine products). However, installations whose output was

primary for diesel engines used for construction equipment, over the road trucks

and the like were put is a separate category, diesel power. Only one of the

installations was owned by an actual auto company. The rest were independent

Table 4-34 Auto Parts

    

' Variable

 

  

 

   

     

   

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Installat Installat Installat Installat Installat Industry Sample
ion 1 ion 2 ion 3 ion 4 ion 5 AVL Avg
Batch Size 50,000 1 100 2400 220,00 140,00 82,700 19023
0 0

Churn 10 0 14 16 7 9.4 56.15
Families 2 1 3 1 70 15.4 1108
MP8 20 20 20 10 5 15 1 1.9
Product -2.34 -1.01 -.54 -7.95 -5.53 -3.48 .467
Process
External 10.5 8.5 13 8.5 9.8 12.1
Skill 1 2 1 2 14. 68 120.01
Compet. IggF 10 4 1 3 II77. .828
Size [F 180 450 600 900 367 499. 4 621.5
Discretion high high high high low high
Technology CNC FMS CNC CNC CNC
Self 6 5 6 6
Assess
Success .76 .78 .85 .86

 

 

 

 

 

 

 

233

(and usually non-union) suppliers to the auto companies. Table 4-34 indicates
that this industry is similar to the die industry in that it is fairly homogeneous.
Auto assembly may be moving towards JIT production, but the making of
auto parts has not progressed as far. The average batch size for companies in
this industry was 82,700 with a few of the installations making parts in a
continuos mode. Additionally the smallest batch size was 1100 parts. The
installations varied greatly on batch sizes, but not on the majority of other
product process variables. All but one installation had a low overall level of
product/process change. In addition most of the installations had low levels of
churn and only made a few families of parts. Internally these installations are still
making large batches of a small variety of part families, with little change from
year to year. However there are some limited signs of change such as the
smaller batch sizes, as well as the installation with 70 part families.
Environmental uncertainty also does not vary a great deal for this industry.
The installations supply parts to a limited number of companies who are
generally the big three North American auto makers and one or two Japanese
transplants. In addition recent industry initiatives to reduce supply bases have
left those companies that remain with longer contracts for more parts, increasing
stability. Finally each of these companies tends to specialize in a specific type of
auto part. One makes small round parts for fuel and brake systems, while
another specializes in larger round parts for drive trains. This specialization

reduces the number of competitors the companies have.

234

The low levels of product/process change and environmental uncertainty
are also evidenced in the skill choices. Four of the five installations have chosen
very low skills, while the fifth (l1) has chosen medium levels of skill. Sixteen
(highest for industry) is also the least successful Of the auto installations,
providing some indication that high skills are not needed for success and that
matching skills to the environment can increase success.

Like the die industry the auto part industry does display some
homogeneity. In general simple manufacturing environments that are
characterized by a limited number of parts that are run in high volumes allow the
companies to choose low skills and be successful. In addition the one company
that does not follow the industry pattern is the least successful (although by a
minimal amount). Low skills may be the choice for auto parts makers at this
time. However, there are some indications that low skills do not have to remain
the choice. Batch sizes at some of the installations are falling. And many of the
auto companies are working hard to shorten life cycles, which should induce
more churn. Therefore in the future these installations may need to respond to
increasingly complicated environments by increasing training beyond the present
industry standard one to two weeks.
4.5.3.4 Diesel Power

The installations in this industry make parts for diesel engines that are
used in a variety of applications, primarily on highway trucks, construction
equipment and power generation. four of the five installations were in engine

assembly plants, while the fifth (l3) installation is a separate facility that only

235

makes parts and does no engine assembly. The four installations where the

AMT was in the assembly plant used the AMT to fabricate parts for which there

were a large number of options. For instance one of the engine manufacturers

had over 900 possible configurations of one engine size for one customer.

These configurations varied by oil pan, intake and exhaust manifolds (for

differences in horsepower and/or emissions), water pumps and many other items

that bolted onto the basic block, pistons and head. The fifth plant (I - 3) made

the higher volume parts that were standard across various configurations.

Table 4-35 Diesel Power

 

 

 

 

 

 

  
 

 

 

 

 

 

 

    

 

   
    

 

 

   

 

   
   

 

 

 

 

 

Variable Installat Installat Installat Installat Installat Industry Sample
ion 1 ion 2 ion 3 ion 4 ion 5 Avg Avg
Batch Size 10 15 10000 25 45 2019 / 19023
24

Churn 0 15 19 0 30 12.8 56.15

Families II 1 2 8 3 6 . 4 1108
IMPS 5 20 20 2.5 1 9.7 11.9

Product -.97 -.43 -.63 -.98 .10 -.58 .467

Process

External 19 17 8.5 15 13 14.5 12.1
Bill 105 81 50 48 8 58.4 JI 120.01

Compet. Jr 3 4 0 2 4 2.6 II 7.828
[Size 675 1500 1000 2100 600 I 1175 621.5

Discretion flgh low low low low

Technology FMS FMS FMS CNC CNC

Self 5 7 5 6 6 5.8 5.87

Assess

Success .69 .70 .74 .76 .79 .74 .78

 

   

The plants have similar, but not identical internal environments. Three of

the plants have low levels of product/process change, while the other two have

236

medium levels of product/process change. The key differences are on the
aforementioned batch sizes as well as the number of families and the amount of
churn. One plant (I1) is making a single part family in small batches on an FMS.
This part family does not change from year to year so there is no churn. This is
in contrast to l5 who is making many non-standard parts in low volumes, some of
which they will only run for a single batch. The end result is that the companies
are similar but not the same, due to differences in the types of parts and the level
of customization.

Four of the five installations view environmental uncertainty as being high,
and this industry has the highest industry average (14.5) on this index. The fifth
plant (I3) is the parts plant, which is a subsidiary of one of the engine
assemblers, and does not have to sell end products. The uncertainty is not due
to the number of competitors because all of the markets for diesel power are
dominated by a small number of large firms.

The diesel power installations are making some parts that are similar to
the auto parts installations (manifolds, water pumps and the like) but they are
doing so in smaller batches with more churn. In addition the majority perceive
their external environment to be fairly uncertain. Therefore it is no surprise that
the average level of skills for the diesel installations is 58.4 which is much higher
than the level chosen for the auto parts producers. However all but one of the
plants is unionized (l1) and all of the union plants have low levels of discretion.

The levels of skill chosen vary significantly for these installations, but the

outcomes do not. The installations were on average less successful than the

237

overall sample. Additionally only one installation had levels of success that were
higher than the average (I5). Finally, success was inversely related to skills.
The three highest skilled facilities were the FMS facilities, who had the worst
outcomes of the group. Because of the previously discussed issues regarding
the proposed model and FMS installations it is difficult to conclude why these
companies are doing relatively poorly. However, one possibility is that as the
companies move towards higher and higher levels of customization they have
seen an increased need for skills. However their limited discretion may be
limiting who they promote to jobs that are becoming more complex.

Increased customization compounded by the effects of an industry move
toward JIT, may be making these jobs more complex than they were in the past.
This may cause unions in the plants to respond (especially in the FMS settings)
with increased training demands, that may not be needed. An alternative is that
the increased needs for customization have not yet been met with increased
skills in the workforce.

4. 5. 3. 5 Injection Molds

Injection mold makers and die makers are part of the same four digit SIC
code (3544). However, molds and dies are different, and the companies in this
groupsee themselves as being in a separate industry from the die makers. The
markets may differ, but the installations in this industry are very similar to the
installations in the die industry in terms of skill level and product/process change.

In addition Table 4-36 shows that injection molds makers in this sample were

238

generally small plants, with highly skilled employees and generally high levels of
success with CNC technology.
The mold makers have the highest average product/process change

score. All of the molds makers do mostly new melds, leading to average churn

Table 4-36 Injection Molds

 

 

 

 

 

 

 

  
  

 

 

 

 

 

 

 

  

 

 

 

  

 

 

 

 

 

Variable Installation Installation Installation Industry Sample
1 2 3 Ave rafige Average
Batch Size 9 1 1 3.67 19023
Churn 130 180 160 157 56.15
Families 5 5000 1 1668 1 108
MP8 7 30 5 14 1 1 .9
Product 3.9 5.78 5.03 4.90 .467
Process
External 15.5 11.5 20 15.67 12.1
Skill 342 195 210 249 120. 01
Compet. 6 7 12 8. 33 |I 7. 828
Size 260 37 10 102 621. 5
Discretion high high high
Technology CNC CNC CNC
Self Assess 7 7 7 5. 7887
Success .66 .84 .89

  

 

 

of 157 for the industry. An additional issue is batch size. One of the mold

installations specializes in molds used for plastic beverage containers. This

installation runs larger batch sizes because they sell some standard products,

that are used on machines that have a 4 mold capacity. The other two

installations tend to make molds that are used on machines with a single mold

capacity, hence the average batch size of one. However l3 makes multiple

cavity molds that are used for rubber products. It could be argued that the batch

239

size is really the number of cavities (since each cavity is identical) therefore their
batch sizes are really in the 6-12 range.

In addition to having high levels of product/process change, two of the
three installations report high levels of environmental uncertainty. This is in
contrast to the more predictable stamping die industry. The one installation (l2)
that reported low levels of environmental uncertainty is a facility that makes all of
its molds for a parent company whose needs exceed the installations capacity.
Therefore everything the installation can make is already “sold” and they do not
have to worry about finding customers or competitors taking their work.

The higher levels of product/process change and environmental
uncertainty are linked to high skill levels as well. The mold makers use similar
apprentice programs to those used by the die makers, resulting in the same
highly skilled employees. The high skills are linked with high levels of goal
achievement at two of the three installations. The third insulation (l - 1) is still in
the process of installing CNC equipment and the plant manager noted that six
months ago they were doing much worse and in six more months they would
probably be where they expect to be. Within this industry there is a great deal of
homogeneity. All of the installations had complex shop floor environments,
highly skilled employees and were very satisfied (self assessment was 7 at all
the installations) with either the present or expected performance of their

systems.

240

4. 5. 3.6 Office Furniture

The two office furniture installations are in different SIC codes because
one plant makes wood furniture and the other makes metal furniture. However,
both companies consider the other to be a major competitor in their main market.
Therefore the two companies are examined simultaneously in Table 4-37. The
table shows the types of differences the literature would predict for FMS.
Namely l2 has a much more complex manufacturing environment because they
are running small lots of many different frequently introduced products on a
single FMS. This is in contrast to l - 1 which is using CNC to run medium sized
lots of a limited number of parts, with no new product introductions or

retirements.

Table 4-37 Furniture

Variable Installation 1 Installation 2 Industry Sample
», Average Average

          
    
  
   
  
  
   
    
  
 

   

 

 

 

 

 

 

 

 
 
    
   
     

 

   

 

Batch Size 500 10 255 19023
I Churn 0 40 20 56.15
Families 4 30 17 1108
I MP8 .5 1.5 1 1 1.9
Product -.99 .46 * -.27 .467
Process
I External 6 12.5 9.25 12.1
Skill 38 32 35 120.01
Compet. 3 5 4 7.828
Size 1 750 650 II 700 621 .5

Discretion
Technology
Self Assess
Success

|| high high
CNC FMS I

 

   

    
 

 

7 6
.80 .86

6.5
.83

5.87
.78

  
    

 

 

 

241

In addition to having a more complex internal environment, the
management at l2 perceives the external environment to be much more
uncertain than the management at l - 1 perceives their environment to be.
However, neither company perceives the environment as having a high level of
uncertainty.

Both companies are using workers with medium levels of skills, which is
not the skill level that is predicted for either installation. I - 1 would be predicted
to have low skills to match their low levels of product/process change and
environmental uncertainty, while I2 is predicted to have high skill because of the
high level of product/process change. While neither fits the hypothesized model,
both have above average levels of achievement.

The furniture installations are not totally homogeneous. One plant uses
their AMT to make standard parts in large batches while the other makes much
smaller batches of a wider variety of parts. However, they both view the external
environment as fairly certain and have similar levels of skills for their employees.
The lack of fit for these installations may also be due to the industry. One plant
is working with wood rather than metal, which may lead to different requirements.
In the other plant the CNC equipment is used to punch the metal, not machine it,
which may also change requirements from the employees.

4. 5. 3. 6 Industry conclusions

Breaking the companies into industries provided support for industry

effects in the sample. This is not a surprise since the variables of interest were

measures of the internal and external environment.

242

Stamping dies and injection molds were two industries where the
installations were very similar. However even within these groups there were
differences such as the die makers who specialized in very large dies who
contrasts with the die makers who do lots of rew6rk and/or repairs. The auto
parts makers, diesel power providers, and furniture makers also displayed some
similarity, although there were more differences among these companies than
among the die and mold makers. Finally the machine tool industry provided a
microcosm of the overall sample with eight very diverse installations.

The existence of industry effects suggest that if a manager has in depth
knowledge of an industry that may be the place to start when trying to mate skills
with AMT. However, there are a number of confounding factors such as
differences in manufacturing strategy that can effect skill choices. In addition
one of the key differences among the machine tool makers was the choice of
which parts to make with the AMT. The degree of outsourcing, the level of
customization, volume and a host of other factors will affect the level of internal
uncertainty, while the choice of customers and suppliers will impact
environmental uncertainty. When differing levels of managerial discretion are
introduced there is some evidence that industry alone is not a good predictor of
skill level, especially if the industry is changing or the company is attempting to
compete in ways that differ from industry norms. Within this small sample, there
is some evidence that in some industries there is a correct skill level to use with

AMT, but that skill level varies with the level of product/process change,

243

environmental uncertainty and managerial discretion for the individual firms.

Future research should control for industry differences.

Chapter 5
DISCUSSION OF RESULTS

5. 1 Overview and chapter contents

Chapter five synthesizes the results of the analysis discussed in Chapter
4. Section 5.2 summarizes the results of the analysis and discusses the level of
support for each of the proposed hypotheses. Section 5.3 discusses the findings
on the relationship between skills and AMT adoption success. Section 5.4
examines potential of no relationship existing between the drivers of choice and
the best of the FIT measures. Section 5.5 examines the impact of industry
effects on the findings. Section 5.6 examines the differences between CNC and
FMS installations. Section 5.7 summarizes the key conclusions of the research.
Section 5.8 re-examines some of the limitations of the research. Finally section
5.9 looks at future research possibilities that arise as a result of this study.
5.2 Level of support for proposed hypotheses:

Tables 5-1 and 5-2 summarize the results of the analysis performed in
chapter 4. Table 5-1 is the results of the primary analysis as proposed in chapter
3, while table 5- 2 details additional tests that were performed either to test

alternative variables or as controls.

244

Table 5-1 Primary Results

245

 

 

 

 

 

Hypothesis Method Result Industry FMS/CNC
Effect Differences
H1A: There is Correlation Not Supported Yes No
a positive -
relationship
between skills
and adoption
success
H1 8: Correlation Supported no yes
Appropriate (support
skills lead to limited by
higher levels restricted
of adoption range of goal
success achievement)
H2: Higher Moderated Supported yes no
levels of regression
productl
process
changelead
to higher skills
H3: Higher Moderated Supported no no
levels of regression
environmental
uncertainty
lead to higher
skills
H4: Moderated Partially no no
Managerial regression supported -
discretion managerial

moderates the
relationships
in H2 and H3

 

 

discretion has
a direct effect

 

 

 

All of the hypotheses, but H1a, are at least partially supported. This support is

diminished somewhat by the industry and technology effects, but not enough to

dismiss the overall findings. Additionally the alternative formulations of the

variables displayed in table 5-2 do not change the results. Controlling for plant

age does not affect the results. Finally plant size is negatively related to skills,

 

246

but the variance accounted for by size seems to be captured by the managerial

discretion variable. That is the addition of plant size into the moderated

regression does not alter the results or reduce the size of the significant beta for

managerial discretion (see section 4.5.1.5).

5.3 Skill Level of the Workforce and Adoption success

A key assumption (tested by hypothesis H1A) of this dissertation was that

the skill level of operational employees was not related to the successful

Table 5-2 Secondary Results

 

 

 

 

Alternative Hypothesis(s) Method Result Change From
Variable tested Prima ?
Self H1A Correlation no no
Assessment relationship
of Success for between self
Goal assessment
Achievement and skills
Direct H2 - H4 Moderated Direct yes
Competitors Regression Competitors
for not a
Environmental predictor of
Uncertainty Skills
Age of H1A: Correlation Age does not NIA
Installation affect skill
choices or
outcomes
Plant H1 - H4 Correlation Size related NIA
Employment and to skills and
Regression Union - in
moderated
regression not
significant -
vafiance
accounted for
in managerial

 

 

 

discretion

 

 

 

247

adoption of AMT. There is no statistical evidence of a relationship (linear or
curvilinear) between skill level and adoption success. The restricted range of the
goal achievement measure may partially explain the statistical results. However,
there are a number of highly successful low skill firms in the primary sample,
using both types of technology. Therefore, the evidence from this sample
suggests that skills in and of themselves are not a predictor of adoption success
for either CNC or FMS. This result sheds doubt on prior research (i.e. Walton
and Sussman 1987) which posits that AMT adoption success is dependent on
high skills.

The alternative proposed in this research is that an “appropriate” choice of
skills will lead to higher levels of success than “inappropriate choices”. Choices
were deemed appropriate if management had matched skills to environmental
factors. In the remainder of the chapter the term FIT will be used to designate
installations with appropriate choices. Hypothesis H1 B which addresses
appropriate choices is supported (noting statistical limitations). There is
evidence that for this sample, installations where skills were matched to the
environment had higher levels of success than installations with inappropriate
skill choices.

Due to the nature of this sample, this result must be interpreted with
caution. The restricted range of the goal achievement measure suggests that
the correlation between FIT and success could be even higher. However, the
sample is small and limited to successful installations, making it difficult to

generalize the findings. Therefore it is suggested that strategic choice (Child

248

1972) of skills by managers is a possible explanation for the existence of
successful low skilled installations. This possibility needs to be addressed by
future research that includes failed installations.

A causal relationship between appropriate skills and success has not
been established in this dissertation. It is very unlikely that increased levels of
success would lead to increased fit. Such a proposition implies that firms do
things well and then make changes that increase fit. The notion that successful
firms make changes to increase fit seems unlikely, especially in light of the
literature on organizational inertia. For instance, Snell and Dean (1991) suggest
that successful companies may be less inclined to change structure or
infrastructure than unsuccessful firms. However, it is possible that the level of
product/process change may influence fit, in some situations. For instance,
when employees had abilities beyond the present requirement of the system,
management could change the product or process, which would increase fit and
might also increase goal achievement. However, overtime the result would still
be a relationship between fit and success.

In addition to the possibility that FIT may explain the existence of
successful low skilled installations, a number of industry effects suggest that
within specific industries high skills may be required. In this sample all of the
mold and die makers had high skilled employees. However, there are other
industries that have a large variance in skills, (i.e. machine tools) where skills are

not related to success. Finally, a few industries such as auto parts and diesel

249

power had inverse relationships between skill and success, suggesting that
additional training was not required.

These industry effects can also be explained by the product/process
change differences between industries. Dies and molds are generally made in
very complex manufacturing environments, while the auto parts makers were
generally operating in the simplest manufacturing environments (from the
standpoint of operator tasks). Therefore the industry effects are likely due to the
fact that companies in an industry have similar manufacturing environments. It a
company decides to compete in a different manner than industry norms, or to
use a different type of technology or level of outsourcing than their “average”
competitor, they then may need to adjust skill levels.

The results also show that firms with lower levels of discretion choose
lower skills regardless of the industry norrn. In sum the differences between
industries are captured by the variables of interest. It follows that industries such
as tool and die making with very complex internal environments will have higher
skills than industries where the shop floor is generally less complex, assuming
equal levels of discretion.

The first key finding of the dissertation is that low skilled companies can
be successful adopters of AMT. This result raises questions about previous
research (i.e. Meredith 1987) that suggests that low skills lead to failure. In
addition the analysis, while difficult to generalize from, indicates that FIT may
explain the existence of successful low skilled firms. The dissertation also points

to the possibility that those industry characteristics captured by the

250

product/process measure may help predict the “appropriate choice” for a given
organization.

The die makers as well as the mold makers provide an example of such
an industry standard “appropriate” choice. All eight installations in these two
industries used journeymen machinists to operate their AMT. These journeymen
had all been through the same four year apprenticeship program that included
4000 hours of classroom time as well as 4000 hours of on-the-job training.
Managers at each firm were convinced that the best way to meet the needs of
their systems (characterized by high product / proceSs change) was to use
employees who had completed recognized apprenticeship programs.
Employees with this training were assumed to be proficient. Journeyman status
was considered a baseline requirement in the hiring and / or promotion decisions
at these firms.

Industry standards may arise due to similarities in the levels of product /
process change across companies. However, each company is unique and
differs with respect to the types of products they outsource, the way they
compete, the companies they compete with, and their level of managerial
discretion. These difference can lead a company to choose a skill level that
differs from the industry norm. If the choice to differ is made based on real
differences in product/process change and/or managerial discretion, the results

indicate that goal achievement may increase.

251

5.4 Skill choices and appropriate choices

The size and bias of the sample makes strong conclusions about fit
suspect. However, there is some evidence that appropriate choices lead to
higher success levels (H1B).The remaining hypbtheses on the drivers of skill
choices were also supported. Combining the two sets of results points out a
potential discrepancy between the drivers of skill choices and the best choice of
skills. The analysis for H2 - H4 suggests that managers choose (the choice may
be explicit or implicit) skill levels based on the level of product/process change,
the level of environmental uncertainty, and their level of discretion. However, the
best FIT measure indicates that higher performance is achieved in installations
where skills are matched to product/process change and managerial discretion
only. Section 5.4.1 discusses the drivers of skill choices. Section 5.4.2 looks at
the various operationalizations of appropriate skill level that were used. Finally
section 5.4.3 looks at the important managerial implications of the discrepancy
between the drivers of skill choices and the best choice of skills.
5.4.1 Drivers of skill choices

The model proposed by this dissertation hypothesizes that three elements
of the environment influence skill choice: product I process change,
environmental uncertainty and managerial discretion. Product / process change
and environmental uncertainty were hypothesized to have direct effects on the
skill decision while the level of managerial discretion was hypothesized to
moderate the choice. The results support hypothesis H2 (product/process

change) and hypothesis H3 (environmental uncertainty) and provide partial

252

support to hypothesis H4 (managerial discretion). The regression model is
strongly significant ( R squared = .572, p < .0000) as are the beta coefficients for
all of the individual variables. However, managerial discretion has a direct effect
(not the hypothesized moderating effect).

The results therefore indicate that the companies in the sample have
made skill choices (implicitly or explicitly) based on the level of product/process
change, the level of environmental uncertainty, and the amount of managerial
discretion at the installation. However, the analysis of alignment and adoption
success suggests that the highest level of success may be achieved when
choices are made to match skill requirements only to the internal environment
(composed of product/process change and managerial discretion).

5. 4.2 Alternative operationalizations oLaippropriate skill level

The original formulation of appropriate skill level, FIT1, indicated that skill
should be matched to the level of environmental uncertainty, product I process
change, and managerial discretion. The results of the analysis, which must be
carefully interpreted due to the limited range of the dependent variable, indicate
that matching skills to these three factors results in a strong (r = .311) and
significant (p = .095 ) relationship between FIT and success. The literature
suggests that other possible forms of alignment may exist as well.

Wood and Albanese (1995) found that the choices of human resource
policies were driven by factors internal to the firm. Therefore, FIT2 was
operationalized such that choices were deemed appropriate when an installation

matched skills to the level of product/process change and the level of managerial

253

discretion, regardless of the external environment. FIT2 has a stronger
relationship with success than FIT1 (r = .383) and is significant at p = .037. This
suggests that the likelihood of success increases for installations where skills are
matched only to the internal environment.

5. 4.3 The drivers of skills a_nd the appropriate skill level

The results suggest that skill choices are made based on the internal and
external environment. However, the results also indicate that the best measure
of alignment (to predict success) examines the internal environment only. This
discrepancy is the third key finding of the dissertation. The results indicate that
for this sample the best performance is obtained by matching skills to the internal
environment only. These results corroborate previous studies by Wood and
Albanese (1995) and Jackson et al. (1989).

Much of the literature on organizational structure, built on Burnes and
Stalker’s (1961) seminal work, suggests that increases in environmental
uncertainty require increasingly organic structures. Organic structures are more
flexible and require employees with broader, easily modified jobs (hence higher
skills). The results of this analysis suggest that managers make their skill
choices based on the proposition stated in Burnes and Stalker’s work, but that
doing so may reduce goal achievement.

Manufacturing environments may be a response to the manufacturing
task (Skinner 1969, 1986) rather than the level of environmental uncertainty.
Figure 5-1 is a process map of a proposed model of the managerial choice of

appropriate skills. The map suggests that environmental uncertainty will affect

     

254

the choice of manufacturing strategy. And the choice of strategy will determine
what the key tasks of the manufacturing portion of the operation are. It is from
these tasks, as well as the level of managerial discretion that the most

appropriate choice of skills can then be determined.

Figure 5-1 A Process Map of the Appropriate Level of Training

 

 

 

IBusiness Strategy I IIManufacturing Strategy]
I ,

 

 

Manufacturing Task

 

 
   
   
 
 
   
   
 

Product/Process
Change

I High I Medium Low

Low High Low High Low
Discretion

Discretion Discretion Discretion Discretion Discretion
Medium
Skills

High Skills Low skills Low Skills
I

Low Skills: Minimal job specific training. Operator tasks will be limited to loading and
unloading the equipment, entering part numbers, and perhaps simple quality checks with
preset jigs and or fixtures.

 

 

 

     

 

High

     

     

 
     

Medium Skills: Comprehensive job specific training along with limited training on non-
job related tasks. Operator tasks can include loading and unloading equipment, setting
tools and fixtures, SPC or other quality checks, day to day maintenance activities, and
limited problem solving.

High Skills: Comprehensive training for job specific and non-job related tasks. Operators
tasks can include loading and unloading the equipment, choosing and setting tools and
fixtures, all quality checks, day to day maintenance and some trouble shooting of
equipment, active problem solving often in team environments, and at least limited
programming of the equipment.

255

For example a firm in an environment where customer demands are
constantly changing may decide that they need to compete by varying their
product mix more than the competition. This business strategy translates into a
manufacturing strategy focused on product mix flexibility. The key task for
manufacturing might then be to change from one product to another with
negligible set-up times. The system may have a large number of product
families and relatively small batch Sizes. The focus on mix rather than new
product introductions means that churn will be relatively low. The result is a
medium level of product/process change. Medium levels of skill would be
appropriate for the company if it had a high level of discretion, while low levels of
skills would be appropriate if the company had a low level of discretion.

Another company in the same environment may decide that the key to a
competitive advantage is not just to vary the mix, but to constantly introduce new
products. New manufacturing must be able to set-up quiCkly, but not just for
products with which they are familiar. Frequent new product introductions will
increase uncertainty in terms of operator tasks (over the previous company) and
increase the skill requirements. Because of different responses to the same
external environment the companies will require different levels of skills.

5.5 Industry effects

In section 5.3 there was a brief discussion of the relationship between
industry and skill choices. This section expands upon that discussion of industry
effects. The key conclusion from the previous discussion was that in general the

industry effect was a function of product/process change. In those industries

256

where product/process change was stable across installations there was often a
standard “appropriate” skill choice for the industry. Meanwhile, in those
industries where the level of product/process change varied significantly across
installations there was not a standard “appropriate” skill choice and the overall
model applied very well.

These differences can best be illustrated by examining two very different
industries: machine tools and stamping dies. The stamping die makers all have
very complex internal environments. All of the companies are using CNC
equipment to machine the faces of dies. Dies tend to have complex shapes and
are made one time. In other words every part is difficult to make and totally
unique. The model would predict that in such a situation of very high
product/process change all of the companies would choose high skills (which
they do).

The machine tool makers form a much more diverse group. These
companies were using AMT for a variety of tasks. Some plants have FMS
and/or CNC installations where literally thousands of different parts are run in a
single year. Other plants have CNC equipment dedicated to a single part or part
family, and have less variety and churn. The end result is that among the
machine tool makers there is a wide range of product/process change scores, as
well as a wide variety of skill choices.

The implication is that choices are made based on the three proposed
variables, and that good choices are still a result of looking at product/process

change and the level of managerial discretion. In industries such as the making

257

of dies, where there is little difference among installations, in terms of how the
AMT is used, it stands to reason that an industry standard “appropriate” choice
will evolve. However, in situations such as that faced by machine tool makers
where the AMT can be used for a variety of different tasks there is no expected
level of product/process change, and hence no standard choice will arise.

The fourth key finding of the dissertation is that even in an industry where
there is a defacto standard skill choice (such as the use ofjourneymen machinist
by all the die makers), management can improve performance by examining
their internal environment. Companies will outsource different activities and
parts, which will impact their level of product/process change. They will also
compete in different manners that may effect the level of customization. Finally
they will serve different customers who may place different demands on the
manufacturing system. The end result (see Figure 5-1) is that in some industries
there may indeed be a standard Skill level, but for companies whose
manufacturing environment is different than the industry norm, the standard skill
choice may be inappropriate. Therefore the industry standard, if one exists, may
be an appropriate starting point in determining skill requirements, but it is not an
end point.

5.6 Differences between FMS and CNC

FMS is composed of a number of linked CNC machines. However, the
level of integration of FMS makes the technology much more difficult to
successfully install. The analysis indicates that high skills are not a prerequisite

for successful adoption of either technology. However, the evidence on fit is not

258

as clear. CNC users have much higher levels of success when they match their
skill choices to their environment. The correlation between FIT1 or FIT2 and
performance of CNC installations is .488 (p = .021) which is a very large effect
when one considers the number of factors that influence adoption success.

It is probably the number of factors influencing success that explain why
there is no relationship between “appropriate” skills and adoption success for the
FMS installations. FMS does not require high skills to be successful, but the
increased integration of the technology may make adoption success a function of
many more factors than have been tested here. This premise is also given some
support when one examines the analysis of the drivers of skill choices.

The proposed model of the drivers of skill choices is a very good predictor
(Adjusted R squared = .685) of the choices made by the CNC installations. In
addition there is some evidence that managers at FMS installations are also
responding to similar cues when making their skill choices. Therefore, the users
of FMS are making choices in a manner similar to the users of CNC, but the
choice is not in and of itself a predictor of success.

The fifth key conclusion of the dissertation is that while neither CNC or
FMS require high skills, the proposed alignment model does not predict FMS
success. This may be because the complexity of an FMS makes success a
function of many more variables. This increased complexity may also help to
explain why FMS has not diffused very quickly.

The users of CNC equipment had more complex internal environments

than the users of FMS. This may indicate that a well designed CNC installation

259

can do everything that an FMS can, without the increased technological
complexity. This simpler technology allows companies to achieve the flexibility of
an FMS, in a simpler, lower cost fashion with a higher likelihood of success.

5. 7 Summary of key conclusions:

There are five key conclusions derived from the analysis:

0 There is no evidence of a relationship between skill level and adoption
success.

0 Appropriate choices may be an explanation for the existence of
successful low skilled installations.

o In general managers make skill choices based on both lntemal and
external factors. However, the best choices may be made based only
on the internal environment.

- Even in an industry where there is a defacto standard skill choice,
management may be able to improve performance by examining their
internal environment and adjusting skills accordingly.

0 While neither CNC or FMS require high skills, the proposed alignment
model does not predict FMS success. This may be because the much
higher levels of complexity of an FMS make success a function of
many more variables.

These conclusions indicate that for this sample the answer to the research
question proposed in chapter one is: there is a relationship between the
appropriateness of management decisions regarding the skill level of operational
employees in AMT installations and adoption success.

5.8 Limitations of the research

The results generally support the hypothesis proposed in chapter one.
However, there are limitations to the research. Chapter 3 presented four
potential limitations of the research: selection bias, the small sample size, non-

respondent bias, and the potential for respondents to have a preconceived

notion that a high skilled workforce is more desirable. This section will discuss

260

these limitations, as well as limitations imposed by the restricted range of the
goal achievement measure, and their impact on the generalizability of the
research.

The sample is clearly a convenience sample and not a random
representation of the larger population of AMT users. Despite this limitation, the
generalizability of the study may not be seriously threatened by this bias. There
are two different technologies represented in the sample and there is evidence
that there are differences between CNC and FMS. The finding that high skills
are not required for adoption success can be generalized across these two
technologies, while the appropriate choice of skills may be better suited to
“simpler” technologies such as CNC. Additionally, firms in a variety of industries
were examined and there is evidence of both industry and generalizable effects.
Finally, the method of data collection mitigated some generalizability concerns by
having the benefits of the valid theory developed through a case study. The
ability to make comparisons through the structured interview questions also
improved generalizability.

A second concern is the sample size. The small sample size makes
statistical tests questionable. However, the use of visual tools lends credence to
some of the findings. Also, the existence of multiple installations that are
successful low skill users of AMT provides compelling evidence that high skills
are not required for adoption success. Finally, the in-depth treatment of a small
number of installations was much more enlightening than a broader study with

less depth. The in-depth nature of the study allowed for discussions not possible

261

with a mail survey. The design also allowed for the collection of multiple
measures of each construct, which supports the validity of the results.

A third concern is the possibility of respondent bias. The users of AMT in
the primary sample differed from the sample of manufactures in general. At
some level this is to be expected, since AMT is a relatively new way of producing
parts. However, the differences were not in the manufacturing realm. Instead,
the respondents in the primary sample tended to be larger firms with predictable
external environments. Therefore the results may not be generalizable to very
small plants, or plants in extremely unpredictable environments (note that there
is at least one very small plant with a highly uncertain external environment in the
sample which fits the predicted model).

A fourth concern was that many firms may view specific workforce
practices as either cutting edge, more professional or perhaps socially desirable.
Scott (1987) notes that firms may adopt an innovation because they feel it is the
most professional or somehow the best choice. Respondents may claim that
they have adopted practices when in fact they have not. This problem appeared
in discussions of skill levels. Many recent works (Walton 1985, Saraph and
Sebastian 1992, Arthur 1992) make a compelling case for using a commitment
strategy for human resources in all settings. Respondents may have believed
that this was the response that the researcher was looking for and hence
provided socially desirable answers that did not reflect actual practices. The use
of multiple respondents and the tour of all facilities brought the “problem" to the

researcher’s attention. Generally a few focused questions would clear up the

262

confusion. What often had occurred was the respondents would discuss either a
pilot project or something they were planning, rather than what was occurring in
the plant at that time.

The final limitation of the research is the lack of failed installations in the
primary sample. The focus on successful firms was not an intentional part of the
research design. However, the end result of this sampling method was that all of
the respondents viewed themselves as at least somewhat successful users of
AMT. Therefore it is difficult to come to any conclusions about AMT failure. In
addition this limited coverage restricts the range of the goal achievement
measure, limiting the validity of the statistical tests of fit.

This research is limited by the small convenience sample that is probably
not representative of AMT users in general, especially in light of the absence of
failed installations. The literature suggests that failure among AMT users in
common, however they failures may be difficult to study because they will only
be in existence for a short period of time. However, the research design may
limit the effects of some of the other issues raised in this section. Multiple
respondents decrease the likelihood of respondent biases. And the coverage of
eight industries, means that many different environments were included to test
the model. Finally the existence of successful low skilled CNC and FMS
installations can not be denied. Therefore alternative explanations of how the
workforce affect AMT adoption success are needed. The appropriate choice
model tested here is one such alternative that may or may not stand up to

additional testing.

263

5.9 Future research and conclusions
5. 9.1 Future research

The results of the dissertation raise questions about existing paradigms
regarding the skill level of operational employees in AMT installations. However,
the limitations of the research that are detailed in section 5.8 make any
generalizations about appropriate choices difficult. Therefore future research
should attempt to address the issue of appropriate skills in a manner that will
address most of the limitations of the present work.

The first step towards this goal would be to study failed installations. The
results of this dissertation are based on a sample of successful companies. The
literature (i.e. Snell and Dean 1992) suggests that failures are common. The
experience of this research suggests that failure may not be as common as
previously thought. However, it is highly unlikely that all companies successfully
install AMT, even if the failures are rare. Therefore, the results of the
dissertation would be made much more compelling after an examination of failed
installations.

At a more fundamental level is the question of what exactly constitutes
success? It is possible that the absence of failed installations in this sample (as
well as the preponderance of failed installations in other research) is a result of a
flawed understanding of success. It is suggested that an ethnographic study,
where the researcher participated in the adoption of an AMT, would aid greatly in

developing both a deeper understanding of practitioner definitions of success, as

264

well as potentially developing more useful measures of success (including pre
and post hoc measures).

The process map presented in Figure 5-1 suggests that the external
environment affects skills indirectly based on how the organization responds to
their environment. Future research should explore the external environment in
more detail. Arthur (1992) linked human resource strategy to business strategy.
Future research could expand Arthur’s research to additional industries, and
more importantly look at how various manufacturing strategies are related to skill
choices. A final external driver that should be examined is the role of supply
chain management in uncertainty reduction. Future research could examine how
(or if) firms use their supply chain to manage their level of uncertainty.
Reductions in uncertainty through supply chain management could translate into
simpler manufacturing environments and hence lower skill requirements.

The dissertation suggests that a number of other areas should also be
explored. The issues were raised in either the data collection or data analysis
stages of the process. Two key issues raised during the data collection effort
involved measuring the performance of the manufacturing system and instituting
teams in machining environments.

A number of the installations that were visited were using performance
measurement systems for their AMT that were not well matched to their stated
goals for the system. The most common measure for AMT success was a
utilization and/or up-time measure. However, the majority of installations were

interested in quality and/or flexibility more than cost. Measuring up-time as a

265

indication of success encourages long production runs or the avoidance of things
such as scheduled maintenance. The end result is a measurement system that
is dysfunctional, because what is measured is not what the company is trying to
achieve. Future research should address the types of measurement systems
used in AMT installations, and how measurement affects both the overall
achievement of the system as well as achievement of Specific goals.

The second issue raised during data collection was the difficulty
experienced by many plants in instituting teams in machining environments.
Unlike assembly areas where the work of one person is clearly linked to the next,
machinists often work in isolation. This isolation makes it difficult for teams to
function because there is no interdependence among team members. Future
research should address both the appropriateness of teams in different types of
machining environments and how to help insure the success of teams in
environments where they are appropriate.

There were a number of interesting findings from the data analysis effort
itself that should also be addressed in future research. Key issues include: the
relationship between internal and external environments, the lower level of
environmental uncertainty for AMT adopters, and the examination of the
successful installations that do not fit the suggested model.

Bluedorn (1993) notes that the concept of fit or alignment has been
addressed by the strategic contingency school of thought for nearly thirty years.
The basic premise of this organizational school of thought is that when various

parts of a firm’s structure are aligned the firm will perform better. The evidence

266

from this dissertation is that there is no alignment between the perceived
external environment and the internal manufacturing environment, for companies
that are generally meeting their manufacturing goals. This finding may provide
some support for early organizational research (i.e. Lawrence and Lorsch
(1967)) that suggests that manufacturing may be buffered from the firms external
environment. This dissertation suggests that rather than being buffered,
manufacturing environments differ from the external environment. Future
research should address the performance implications of matching
manufacturing structure and infrastructure to the external environment.

In addition to looking at the relationship between the internal and external
environments; future research also needs to address a counterintuitive finding of
this dissertation; namely that the adopters of AMT perceived their environments
as more certain than the sample of manufactures in general. The flexibility
literature (Swamidass and Newell 1987, Genrvin 1993) suggests that increases in
environmental uncertainty lead firms to increased flexibility and hence the
adoption of AMT. For this sample, that is not the case. Future research should
therefore address the linkage between environmental uncertainty and adoption
of AMT for a representative sample of AMT users. In addition, future research
should address the possibility that it is not uncertainty, but rather other
environmental factors (Dean and Snell 1996) that drive the AMT adoption
decision. Finally, as noted by Boyer et al. (1996) there are different types of

AMT, as well as different types of adoption patterns. Future research should

267

address the possibility that either specific AMTS and/or specific investment
patterns are used in response to specific environments.

Finally, as noted in Chapter 4, there are installations in the primary data
set that do not fit the proposed model. Specifically, one installation had productl
process change of -5.5 and total preparation time of 110 weeks. This
installation, which makes glassware, does not fit the proposed model at all. One
possible explanation is that this is a very unique use of CNC controls which
requires very site specific training. Because CNC controllers are being use in a
number of non-machining environments it is possible that the literature used to
build the model, which is based primary on machining, is not applicable to other
environments. Therefore it is proposed that part of future research efforts should
address the applicability of the model in non-traditional AMT environments.

5. 9.2 Conclusions

This dissertation examined the impact of one managerial decision on the
success of AMT adoption. In contrast to previous research (i.e. Walton and
Sussman 1987) there is no evidence that high skills are required for adoption
success, a finding that is supported by the existence of a number of successful
low skilled installations.

An alternative offered by this research is that installations where
management has (implicitly or explicitly) fitted skills to the environment will be
more successful than installations where there is no such fit. The conclusions
available from the data analysis are limited by the sample, but there is some

evidence that fit may improve the level of goal achievement. More specifically

268

the evidence suggests that installations are most successful when skills are fitted
to the internal environment composed of product/process change and
managerial discretion. Therefore it is proposed that managers who follow the
process map presented in Figure 5-1 may increase their probability of
successfully installing these technologies.

In sum, the dissertation results indicate that previous research which
suggested that high skills were a prerequisite for adoption success must be
questioned. Researchers and managers who are examining the issue of what
types of human resource policies will be required of new technologies and/or
processes should examine the fit between the planned manufacturing

environment and human resource policies.

APPENDICES

APPENDIX A

FACTORS THAT IMPACT THE SUCCESS OF AMT ADOPTION

APPENDIX A

FACTORS THAT IMPACT THE SUCCESS 0F AMT ADOPTION

Success Factor

 

 

Wanagerial focus rather
than a technical focus -
importance of

infrastructure in general
and people in particular

I Exnlanation Imme-

in general managerial
problems are more
prevalent, harder to
solve, and more likely to
inhibit installation than
technical problems

Shani, Et. Al. 1992 I
Meredith 1987 (p)!
Gerwin 1982 I Elango
and Meinhart 1994/
Gupta et. al. 1993]
Bessant 1993 l Boer, Hill,
and Krabbendam 1990/
Adler 1988

 

Integration across
functions

Installation of FMS
requires input from all
functions because the
technology will impact
more than just the
manufacturing part of the
organization

Shani, et. al. 1992 I Boer
and Krabbendam 1992/
Knorr and Theide 1991 /
Diaz 1991 I Goldhar and
Lei IT ranfield Et. Al. 1991
ISpano 1993 I Bessant
1993 I Boer, Hill, and
Krabbendam 1990

 

Purchasing: integration

suppliers of inputs should
be included in integration
efforts

Shani, Et. al. 1992/
Knorr and Theide 1991 /
Fry and Smith 1989/
Elango and Meinhart

 

to its sensitivity

1994 [Adler 1988 / Zairi
1992
purchasing supplier the technology demands Boer and Krabbendam
quality higher quality inputs due 1992 I Knorr and Theide

1991 IMieskonen 1991 I
Boer, Hill, and
Krabbendam 1990

 

Top Down (top
management support)

because of the capitol
intensity of adoption, only
top down programs work,
and need top
management support

Shani et. al. 1992/
Gupta et. al. 1993/

 

Phased Approach to
installation

 

the islands of automation
view - instead installing
everything at once we
install small sections at a
time tofiget familiar with

 

 

Shani et. a. 1992 l Miller
and Gilbert 1992 I
Gerwin 1982/
Vonderembse and
Wobser 1987 / not an

 

269

 

270

 

the new technology

impact according to

 

Beatty 1993
Knowledge workforce - a a catch all for the various Boer and Krabbendam
commitment or strategic but convergent views of 1992 / Walton and

 

HR focus to the the way the workforce Sussman 1987 / Goldhar
workforce (Arthur 1992/ should be organized. and Lei 1994 l Tranfield
94) Authors taking this view et. al. 1991 [Gerwin
see FMS operators as 1982 I Gupta et. al. 1993
multi-skilled, flexible, and l Prickett 1994 / Smith et.
working in teams al. 1992 lAdler 1988/
Gupta and Yakimchuk
1989/ Saraph and
Sebastien 1992 I
Majchrzak 1988
Software poor software decision Meredith 1987 I Fry and

especially when it comes
to interfacing various
components can be
disastrous

Smith 1989 lGupta et. al.
1993

 

Tool rules I control

dealing with the costs
and complexity of tool
decisions

Meredith 1987 I Fry and
Smith 1989 lGupta et. al.
1993

 

Technology Strategy

related to integration. the
technology should be
installed to meet specific
goals that span functions
-not piecemeal or
because we want an
FMS

Vineyard 1993 [Adler
1988

 

Fit with strategy and
products

there should be a link
between the firms
strategy and the chosen
technology as well as a
fit between the product
mix and the technology

Vineyard 1993 / Adler
1 988

 

Process Champion

Successful adoption
requires a factory
automation champion
who pushes the project
to completion

Meredith 1986 l Gerwin
1982 / Beatty 1993
(present at all successful
installations studied)

 

 

Information Systems

 

methods of gathering and
communicating data are
a key to success

 

Attaran 1992 / Badiru
1990

 

 

271

 

Push verses pull

is the technology being
pushed upon the
organization to test it for
a larger company or the
government, or is there a
market need pulling the
techncfly ?

Groth 1993

 

Preventive Maintenance

because the technology
is so complex it is less
expensive to do
preventive maintenance
than to wait for it to break
- the maintenance should
be done by the operators
to increase their
awareness of the
process

Diaz 1991 [Jackson and
Wall 1991 / Boer, Hill and
Krabbendam 1990/
Majchrzak 1988

 

Time Between
technological changes

The longer a plant goes
with the same technology
the harder change

Schroeder, Congden,
and Gopinath 1995

 

becomes
Union management unions and management Walton and Sussman
cooperation must work together to 1987 / Gupta and
make the new technology Yakimchuk 1989

succeed

 

Changes in Performance
measurement / cost

Measures have to reflect
the new group based

Genrvin 1981/Gerwin
1982 I Saraph and

 

 

accounting structures as well as the Sebastion 1992 I Tayles
fact that labor efficiency and Drury 1994/
is no longer important Coulthurst 1989 I Currie
1992 / Maskell 1989 /
Son 1990 / Fry 1992
Selection There are a number of Hunter, Schmidt and

 

reasons that selection of
workers becomes more
important including the
need for new skills, and
the costs of failure

 

Juduesch 1990 I Walton
and Sussman 1987

 

 

APPENDIX B

VARIABLES THAT IMPACT THE STRATEGIC CHOICE OF SKILL LEVELS

APPENDIX B

VARIABLES THAT IMPACT THE STRATEGIC CHOICE OF SKILL LEVELS

_Imlm

Complexity of products I
level of conversion
uncertainty

more complex products

lead to higher skills I the
more change the higher
the needed skill level

Kelly 1990 (complex defined
as batches of less than 10)
Zicklen 1987 (smaller batch
equal higher skill) / Jackson
and Wall 1991(system
variance) / Milkman and
Pullman 1991/ Brass 1985
IHazeIhurst, Bradbury and
Corlett 1969 (batch size)/
Jackson and Wall 1991/ Clegg
and Wall 1987

 

NC Share - percent of
firm using programmable
automation

the more automation the
less likely workers are to
program their own
machines - more
automation leads to
deskilling

Kelly 1990

 

Unions

lead to deskilling

Clegg and Wall 1987/Kelly
1990 I Chaykowski and
Slotsve 1992 /Wood and
Albanese 1995

 

Unions with seniority
systems

deskilling

Kelly 1990(skills =
programming)

 

 

more likely to follow a
deskilling model

professional decentralization of Kelly 1990
management decisions and skill
upgrading
Size (deskill) larger organizations are Kelly 1990/ Clegg and Wall

1987

 

Size (skill upgrading)

large organizations have
the resources and
knowledge to institute
“advanced” programs

Dean and Snell 1991

 

recent adapters

 

 

more recent likely to
follow skill upgrading

 

Kelly 1990 / Meredith 1987

 

272

 

273

 

Uncertainty or task
complexity

indicator of policies
linked to skilling

Dean and Snell (1991)/
Hunter, Schmidt and Judiesch

~ 1990/ Reshef 1993

 

High firm performance

leads to skilling because
the resources exist

Dean and Snell 1991

 

Point in Product Life
cycle

In intro. and growth
stages we focus on
performance of the
product and increase
skills, as the product
(process) matures we
increase control to lower
costs

Smith 1992 I Schler 1989

 

Supplier involvement
(vendor)

The vendor of the
technology has a large
impact on job design

Preece 1989 / Thomas 1991

 

Skills in labor market

if high skill workers do
not exist or we can not
pay them, we deskill

Bessant (in Preece 1989 p.
26) Davis 1986, Buchanan
and Boddy 1983/ Osterman
1987/ Burnes 1989/ Clegg
and Wall 1987

 

Better relations equal a
better chance for skilling

Thomas 1991 / Smith 1992/
Davies 1986/ Burnes 1989

 

Maturity of technology

mature or debugged
technology does not
pose the same
challenges and does not
need the same skills

Zicklen 1987

 

 

Environmental variability
(uncertainty) and
complexity

 

Frequency of changes /
degree of difference
between changes /
degree of irregularity in
the pattern of change
(according to child 1985
these are the key
reasons for differences in
org. structure given that
tech. is the same -they
also link well to the
reason given for using
skilling with FMS)

 

Child 1972 / child 1985 (with
additional cites) lburns and
stalker. Lawrence and Lorsch
-etc. [Well and Davids 1992

 

 

274

 

Age of workforce

older workers (over 45)
are far less likely to want
their skills upgraded -
long tenure induces a
dislike of change - may
be a big problem with
union environs

Kochan, McKersie, and
Cappelli 1984/Gupta

. 1989

 

Business Strategy

low cost producers use a
deskilling (containment
strategy) while
differentiators use a
commitment (skilling )
strategy

Arthur 1991/ Schluer
1989

 

Cost of Errors

The higher the cost of an
error, regardless of the
actual complexity of the
job, the more likely a firm
is to want high skilled
workers so that they help
prevent errors

Osterman 1987

 

 

Time since last change /
established social
relations

 

the longer we have done
things this way the less
likely we are to be able to
change

 

Schroeder, Congden,
and Gopinath 1995
Walsh 1991

 

 

APPENDIX C

INTERVIEW PROTOCOL

APPENDIX C

INTERVIEW PROTOCOL

Questions for Operations Manager:

 

 

Respondent information

1. Name:

2. Title:

3. Tenure in position and tenure in company:

4.

 

Background:

General Company lnfonnation:

u—L

999°.“P’9‘PP’N

_\
.o

11.
12.
13.

Company / plant name:

 

Plant location:

 

Plant size (employment)

 

 

 

 

Workforce composition Average age: Average Tenure:
Plant sales (last three years) 95 94 93
Plant age:

 

Union status: yes or no and if yes number of unions:

 

What products are made at this plant?

Are there other plants within the company that do the same or similar
things? yes / no

If the answer to 9 is yes, what type (s) of production technology do they
use?

What type of markets do you sell your products in?
Would you describe your firm as, MTO, MTS, ETO, other?

Is demand cyclical? seasonal? if so what is the deviation per quarter?

275

276

Specific information about the AMT

1. Please describe the shop floor technology (in plants with both AMT and non-
AMT technologies the respondent will be directed to focus on the AMT):

2. Has your firm / plant installed similar technologies before? If so how many?

3. When was the technology installed?

 

4. Why was the Technology installed?

5. Please describe the planing and installation process used for the installation
(who was involved in planning, how was installation carried out).

6. Were the operators of the equipment retrained when the AMT was installed?
If so please explain the type and duration of training.

7. What type of products I parts are made on the AMT?

8. Who is responsible for programming the AMT? How is this accomplished?
Why?

9. Who is responsible for determining quality and how is this accomplished?
10. Who is responsible for set-ups and tooling on the AMT? Why?

11. How is the work for the AMT organized? Are operators in teams? Cells?
Quality Circles? Other arrangements?

12. Who is responsible for maintaining the AMT? Why?

At this point it would be requested that the plant tour take place (if they are
willing) - a tour before getting into details will allow better questions regarding
responses than would occur if the tour came second

Post tour questions:
Product process measures:

The following questions deal with the products made using the AMT or its
operators. If there are products that run in the plant but do not require any input
from the AMT and/or its operators please exclude them from your answers.

1. How many distinct product families are made on the AMT?

 

Note how respondent defines a product family and the actual number of
families as well as the category.

277

. What is the average batch size for the AMT in hours of supply? / in actual

part counts? I

Note: the respondent will be asked for an actual number as well as a
definition of batch size to be sure that both respondent and researcher view
the concept the same way

How much churn (new part/product introductions/ old part/product retirement)
in percentages is there among the parts/products made on the AMT?

new parts/products as percent of part/product mix

 

retired parts/products as percent of part/product mix

 

Note: the respondent will be asked to define introduction and retirement to
be sure that minor changes or changes in name only are not included. Actual
percentages will also be recorded

What is the percent deviation from the master production schedule in volume
in an average week ?

 

How many alternative processes or routes are there in the production
system?

 

Note: respondent will be asked to explain various alternatives if they exist.

How often are engineering changes made on products I parts that have been
released to the shop floor ?

never rarely often always
1 2 3 4 5 6 7

Note: engineering changes will be defined by respondent, and respondents
will be asked to explain the magnitude of these changes, as well as their
frequency. Finally respondents will be asked to explain why EC’s occur.

External Environment questions:

1.
2.

How many distinct customers do you sell to and or service?

 

How many competitors do you have in the markets you serve?

 

How predictable are the actual users of your products ?
never usually not 50% usually always

1 2 3 4 5 6 7

3.

278

How predictable are the competitors for your supply of raw materials?
never usually not 50% usually always
1 2 3 4 5 6 7
How predictable are the competitors for your customers?
never usually not 50% usually always
1 2 3 4 5 6 7
How predictable is government regulation controlling ygg industry?
never usually not 50% usually always
1 2 3 4 5 6 7

How predictable are the public’s political views and attitudes towards your
industry?

never usually not 50% usually always
1 2 3 4 5 6 7

How would you define your firms external environment, and why? What
factors make it ?

Does the firm have a formal grievance procedure? If so do you feel the
number of grievances are:

excessive about average less than average
1 2 3 4 5

Note: questions regarding number of grievances, strikes related to the AMT,
as well as for the entire plant will be asked but it is possible that this person
will not have the information.

if union: How often does the union (for each individual union) inhibit your
ability to make decisions?

never usually not 50% usually always

1 2 3 4 5 6 7

279

10. How would you describe management’s relationship with the shop floor

employees? Please explain.
combative

somewhat distrustful
somewhat trustful

trustful

collaborative

partners

Adoption Criteria and Success

1.

Please rank the following goals (for the AMT) in order of importance with a
brief explanation of how the AMT was expected to impact this criteria. Note
areas that were considered of no importance can be ranked zero Ill!!!

quality

 

cost I efficiency

 

delivery I responsiveness

 

Flexibility

 

Innovafion

 

Other

 

For each of the above goals rank the success of the installation in meeting it
(1 totally unsuccessful - 7 totally successful) and explain.

goal 1:
goal 2:
goal 3:
goal 4:
goal 5:
goal 6:

9"

599051.03

280

Did the installation have any unintended consequences? If so what where
they?

Overall how successful do you consider this technology to be? Why?

Please provide the following information for the plant for the last three years
(all information is confidential) '

ROI

 

Sales % increase/decrease

 

Profits

 

 

 

Market share for major product made on AMT

 

 

Miscellaneous:

1.
2.

What role did the suppliers of the AMT play in installation design?
Did the technology supplier provide options for the user interface?

Do you have a service/maintenance agreement with the supplier of the
technology or another company? Why?

How would you describe the way your firm competes?

How do you measure performance of the AMT? Does this differ from the
measures you use for the entire plant?

How has the introduction of programmable equipment effect workers skills?
(e.g. Doug Hartmens Comment on craftsmanship) How about centralized
programming?

If you were to install another, similar piece of equipment, what would you do
differently?

QUESTIONS FOR HR MANAGER:

Respondent information

1.

2.

Name:

 

Title:

 

3.

4.

281

Tenure in position and tenure in company:

 

Background:

Questions about the shop floor technology

1.
2.

Why Was the Technology installed?

Please describe the planing and installation process used for the installation.
(Who was involved in planning? How was installation carried out?)

Who is responsible for programming the AMT? How is this accomplished?
Why?

Who is responsible for maintaining the AMT? Why?

Education and Training: the following questions address the time it takes
for an average employee to become proficient at theirjob. Please include all
training or education that is relevant to the job the operator is presently
doing, but not developmental education that will allow the operator to move
to another job. Additionally if there are different combinations of types of
training please discuss all “normal” options. In other words if you are willing
to hire people with either 2 years of experience or a vo-tech degree note
both options. Finally if you have a formal job description a copy would be
appreciated.

What is the average level of general education reguired to perform
operational tasks (of the AMT) at an average (acceptable) level? (space
given for explanation)

How much vocational training (in months) does an average operator require
to perform operational tasks at an average (acceptable) level? Please
include any training that is required for the operator to be hired, not training
that occurs after the operator is hired. Vocational training may include
specialized schools, union apprenticeship programs and or other job training
programs.

How much formal training (in hours) does an average operator receive to
perform their present job at average (acceptable) levels? Please include all
time that is specifically set aside for training after the operator is hired. This
may include classroom sessions, apprenticeship classes in the evening and
or time spent following another operator. Please explain type of training as
well as who performs the training.

10.

11.

12.

13.

282

How much experience (in months) are average operators required to have to
be hired/assigned to the job? Please explain the type of experience.

How much continuing education (hours per year) does the company
request/require of the operators?

Please rank the types of training in order of importance:

general education
vocational training
formal on the job training
expenence
continuing education

 

 

 

 

 

Please explain the reasons for the rankings:

How often do you have trouble finding employees with the skills you need in
the local labor market?

never rarely often always
1 2 3 4 5 6 7
Is it difficult for you to keep employees once they have reached proficiency?
never rarely often always

1 2 3 4 5 6 7

Industrial Relations

1.

4.

If unionized, what union represents the operators of the AMT and how long
have they been organized?

If unionized are there other unions in the plant besides the one mentioned in
1 above?

Is there a seniority system in place in the plant? for the AMT?

Note: respondents will be asked to explain how the system works. Is it
straight time in or are there specific jobs where you must have a certain type
of training regardless of time (and who has access to the training)?

If no seniority system, how are promotions and job vacancies handled?

10.

283

When hiring from outside the plant how are workers selected (by job specific
skills, by experience, by ability to learn, by work group, etc)?

Does the firm have a formal grievance procedure? If so do you feel the
number of grievances are:

excessive about average less than average
1 2 3 4 5 6 7

How many grievances for the work area where the AMT is installed were
there the year the technology was installed? in the year before the
technology was installed? in the present year?

Has there been a strike or slowdown over the installation of the AMT? Over
work Practices for the AMT after installation?

If union: how often does the union (for each individual union) inhibit your
ability to make decisions?

never rarely often always
1 2 3 4 5 6 7

How would you describe management’s relationship with the shop floor
employees? Please explain.

combative
somewhat distrustful
somewhat trustful
trustful

collaborative
partners

Specific questions for HR manager:

11.
12.
13.
14.

Are the operators of the AMT paid hourly or a salary?
Do the operators have any pay at risk?
How is operator pay determined?

Are the employees operating the AMT compensated differently than other
shop floor employees?

15.

16.

17.

18.

19.

284

How is the performance of AMT operators appraised? Does this differ from
other shop floor employees?

Do the operators work in teams? And if they do, are there specific rewards
for team performance?

How are jobs on the AMT assigned? Does this differ from other jobs
assignments in the plant?

How were the jobs for the AMT designed? Who took part in these
decisions?

What do you think the effect of programmable equipment has been on the
shop floor employees? (Noting Doug Hartmens conclusions) If they have
centralized programming of equipment ask about that as well.

APPENDIX D

CONSTRUCT VALIDATION INSTRUMENT

APPENDIX D
CONSTRUCT VALIDATION INSTRUMENT

Name: Company

 

Position / title

 

Number of employees at location

 

Would you like a summary of this study’s results? yes no
Would you like a summary of follow up studies? yes no

Please answer all questions for the facility (plant) that the questionnaire was
mailed to.

1. What SIC code(s) best describe the products you make?

 

2. How would you rate your plant’s performance relative to others in your

industry?
below average average above average
1 2 3 4 5 6 7

3. What type(s) of production technology are used in your plant? Please
indicate the percentage of production (in $ of volume) produced on each
type of technology?

 

 

 

 

Technology Percent of $ of volume
Technology Percent of $ of volume
Technology Percent of $ of volume

 

 

4. Is your plant unionized? yes no

5. How do you promote operational employees?
seniority performance Other

6. How many distinct part families or product lines does your plant produce? _

7. What is your average batch size (for the most important type of technology
by $ of volume) in actual number of parts or products?

 

8. What is your average batch size (for the most important type of technology
by $ of volume) in hours of supply?

 

285

10.

11.

12.

13.

14.

15.

16.

17.

18.

286

How many new parts/products (as a percent of existing part/product mix) are
introduced on the shop floor in an average year?

 

How many parts/products ( as a percentage of existing part/product mix) are
retired from the shop floor in an average year?

 

What is the average percentage deviation, in volume, from the master
production schedule in a week ?

 

What is the average number of product and/or process changes made in a
month ?

 

Product and process changes are:

Always Predictable Never
Predictable 50% of the time Predictable

1 2 3 4 5 6 7
The actual users of your products are:
Always Predictable Never
Predictable 50% of the time Predictable

1 2 3 4 5 6 7
The competitors for your supply of raw materials are:
Always Predictable Never
Predictable 50% of the time ‘ Predictable

1 2 3 4 5 6 7
The competitors for your customers are:
Always Predictable Never
Predictable 50% of the time Predictable

1 2 3 4 5 6 7
Government regulation controlling your industry is:
Always Predictable Never
Predictable 50% of the time Predictable

1 2 3 4 5 6 7
The public’s political views and attitudes towards your industry are:
Always Predictable Never
Predictable 50% of the time Predictable

1 2 3 4 5 6 7

APPENDIX E

FACTOR ANALYSIS

APPENDIX E

FACTOR ANALYSIS

Table E-1 shows the factor loadings of the exploratory factor analysis

done using all of the product/process variables and all of the environmental

uncertainty variables. This test was done to ensure that there was not a factor

that was composed of variables from both of the proposed measures. As the

table shows the product process variables do not load on any single factor, nor

do they load on the external environment variables.

Table E-1 Exploratory Factor Analysis

Families
Batch
Size
Churn
MPS
deviation
Q14
Q15
Q16
Q17
Q18

Tables E-2 and E3 respectively show attempts to force the

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Factor Factor Factor Factor Factor Factor Factor Factor
1 2 3 4 5 6 7 8
-.004 -.174 .147 .004 .339 -.008 .009 .007
-.211 .277 -.302 -.021 .055 .181 .056 -.010
-.162 -.124 .548 .037 .020 .106 .045 -.011
.051 -.479 -.373 .125 .096 .011 -.035 .011
.488 -.412 .004 .096 -.148 .049 .110 .005
.874 .210 .090 .063 .037 -.057 -.035 -.018
.680 .012 .052 -.346 .058 .083 .008 .009
.308 .199 .015 .399 .025 .092 -.053 .013
.050 .210 -.100 .107 .083 -.097 .180 .000

 

product/process variables into a single factor and attempts to force the

product/process and environmental uncertainty variables into 2 distinct factors.

In both cases explicit factors do not emerge and the results indicate that both

measures are indexes rather than scales. In other words the various

287

 

288

components are not really measuring the same underlying construct. Rather, the
indexes are measuring facets of the environments which do not necessarily vary
with other facets of the environment. However, there is no indication that any of
the variables that make up the indexes vary With the variables of the other

index, which reinforces the concept of these indices being separate entities.

Table E-2 Loadings For “Forced” Single Factor

Load
Families -.140
Batch Size .189
Churn -.742

MP8 Deviation .158

 

Table E-3 Loadings For “Forced” 2 Factor Solution

Factor 1 Factor 2
Families -.003 .125
Batch Size -.195 . -.239
Churn -.124 .088
MP8 Deviation .032 .313
Q14 .492 .506
Q15 .966 -.257

Q16 .571 .012
Q17 .279 -.140
Q18 .050 -.196

 

APPENDIX F

CORRELATIONS CONSTRUCT VALIDATION STUDY

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oe mmm. 2b. Em. En. N8. mvo. 30. Box vmrr mm:
o._. 50. moo. wwo. $0.. 050. 9.0.. omor omor m3.
09 $0. mmm. Noe. mvor 50.- vmo; 3.0.- moor
oe mom. omm. mmor oFo. moo. mgr mmor
oe vmm. mmor to: 30.- NE..- m9. -
oe mmm. mew. omo. 30.- SW.
C; Nvm 5m. m5. vmmr
oe Nmo. m3: mmor
09 0:. m3:
09 m3:
0;
\EonEoo wmoooi
_mEonw med :0 90 30 :0 Heaven. co=m_>mu was. mo___Emu_ Ezco mum 52mm

Xm
30
:0
90
30
:0

mm.

was.
Emu.
Ezco
coamm

Nm "2 - a_aEmm cosmc=m> uosbmcoo Low xEmE cozflotoo

>DDhm zOF<n=._<> hoamhwzoo mZO_._.<._mm_m_OU

"— x_ozmn_n_<

289

290

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o ooo. ooo. ooo. ooo. ooo. moo. ooo. ovo. omm. E.
o mov. ooo. mmv. ooo. oov. who. ooo. «3. go.
o vov. moo. ooo. moo. mom. Foo. voo. ovo.
o ooo. ooo. mom. nmo. woo. mm... «mm.
o moo. ooh. Fov. Foo. mum. o9.
o oFo. Foo. ooo. ooo. ooo.
o moo. ooo. ooo. mom.
o Foo. owe. vow.
o Rm. om.
o ooN.
o
a_xanoo wmmooi
.9:me ma :0 o5 30 v5 632.”. co:m_>oo was. mm___Emu_ 25:0 Bow coomm

 

ma
:0
9.0
9.0
:0
n_n_
was.
Emu.
Ezco
seam

macaw :o_§__m> 82880 .2 5.5.2 55205

APPENDIX G

STATISTICAL CHARACTERISTICS OF CONSTRUCT VALIDATION
VARIABLES

APPENDIX G

STATISTICAL CHARACTERISTICS OF CONSTRUCT VALIDATION

 

 

  
 

 

 

 

  

 

 

 

 

 

 

 

 

   

    

 

 

 

 

 

VARIABLES
Variable Minimum Maximu Mean Standard Skewnes Kurtosis
m deviation 3
Batch Size 1.0 200,000 1031 1 27,229 5.1 29.4
| Churn 0 200 33.6 48.6 2.3 4.67
MPS 0 100 14.4 14.6 3.5 15.9
Deviation
Families 1 400 21.5 54.7 5.2 29.9
Product -7.5 8.8 0.00 2.415 .355 2.8
Process
Q14 1 7 3.3 1.7 .381 -.79
Q15“ 1 6 2.8 1.3 ..336 -.511
Q16 1 7 3.4 1.5 .280 -.529
017" 1 7 3.4 1.33 .391 -.114
018" 1 6 3.3 1.3 .127 -.426
External 6 25 16.1 4.2 -.088 -.627
Complexity

 

291

 

APPENDIX H
STATISTICAL CHARACTERISTICS OF PRIMARY SAMPLE VARIABLES

APPENDIX H

STATISTICAL CHARACTERISTICS OF PRIMARY SAMPLE VARIABLES

 

 

 

 

 

 

 

 

 

 

 

 

Variable Minimum Maximu Mean Standard Skewnes Kurtosis
m deviation 5
Batch Size 1 220,000 19,023 52,683 2.8 6.7
Churn 0 200 56.15 68.4 1.099 -.393
MP8 0 35 11.9 9.4 .609 -.44
Deviation
Families 1 5000 1 108 2018 1.4 .027
Product -7.95 6.53 .467 3.48 -.107 .090
Process
Q14 1 7 2.7 1.33 1.4 2.040
Q15 1 6 2.4 1.3 1.14 .883
(“ilk 1 6 2.6 1.58 1.19 .398
Q17 1 7 2.7 1.77 1.14 .129
Q18" 1 4 1.7 .944 1.33 .865
External 6 20 12.1 3.5 .614 -.309
Complexity
Total 1 342 120.01 105.9 .317 -1.3
Preparatio
n Time

(Skill)

   

 

 

 

 

 

 

292

 

APPENDIX I

CORRELATIONS FOR PRIMARY SAMPLE DATA

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3 now. to; 8? $4.. 8.0... 50... 80.- BQ. 8N. EN- QR. N8. 80
3 23.. So. 30.. Km. 88. 20.- 3.- SN- 3. SN. RN- wmo
3 Sm. $0.. 83. En. mi. mm: :8. Bo. 2v. 8m 0%..
3 SN. 5v. v5. 8v. 8m. EC. 80.- 80.- one. 08.-
3 mmo. mg. 9.3.- v8. v3- 8.0.. E. v8. om:
3 v8. omo- mm? 38. EN. 2: 08.- $0.-
3 02. van. 8m. 3. 93.. 9a. vow-
3 own. So. $3.- 38.- So. 09.-
3 83.- So. 80.- $9. 83.
3 .3. 9%. 8m. 80..
3 vs. BF. ST
3 owe. mm?
3 SN-
3
$5 E
:25me zooms 38258 $893 coomSwv mum
$826 .9585: 393:6 .9:me So to 90 20 30 8:85 was. mm___Emu_ EEG 22mm

3825
mews.
___xm
xm
90
to
9.0
m5
v5
on.
was.
Emu.
c520
coumm

om uZ - <._.<Q m.._n=2<w >m<s=mm mOm x_m_._.<_>_ ZO_._.<._wmmOO

<._.<n_ w._n=>_<m >m<s=mm m0". mZOFSWMKOU

_ X_Dzmn_m<

293

294

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o3 no? moo. mmo. tv. moo. ooo. Foo. mos. mum. ovw. on... woo. own.
o So. :8. vow. ov_.. Foo. owo. mom. 9N. omv. mow. 3N: Von.

o moo. omo. row. ovo. omv. oov. ooo. nos. 09. ooo. N3.

o 3. So. ooo. moo. roo. Non. v.8. moo. So. ooo.

o moo. ooo. omv. ooo. 3N. ooo. oom. 3N. Rm.

o one. mno. own. mum. ovr. to. o5. ooo.

o ooo. _.oo. FNF. Em. vmv. vo_.. oom.

o moo. woo. mom. o; moo. ooo.

o ooo. ooF. mom. ooh. ooo.

o ovm. woo. ooo. ooo.

o ooo. Rm. «mm.

o ooo. 9v.

oo ooN.

oo

a_xmano mmmooi coomsmu mum

$826 c8285 5% .2580 o5 to 90 20 v5 639a was. mo___Emn_ EEO 22mm

.3825

010—2
Exm
Xm
9.0
5.0
90

m _0
30
an.
was.
Emu—
5350
28mm

<._.<O m._n=>_<w >m<_2_mn_ m0”. X_m._.<_>_ >._..__m<mOmn_

APPENDIX J

ASSESSING FIT FOR THE 30 RESPONDENTS

APPENDIX J
ASSESSING FIT FOR THE 30 RESPONDENTS

Company Skills Product External Discretion ' ' ' '
Process Complexity . 1 2 3 5

m
=6-
11
=-
m
3
m
3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

22 ll

 

 

 

—7_‘——_—_T'TC_—TT—_
: a

”I

 

 

 

24
25
“II
27

 

28"

 

 

 

 

 

 

 

 

 

E
IIIIIIgl—ZZIl-l—Il—III—gl-IIgl-ggzll-g
IIIIIIi—ZI—IIZr-II-IIZZI—IZi—Zr-i-r-gr-r-
r—I-r-r-r-r-I-r-r-i-r-r-i-r-r-Ir—r-Ir-i-II-i—r-i—Ir-r-I—
IIIIIIPIIIIl—l—l—l—IIr-l—IIf—Il—IIIIII—
-<-<-<-<-<-<ZZZZ-<-<-<Z-<-<-<-<Z-<-<ZZ-<ZZ-<Z-<Z
-<-<-<-<-<-<ZZZZ-<-<-<Z-<-<-<-<Z-<-<ZZ-<ZZZZ-<Z
-<-<-<-<-<-<ZZZZ-<Z-<-<-<-<-<Z-<-<-<ZZZZZZZ-<Z
-<-<-<-<-<-<ZZZZ-<-<-<-<-<-<-<-<Z-<-<-<Z-<ZZ-<Z-<Z

 

 

 

_fil

FIT1 = the fit measure as originally formulated

FIT2 = fit assessed with just product process change and managerial discretion - noting Wood
and Albanese (1995) who found that external environments did not drive choices of HR
policies

FIT3 = using just product process change

FIT4 = not included

FIT5 = internal and external environments- no discretion

295

APPENDIX K

SORTED FIT MEASURES

Success

0.571429
0.657143
0.666667
0.666667
0.693878
0.695238
0.714286
0.742857
0.747253
0.757143
0.758242
0.77551
0.780952
0.785714
0.790476
0.791667
0.795918
0.8
0.809524
0.831169
0.834586
0.835714
0.838095
0.85
0.857143
0.857143
0.857143
0.880952
0.885714
0.892857

APPENDIX K

SORTED FIT MEASURES

FIT2
n

296

FIT3
n

 

APPENDIX L

DESCRIPTIVE STATISTICS BY TECHNOLOGY

APPENDIX L

DESCRIPTIVE STATISTICS BY TECHNOLOGY

Overall Sample

FMS Sub-Sample

 

CNC Sub-Sample

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    

 

 

 

     

 

Variable Min Max Mean Min Max Mean Min Max Mean
Batch Size 1 220000 19,023 1 220000 25,41 1 2.8 10000 1455
Churn 0 200 56.15 0 200 69.75 0 51 18.75
Families 1 5000 1 108 1 5000 1281 1 5000 633
[ MPS Deviation 0 35 11.9 0 35 12.18 1.5 20 10.93
Product Process || -7.95 6.53 .467 -7.95 6.53 .759 -1.01 .954 -.335
L QMI 1 7 2.7 1 7 2.7 1.5 4 2.6
|_ Q15 1 6 2.4 1 5 2.2 2 6 2.9
Q1 61k 6 2.6 1 6 2.6 1 6 2.6
Q17 1 7 2.7 1 7 2.6 1 6 3.1
r Q18 r 1 4 1.7 1 4 1.7 1 4 1.8
Environmental 6 20 12.1 6 20 11.78 8.5 19 13
Uncertainty
Total 1 342 120.01 1 342 127.8 1 265 98.5
Preparation
Time
Segments 1 20 4.717 1 20 5.3 1 6 3.3
+Direct o 50 7.828 0 50 10.8 o 12 4.9
Competitors
Success .571 .893 .781 .571 .893 .788 .694 .857 .76
Self Assessment F 4 7 5.867 4 7 5.864 5 7 5.9
Employment 10 2500 621.5 10 2100 - 437 450 2500 1 128

   

 

 

 

 

 

 

 

 

 

 

 

297

BIBLIOGRAPHY

BIBLIOGRAPHY

Abrahamson, Eric, “Managerial Fads and Fashions: The Diffusion and
Rejection of Innovations”, Academy of Management Review, Vol. 16, No. 3, pp.
586-612, (1991).

Abrahamson, Eric, and Rosenkopf, Lori, “Institutional and Competitive
Bandwagons: Using Mathematical Modeling as a Tool to Explore Innovation
Diffusion”, Academy of Management Review, Vol. 18, No. 3, pp. 487-517,
(1993).

Adler, Paul 8., “Managing Flexible Automation”, California Management
Review, Vol. 30, No. 3, pp. 34-56, (Spring 1988).

Adler, Paul 8., “Workers and Flexible Manufacturing Systems: Three
Installations Compared”, Joumal of Organizational Behavior, Vol. 12, pp. 447-
460, (1991).

Afifi, AA, and Clark, V., Computer-Aided Multivariate Analysis, New York,
Van Nostrand Reinhold Company, (1990).

Arthur, Jeffrey B., “The Link Between Business Strategy and Industrial
Relations Systems in American Steel Minimills”, Industrial and Labor Relations
Review, Vol. 45, No. 3, pp. 488-506, (April 1992).

Arthur, Jeffrey B., “Effects of Human Resource Systems on Manufacturing
Performance and Turnover", Academy of Management Joumal,_Vol. 37, No. 3,
pp. 670-687, (June 1994).

Astley, W. Graham, and Van de Ven, Andrew H., “Central Perspectives
and Debates in Organization Theory”, Administrative Science Quarterly, Vol. 28,
pp. 245-273, (1983).

Attaran, Mohsen, “Flexible Manufacturing Systems: Implementing an

Automated Factory”, Information Systems Management, pp. 44-47, (Spring
1992)

298

299

Attewell, Paul, “The Deskilling Controversy”, Work and Occupations, Vol.
14, No. 3, pp. 3232-346, (August 1987).

Badiru, Adedeji B., “Analysis of Data Requirements for FMS
Implementation is Crucial to Success”, Industrial Engineering, Vol. 22, No. 10,
pp. 29-32, (October 1990).

Beatty, Carol, “Critical Implementation Decisions for Advanced
Manufacturing Technologies”, lntemational Journal of Technology Management,
Vol. 8, No. 1/2, pp. 189-196, (1993).

Bessant, John, “The Lessons of Failure: Learning to Manage New
Manufacturing Technology”, lntemational Joumal of Technology Management,
Vol. 8, Nos. 2/3/4. pp. 197-215, (1993).

Birecree, Adrienne M., “Corporate Development, Structural Change, and
Strategic Choice: Bargaining at International Paper Company in the 19808”,
Industrial Relations, Vol. 32, No. 3, pp. 343-366, (1993).

Blackler, Frank, and Brown, Collin, “Management, Organizations, and the
New Technologies”, In Information Technology and People: Designing for the
Future, Frank Blacker and David J. Obome eds., MIT Press Cambridge Mass. ,
(1987)

Bluedorn, Allen 0., “Pilgrim’s Progress: Trends and Convergence in
Research on Organizational size and Environments”, Journal of Management,
Vol. 19, No. 2. pp. 163-191, (1993).

Bluedorn, Allen C., Johnson, Richard A., Cartwright, Debra K., Barringer,
Bruce R., “The Interface and Convergence of the Strategic Management and
Organizational Environmental Domains”, Journal of Management, Vol. 20, No. 2,
pp. 201-262, (1994).

Blumberg, M., and Gerwin, D., “Coping with Advanced Manufacturing
Technology”, Journal of Occupational Behavior, Vol. 5, pp. 113-130, (1984).

Boer, Harry, Hill, Malcolm, and Krabbendam, Koos, “FMS Implementation
Management: Promise and Performance", lntemational Joumal of Operations
and Production Management, Vol. 10, No. 1, pp. 5-20, (1990).

Boer, Harry and Krabbendam “The Effective Implementation and
Operation of Flexible Manufacturing Systems”, lntemational Studies of
Management and Organization, Vol. 22, No. 4, pp. 33-48 (1992)

300

Bourgeois, L.J. Ill, “Strategy and Environment: A conceptual Integration”,
Academy of Management Review, Vol. 5, No. 1, pp. 25-39 (1980)

Boyer, K.K., Leong,G., Ward, PT, and Krajewski, L., “Unlocking the
Potential of Advanced Manufacturing Technologies”, forthcoming, Joumal of
Operations Management, (1997).

Brass, Daniel J., “Technology and the Structuring of Jobs: Employee
Satisfaction, Performance, and Influence”, Organizational Behavior and Human
Decision Processes, Vol. 35, pp. 216-240, (1985).

Braverman, Harry, Labor and Monopoly Capital : The Degradation of Work
in the Twentieth Century, Monthly Review Press, New York, (1974).

Brewer, John, and Hunter, Albert, “Multimethod Research: A Synthesis of
Styles”, Newberry Court CA., Sage Publications, (1989).

Buchanan, DA, and Boddy, D., Organizations in the ComputerAge:
Technological Imperatives and Strategic Choice, Aldershot, Gower , (1983).

Burnes, T. and Stalker, G.M., The Management of Innovation, London,
Tavistock Publications, (1961).

Burns, Bernard, New Technology in Context: The Selection, Introduction
and Use of Computer Numerically Controlled Machine Tools, Avebury, Brookfield
Vermont, (1989). '

Campion, Michael A., and Berger, Chris J., “Conceptual Integration and
Empirical Test of Job design and Compensation Relationships”, Personnel
Psychology, Vol. 43, pp. 525-553, (1990).

Cappelli, Peter, “Are Skill Requirements Rising? Evidence from
Production and Clerical Jobs”, Industrial and Labor Relations, Vol. 46, No. 3, pp.
515-530, (April 1993).

Chaykowski, Richard P. and Slotsve, George A., “The Impact of Plant
Modernization on Organizational Work Practices”, Industrial Relations, Vol. 31,
No. 2, pp. 309-329, (Spring 1992).

Child, John, “Organization Structure, Environment, and Performance: The
Role of Strategic Choice”, Sociology,_Vo|. 6, No. 1, pp. 1-22, (1972).

301

Child, John, “Managerial Strategies, New Technology, and the Labor
Process” in Job Redesign: Critical Perspectives on the Labor Process, David
Knights, Hugh WIlImott, and David Collinson, Eds. Gower, Brookfield Vermont
(1985)

Cohen, Jacob and Cohen, Patricia, Applied Multiple Regression/
Correlation Analysis for the Behavioral Sciences: Second Edition, Lawrence
Erlbaum Associates, Hillsdale New Jersey, (1983)

Cook, TD, and Campbell, D.T., Quasi-Experimentation: Design and
Analysis Issues for Field Settings, Boston, Houghton Mifflin Company, (1979).

Cooper, William W., Sinha, Kingshuk K., and Sullivan, Robert 8.,
“Measuring Complexity in High-Technology Manufacturing: Indexes for
Evaluation”, Interfaces, Vol. 22, No. 4, pp. 38-48 (July-August 1992).

Coulthurst, Nigel, “Organizing and Accounting for the New Factory”,
Management Accounting, Vol. 67, No. 5, pp. 38-41, (May 1989).

Covin Jeffrey G., and Slevin, Dennis P., “Strategic Management of Small
firms in Hostile and Benign Environments”, Strategic Management Journal, Vol.
10, pp. 75-87, (1989).

Currie, Wendy, “The Strategic Management of AMT in Japan, the USA,
the UK, and West Germany (Part 1)”, Management Accounting, Vol. 70, No. 10,
pp. 32-36, (November 1992). '

Davies, A., Industrial Relations and New Techno/09% Croom-Hlem,
London, (1986).

Dean, James W. Jr., and Snell, Scott A., “Integrated Manufacturing and
Job Design: Moderating Effects of Organizational Inertia”, Academy of
Management Joumal, Vol. 34, No. 4, pp. 776-804, (1991).

De Meyer, Arnoud, Nakane, Jinichiro, Miller, Jeffrey G. and Ferdows,
Kasra, “Flexibility: The Next Competitive Battle The Manufacturing Futures
Survey”, Strategic Management Journal, Vol. 10 . PP. 135-144 (1989).

Dess, Gregory 6., and Beard, Donald W., “Dimensions of Organizational
Task Environments”, Administrative Science Quarteriy, Vol. 29, pp. 52-73,
(1984)

Diaz, Ismael Jr., “Back To Basics: Just What is Involved in Implementing a
Flexible Manufacturing System?”, Industrial Engineering, Vol. 23, No. 4, pp. 43-
44, (April 1991).

302

DiMaggio, Paul J., and Powell, Walter W., “Institutional lsomorphism”,
American Sociological Review, Vol. 48, No. 2, pp. 147-160, (1983).

Duncan, Robert B., “Characteristics of Organizational Environments and
Perceived Environmental Uncertainty”, Administrative Science Quarteriy, Vol. 17,
No. 3, pp. 313-327, (1972)

Dunlop, John T., Industrial Relations Systems, Holt, Rinehart, and
VIIInston, New York, (1958).

Eisenhardt, Kathleen M., “Building Theories from Case Studies”, Academy
of Management Review, Vol. 14, No. 4, pp. 532-550, (1989).

Elango, B., and Meinhart, Wayne A., “Selecting a Flexible Manufacturing
System - A Strategic Approach”, Long Range Planning, Vol. 27, No. 3, pp. 118-
126, (1994).

Fink, A., The Survey Handbook: Volume 1, Sage Publications, Thousand
Oaks CA., (1995).

Flynn, Barbara 3., and Smith, Roger, “Meeting the Challenge of the 21st
Century: Infrastructure for the Use of Advanced Manufacturing Technologies by
Small Manufacturers”, Presented at the national meeting of The Academy of
Management, (1995).

Form, William, “On the Degradation of Skills”, Annual Review of
Sociology, Vol. 13, pp. 29-47, (1987).

Form, William, Kaufman, Robert L., Parcelt, Toby L. and Wallace,
Michael, “The Impact of Technology on Work Organization and Work Outcomes”,
in George Farkas and Paula England Eds., Industries, Firms and Jobs:
Expanded Edition, Aldine DE Gruyter, New York, (1994).

Fowler, Floys J. Jr., and Mangione, Thomas W., Standardized Survey
Interviewing: Minimizing Interviewer-Related Emor, Sage Publications, New York,
1 990.

Fry, Timothy D., “Manufacturing Performance and Cost Accounting”,
Production and Inventory Management Journal, pp. 30-35, (third Quarter 1992).

Fry, Timothy D., and Smith, Allen E., “FMS Implementation Procedure: A
Case Study”, IEE Transactions, Vol. 21, No. 3, pp. 288-293, (September 1989).

303

Genrvin, Donald, “Control and the Evaluation in the Innovation Process:
The Case of Flexible Manufacturing Systems”, IEEE Transactions on
Engineering Management, Vol. em-28, No. 3, pp. 62-70, (1980).

Genrvin, Donald, “Do’s and Don’ts of Computerized Manufacturing”,
Harvard Business Review, pp. 107-116 (March-April 1982).

Gerwin, Donald, “Manufacturing Flexibility: A Strategic Perspective”,
Management Science, Vol. 39, No. 4, pp. 395-410, (1993).

Glasser, B.G., and Strauss, AL, The Discovery of Grounded Theory:
Strategies for Qualitative Research, Aldine, Chicago, IL. (1967).

Goldhar, Joel D., and Lei, David, “Organizing and Managing the CIM/FMS
Firm for Maximum Competitive Advantage”, lntemational Journal of Technology
Management, Vol. 9, Nos. 5/6/7, pp. 709-732, (1994).

Groth, John C., “Critical Factors in Exploiting Technologies”, Management
Decision, Vol. 33, No. 3, pp. 34-47, (1993).

Guest, David E., “Human Resource Management and Industrial
Relations”, Journal of Management Studies, Vol. 24, No. 5, pp. 503-521,
(September 1987).

Guest, David E., “Human Resource Management: Its Implications for
Industrial Relations and Trade Unions”, New Perspectives on Human Resource
Management, John Storey Ed., Routledge, New York, (1989).

Gupta, Atul, Chen, lnjazz J., and Rom, Walter 0., “Understanding the
Human Aspects of Flexible Manufacturing Systems Through Management
Development”, Journal of Management Development, Vol. 12, No. 1, pp. 33-42,
(1993)

Gupta, Yash P., “Human Aspects of Flexible Manufacturing Systems”,
Production and Inventory Management Joumal, pp. 30-35, (Second Quarter
1989)

Gupta, Yash P., and Yakimchuk, “Impact of Advanced Manufacturing
Technology on Industrial Relations: A comparative Study”, lntemational Journal
of Production Management, Vol. 9, No. 3, pp. 5-14, (1988).

Hambrick, Donald C., and Lei, David, “Toward a Prioritization of
Contingency Variables for Business Strategy”, Academy of Management
Journal, Vol. 28, No. 4, pp. 763-788, (1985).

Hamper, Ben, Rivethead, Warner Books, New York, (1991).

304

Handfield, Robert, “A Study of Factors Affecting Delivery Speeds in Make
to Order Manufacturing”, Unpublished Doctoral Dissertation.

Handfield, Robert and Pagell, Mark, “An Analysis of the Diffusion of
Flexible Manufacturing Systems” lntemational Joumal of Production Economics,
Vol. 39, (1995).

Hartwig, Fredrick, and Dearing, Brian E., Exploratory Data Analysis, Sage
Publications, Beverly Hills CA., 1979.

Hazlehurst, R.J., Bradbury, R.J., and Corlett, E.N., “A Comparison of the
Skills of Machinists on Numerically-Controlled and Congenital Machines”,
Occupational Psychology, Vol. 42, No. 3&4, pp. 169-182, (1969).

Hill, Terry, Manufacturing Strategy. Macmillan, (1994).

Horvitz, Wayne L., “The New Labor-Management Relations: A Cautionary
Tale”, Negotiation Journal, Vol. 10, No. 3, pp. 281-292, (July 1994).

Hrebiniak, Lawrence G., and Joyce, William F., “Organizational
Adaptation: Strategic Choice and Environmental Deterrninism”, Administrative
Science Quarterly, Vol. 30, pp. 336-349, (1985).

Huberman A. Michael, and Miles, Matthew B., “Data Management and
Analysis Methods”, in Handbook of Qualitative Research, Norman K. Denzin,
and Yvonne S. Loncoln eds., Sage Publications, Thousand Oaks CA., (1994).

Hunter, John E., Schmidt, Frank L., and Juduesch, Michael K., “Individual
Differences in Output in Output Variability as a Function of Job Complexity”,
Joumal of Applied Psychology, Vol. 75, No. 1 , pp. 28-42, (1990).

Huq, Faizul, “Labor Issues in the Implementation of Group Technology
Cellular Manufacturing”, Production and Inventory Management Journal, pp. 15-
19, (Forth Quarter 1992).

Jackson, PR, and Wall, T.D., “ How Does Operator Control Enhance
Performance of Advanced Manufacturing Technology?”, Ergonomics, Vol. 34,
No. 10, pp. 1301-1311 (1991)

Jackson, Susan E., Schuler, Randall S., and Rivero, J. Carlos,
“Organizational Characteristics as Predictors of Personnel Practices”, Personnel
Psychology, Vol. 42, No. 4, pp. 727-786, (Winter 1989).

Jaikumar, Ramchandran, “Postindustrial Manufacturing”, Harvard
Business Review, pp. 69-76, (November-December 1986).

305

Jarvenpaa, Sirkka L., and Ives, Blake, “Organizing for Global Competition:
The Fit of Information Technology”, Decision Sciences, Vol. 24, No. 3, pp. 547p-
579, (1993).

Jelinek, Mariann, “Technology, Organizations, and Contingency”,
Academy of Management Review, Vol. 2, No. 1, pp. 17-26, (January 1977).

Kaku, Bharat, K., “Fitting Flexible Manufacturing Systems to the Task: An
Analysis of Currant Practice”, Industrial Engineering, Vol. 26, No. 11, pp. 38-40,
(1994).

Karuppan, Corinne M., and Schniederjans, Marc J., “Sources of Worker
Stress in An Automated Plant”, Production and Operations Management, Vol. 4,
No. 2, pp. 108-126, (Spring 1995).

Keefe, Jeffrey H., “Numerically Controlled Machine Tools and Worker
Skills”, Industrial and Labor Relations Review, Vol. 44, No. 3, pp. 503-519, (April
1991)

Kelly, Maryellen R., “Programmable Automation and the Skill Question: A
Reinterpretation of the Cross-National Evidence”, Human Systems Management,
Vol. 6, pp. 223-241, (1986).

Kelly, Maryellen R., “New Process Technology, Job Design, and Work
Organization: A Contingency Model”, American Sociological Review, Vol. 55, pp.
191-208, (1990) ‘

Kelly, Maryellen R., “Productivity and Information Technology: The elusive
Connection”, Management Science, Vol. 40, No. 11, pp. 1406-1425, (1994).

Kerlinger, Fred M, Foundations of Behavioral Research, third edition,
Holt, Rinehart, and V\finston, Inc., Chicago, IL, 1986.

Kim, Yearnmin, and Lee, Jinjoo, “Manufacturing Strategy and Production
Systems: An Integrative Framework”, Journal of Operations Management, Vol.
11, No. 1, pp. 3-15, (1993).

Kimberly, John R., and Rottman, David B., “Environment, Organization
and Effectiveness: A Biographical Approach”, Joumal of Management Studies,
Vol. 25, No. 6, pp. 595-622, (1987).

Knorr, Robert O., and Thuede, Edward F. Jr., “Making New Technologies
Work”, The Journal of Business Strategy, pp. 46-49, (January/February 1991).

 

306

Kochan, Thomas A., McKersie, Robert B, and Cappelli, Peter, “Strategic
Choice and Industrial Relations Theory”, Industrial Relations, Vol. 23, No. 1, pp.
16-39, (Winter 1984).

Kotha, Suresh, and Orne, Daniel, “Generic Manufacturing Strategies: A
Conceptual Synthesis”, Strategic Management Joumal, Vol. 10, pp. 211-231
(1989) '

Lawrence, Paul R., and Lorsch, Jay W., Organization and the
Environment: Managing Differentiation and Integration, Harvard Business
School Press, Boston, (1986).

Lawrence, Paul R., and Lorsch, Jay W., “Differentiation and Integration in
Complex Organizations”, Administrative Science Quarteriy, Vol. 12, No. 1, pp. 1-
47, (June 1967).

 

Lund, Robert T., and Hansen, John A., Connected Machines,
Disconnected Jobs: Technology and Work in the Next Decade, MIT Press,
(1983).

Majchrzak, Ann, The Human Side of Factory Automation, Jossey-Bass,
San Franscisco, (1988).

Mansfield, Edwin, “The Diffusion of Flexible Manufacturing Systems in
Japan, Europe, and The United States”, Management Science, Vol. 39, No. 2.
pp. 149-159, (February 1993a). '

Mansfield, Edwin, “New Evidence on the Economic Effects and Diffusion
of FMS”, IEEE Transactions on Engineering Management, Vol. 40, No. 1, pp. 76-
78, (February 1993b).

Martin, Robin, “Attitudes Toward Advanced Manufacturing Technology
(AMT): The Role of AMT Experience, Skill Level and Job Involvement”, Social
Behavior, Vol. 3, pp. 297-305, (1988).

Maskell, Brian, “Performance Measurement for World Class
Manufacturing - 3”, Management Accounting, Vol. 67, No. 7, pp. 48-50, (1989).

Melnyk, SA, and Narasimhan, R., Computer Integrated Manufacturing,
lrvvin, Homewood, IL, (1992)

Meredith, Jack R., “Strategic Planning for Factory Automation by the
Championing Process”, IEEE Transactions on Engineering Management, Vol.
em-33, No. 4, pp. 229-231, (November 1986).

307

Meredith, Jack R., “Implementing New Manufacturing Technologies:
Managerial Lessons Over the FMS Life Cycle”, Interfaces, Vol. 17, No. 6, pp. 51-
62, (November-December 1987).

Mieskonen, Jari, “The Success of FM Investments: Case Studies from
Small Industrial Economies”, lntemational Journal of Technology Management,
Vol. 6, Nos. 3/4, pp. 277-291, (1991).

Miles, Matthew B., and Huberman A. Michael, Qualitative Data Analysis:
An expanded Sourcebook, Sage, Thousand Oaks California, 1994.

Milkman, Ruth, and Pullman, Cydney, “Technological Change in An Auto
Assembly Plant: The Impact on Workers’ Tasks and Skills”, Work and
Occupations, Vol. 18, No. 2, pp. 123-147, (May 1991).

Miller, J.G., and Roth, A.V., “A Taxonomy of Manufacturing Strategies”,
Management Science, Vol. 40, pp. 285-304, (1994).

Miller, VVrlliam A., and Gilbert, Richard, “Computer Control in Tomato
Packing: A Low-Technology Case Study”, Production and Inventory
Management Journal, pp. 64-67, (Forth Quarter 1992).

Mueller, Frank, “Teams Between Hierarchy and Commitment: Change
Strategies and the Internal Environment”, Journal of Management Studies, Vol.
31, No. 3, pp. 383-403, (1994).

Nemetz, Patricia L., and Fry, Louis W., “Flexible Manufacturing
Organizations: Implications for Strategy Formulation and Organization Design”,
Academy of Management Review, Vol. 13, No. 4, pp. 627-638, (1988).

Nobel, David F., Forces of Production: A Social History of Industrial
Automation, Alfred A. Knopf, New York, (1984).

Nobel, M., “Manufacturing Strategy: Testing the Cumulative Model in a
Multiple Country Context”, Decision Sciences, Vol. 26, pp. 693-721, (1995).

Oliver, Christine, “The Collective Strategy Framework: An Application to
Competing Predictions of lsomorphism”, Administrative Science Quarterly, Vol.
33. pp. 543-561, (1988).

Osterman, Paul, “Choice of Employment Systems in Internal Labor
Markets”, Industrial Relations, Vol. 26, No. 1, pp. 46-67, (Winter 1987).

308

Osterman, Paul, “How Common is Workplace Transformation and Who
Adopts lt?”, Industrial and Labor Relations Review, Vol. 47, No. 2, pp. 173-188,
(January 1994).

Pavitt, K., “What We Know About the Strategic Management of
Technology”, California Management Review, Vol. 32, No. 3, pp. 17-26, (1990).

Pennings, J.M., “Structural Contingency Theory: A Multivariate Test”,
Organization Studies, Vol. 8, No. 2, pp. 223—240, (1987).

Poole, Marshall Scott, and Van De Ven, Andrew H., “Using Paradox to
Build Management and Organizational Theories”, Academy of Management
Review, Vol. 14, No. 4, pp. 562-578, (1989).

Preece, David A., Managing the Adoption of New Technology, Routledge,
New York, (1989).

Prickett, Paul, “Cell-Based Manufacturing Systems: Design and
Implementation”, lntemational Journal of Operations and Production
Management, Vol. 14, No. 2, pp. 4-17, (1994).

Porter, Michael, Competitive Strategy, The Free Press, New York, (1980).

Reshef, Yonatan, “Employees, Unions, and Technological Changes: A
Research Agenda”, Joumal of Labor Research, Vol. 14, No. 2, pp. 1 11-129,
(Spring 1993). ‘

Robinson, S.L., Hamessing Technology: The Management of Technology
for the Nontechnologist, Van Nostrand Reinhold, New York, (1991).

Romanelli, Elaine, and Tushman, Michael T., “Inertia, Environments, and
Strategic Choice: A Quais-Experimental Design for Comparative Longitudinal
Research”, Management Science, Vol. 32, No. 5, pp. 608-621, (1986).

Roth, Philip, L., “Missing Data: A Conceptual Review for Applied
Psychologists”, Personnel Psychology, Vol. 47, pp. 537 - 560, (1994).

Saraph, Jayant V. and Sebastian, Richard J., “Human Resource
Strategies for Effective Introduction of Advanced Manufacturing
Technologies(AMT), Production and Inventory Management Journal, pp. 64-70,
(first quarter 1992).

Schroeder, Dean M., Congden, Steven W., and Gopinath, C., “Linking
Competitive Strategy and Manufacturing Process Technology”, Joumal of
Management Studies, Vol. 32, No. 2, pp. 163-189, (March 1995).

309

Schluer, Randall 8., “Strategic Human Resource Management and
Industrial Relations”, Human Relations, Vol. 42, No. 2, pp. 157-184, (1989).

Schultz-VWd, Rainer, and Kohler, Christoph, “Introducing New
Manufacturing Technology: Manpower Problems and Policies”, Human Systems
Management, Vol. 5, pp. 231-243, (1985).

Scott, W. R., “The Adolescence of Institutional Theory”, Administrative
Science Quarteriy, Vol. 32, pp. 493-511, (1987).

Shaiken, Harley, Herzenberg, Stephen, and Kuhn, Sara, “The Work
Process Under More Flexible Production”, Industrial Relations, Vol. 25, No. 2,
pp. 167-182, (Spring 1986).

Shani, A.B., Grant, Robert M., Krishnan, R., and Thompson, Eric,
“Advanced Manufacturing systems and Organizational Choice: Sociotechnical
System Approach”, Califomia Management Review, (Summer 1992).

Skinner, W., “Manufacturing: Missing Link in Corporate Strategy”,
Harvard Business Review, Vol. 47, pp. 136-145, (1969).

Skinner, W., “The Productivity Paradox”, Han/ard Business Review, pp.
55-59, (July/August 1986).

Smith, Anthony E., “The Nature of Technological Change: Its Implications
for Work and Labor Regulation”, Relations Industrielles, Vol. 47, No. 4, pp. 752-
775, (1992).

Smith, Stuart, Tranfield, David, Bessant, John, Levy, Paul, and Ley, Clive,
“Factory 2000: Organization Design for the Factory of the Future”, lntemational
Studies of Management and Organizations, Vol. 22, No. 4, pp. 61-68, (1992).

Smith, Vicki, “lnstitutionalizing Flexibility in a Service Firm”, Work and
Occupations, Vol. 21, No. 3, pp. 284-307, (1994).

Snell, Scott A., and Dean, James W. Jr., “Integrated Manufacturing and
Human Resource Management: A Human Capitol Perspective”, Academy of
Management Journal, Vol. 35, No. 3, pp. 467-504, (August 1992) .

Snell, Scott A., and Dean, James W. Jr., “Strategic Compensation for
Integrated Manufacturing: The Moderating Effects of Jobs and Organizational
Inertia”, Academy of Management Joumal, Vol. 37, No. 5, pp. 1 109-1 140,
(October 1 994).

310

Son, Young K., “A Performance Measurement Method Which Remedies
the ‘Productivity Paradox’”, Production and Inventory Management Journal, pp.
38-43, (second quarter 1990).

Spano, Michael R. Sr., O’Grady, Peter J., and Young, Robert E., “The
Design of Flexible Manufacturing Systems”, Computers In Industry, Vol. 21, pp.
185-198, (1993). -

Spenner, Kenneth L, “Occupation Characteristics and Classification
Systems of the DOT in Research”, Social Methods Research, Vol. 9, pp. 239-
264, (1980).

Spenner, Kenneth L, “Deciphering Prometheus: Temporal Change in the
Skill Level of Work”, American Sociological Review, Vol. 48, pp. 824-837, (1983).

Spenner, Kenneth L, “Technological Change, Skill Requirements, and
Education: The Case For Uncertainty”, in The Impact of Technological Change
on Employment and Economic Growth, Richard M. Cyert and David C, Mowery
Eds., Ballinger Publishing, Cambridge MA., (1988).

Stake, Robert E., “Case Studies”, in Handbook of Qualitative Research,
Norman K. Denzin, and Yvonna S. Loncoln eds., Sage Publications, Thousand
Oaks CA., (1994).

Stecke, K.E., “Design, Planning, Scheduling, and Control Problems of
Flexible Manufacturing Systems”, Annals of Operations Research, Vol. 3, pp. 3-
12, (1 985).

Strauss, Anselm, and Corbin, Juliet, “Grounded Theory Methodology: An
Overview”, in Handbook of Qualitative Research, Norman K. Denzin, and
Yvonna S. Loncoln eds., Sage Publications, Thousand Oaks, CA (1994).

Stone, Eugene, F., and Hollenbeck, John R., “Clarifying Some
Controversial Issues Surrounding Statistical Procedures for Detecting Moderator
Variables: Empirical Evidence and Related Evidence”, Joumal of Applied
Psychology, Vol. 74, No. 1, pp. 3-10, (1989).

Swamidass, Paul M., and Newell, \MIIiam T., “Manufacturing Strategy,
Environmental Uncertainty and Performance: A Path Analytic Model”,
Management Science, Vol. 33, No. 4, pp. 509-524, (1987).

Tawegoum, R., Castelain, E., and Genrina, J.C., “Real-Time Piloting of
Flexible Manufacturing Systems”, European Journal of Operational Research,
Vol. 78. pp. 252-261, (1994).

 

 

311

Tayles, Mike and Drury, Colin, “New Manufacturing Technologies and
Management Accounting Systems: Some Evidence of the Perceptions of UK
Management Accounting Practitioners”, lntemational Journal of Production
Economics, Vol. 36, pp. 1-17, (1994).

Thomas, Robert J., “Technological Choice and Union Management
Cooperation”, Industrial Relations, Vol. 30, No.2, pp. 167-192, (Spring 1991).

Tichey, Noel M., Fombrun, Charles J., and Devanna, Mary Anne,
“Strategic Human Resource Management”, Sloan Management Review, Vol. 23,
No. 2, pp. 47-62, (VWnter 1982).

Tornatzky, Louis G., and Klein, Katherine J., “Innovation Characteristics
and Innovation Implementation: A Meta-Analysis of Findings”, IEEE
Transactions on Management, Vol. EM-29, No. 1, pp. 28-45, (February 1982).

Tranfiled, David, Smith, Stuart, Ley, Clive, Bessant, John, and Levy, Paul,
“Changing Organizational Design and Practices for Computer-Integrated
Technologies”, lntemational Joumal of Technology Management, Vol. 6, Nos.
3/4, pp. 211-222, (1991).

Van De Ven, Andrew H., Delbecq, Andre L., and Koenig, Richard Jr.,
“Determinants of Coordination Modes Within Organizations”, American
Sociological Review, Vol. 41, pp. 322-338, (April 1976).

Venkatesan, Ravi, “Cummins Engine Flexes its Factory”, Harvard
Business Review, pp. 120-127, (March-April 1990) .

Venkatraman, N., and Prescott, John B, “Environmental Strategy
Coallignment: An Empirical Test of Its Performance Implications”, Strategic
Management Journal, Vol. 11, pp. 1-23, (1990).

Vickery, Shawnee, K., Droge, Cornelia, and Markland, Robert, E.,
“Production Competence and Business Strategy: Do they effect Business
Performance?”, Decision Sciences, Vol. 24, No. 2, pp. 435-455, (1993).

Vineyard, Michael L., “FMS Performance: A Long-Term Study”,
Production and Inventory Management Journal, pp. 36-39, (Fourth Quarter
1993).

Vonderembse, Mark A., and Wobser, Gregory 8., “Steps for Implementing
A Flexible Manufacturing System”, Industrial Engineering, Vol. 19, No. 4., pp. 38-
48, (1987).

312

Wall, Toby D., Corbett, J. Martin, Clegg, Chris W., Jackson, Paul R., and
Martin, Robin, “Advanced Manufacturing Technology and Work Design: Towards
3 Theoretical Framework”, Joumal of Organizational Behavior, Vol. 11, No. 3, pp.
201-219, (1990).

Wall, Toby D. and Davids, Keith, “Shopfloor Work Organization and
Advanced Manufacturing Technology”, in lntemational Review of Industrial and
Organizational Psychology 1992, Gary L. Cooper and Ivan T. Robertson Eds. ,
John erey and Sons, New York , (1992).

Walsh, John, P., “The Social Context of Technological Change: The Case
of the Retail food Industry”, The Sociological Quarterly, Vol. 32, No. 3, pp. 447-
468, (1991).

Walton, Richard E., “From Control to Commitment in the Workplace”,
Harvard Business Review, pp. 77-84, (March-April 1985).

Walton, Richard E., and Susman, Gerald L, “People Policies for the New
Machines”, Harvard Business Review, pp. 98-106, (March-April 1987).

Wernerfelt, Birger and Karani, Aneel, “Competitive Strategy Under
Uncertainty”, Strategic Management Journal, Vol. 8, pp. 187-194 (1987)

Whittington, R., “Environmental Structure and Theories of Strategic
Choice”, Journal of Management Studies, Vol. 25, No. 6, pp. 521 p-536, (1988).

VWlkinson, Leland, “SYSTAT: The System for Statistics”, SYSTAT Inc.,
Evanston, IL., (1990).

Wood, Stephen, and Albanese, Maria Teresa, “Can We Speak of a High
Commitment Management on the Shop Floor?”, Joumal of Management Studies,
Vol. 32, No. 2, pp. 215-247, (March 1995).

Yin, Robert K., Case Study Research: Design and Methods, second
edition, Sage Publications, Thousand Oaks, CA (1994).

Zairi, Mohamed, “Measuring Success in AMT Implementation Using
Customer Supplier Integration Criteria”, lntemational Journal of Operations and
Production Management, Vol. 12, No. 10, pp. 34-55, (1992).

Zicklen, Gilbert, “Numerical Control Machining and the Issue of
Deskilling", Work and Occupations, Vol. 14, No. 3, pp. 452-466, (August 1987).

Zuboff, Shoshana, “New Worlds of Computer-Mediated Work”, Harvard
Business Review, pp. 142-152, (September-October 1982).