CHAPTER III

Methodology

The purpose of this study is to examine apparel manufacturers' inventory

performance, which is one of the major aspects of supply chain management (SCM).
First, apparel manufacturers' level of SCM activity was identified. In addition, how
apparel manufacturers' characteristics in terms of product characteristics, production
system, fabric supplier, and retail customers are different according to the level of SCM
activities and how the four characteristics are related to inventory performance (i.e., level
within the type) were identified. This chapter details the procedures for the study in the
following sections: (a) hypotheses, (b) research design, (c) sample, (d) instrument, (e)
data collection, and (f) data analysis.

Hypotheses

This study is designed to examine how the level of SCM activities are related to
selected apparel manufacturers' characteristics and how these characteristics are related to
inventory performance. The following hypotheses are proposed.

H1. Apparel manufacturers show different levels of SCM activities.

Based on the literature analysis, the SCM activities can be composed of six
dimensions: collaborative partnership, utilization of information technology, flexibility of
operations, service and performance measurements, top management commitment, and
knowledge of demand characteristics. Each dimension requires specific activities to
achieve benefits from the successful SCM. According to Subramanian and Nilakanta
(1996), the adoption of innovations is related with a company's environmental,
organizational, and individual factors. Implementation of QR program, and computer
technology usage in purchasing, design, and manufacturing were found to be different
among apparel manufacturers. (Belleau & Didier, 1989; Ko & Kincade, 1998;
Priyadarshi, 1996; Sullivan & Kang, 1999). Therefore, Hypothesis One suggests that the
level of SCM activities be different among apparel manufacturers.

55
H2. Apparel manufacturers' company characteristics (i.e., product characteristics,
production system, fabric supplier, retail customers) are related to the level of
SCM activities.
Company characteristics are varied according to the company's level of
innovation adoption, based on the conceptual framework of this study. SCM is the
innovation that a company may adopt and four apparel manufacturers' characteristics are
the organizational characteristics in the framework. More product variety is a product
characteristic required in the apparel industry (Standard & Poor's, 1998). Product variety
in apparel is measured in terms of fashion level, which interprets to uncertainty of
demand. Fashion level was found to be related to the implementation of QR program in
the apparel industry (Ko & Kincade, 1998). SCM allows apparel manufacturers to better
serve their customers through improved operational activities (i.e., reduced cycle times,
lower inventory levels), while reducing costs; therefore, implementing SCM activities
would be related to production of fashion apparel.

Traditional production system such as bundle system and progressive bundle

system (PBS) may be less efficient to implement SCM activities due to its nature of high
levels of inventory and extended throughput time (Bailey, 1993). On the other hand,
modular system, which reduces the throughput time and WIP inventory, enhances the
operational flexibility and seeks economies of scope (Berg, Appelbaum, Bailey, &
Kalleberg, 1996). Modular system may be related with a high level of SCM activities.

Fabric suppliers and retail customers are adjacent supply chain partners to apparel
manufacturers and are the companies with whom apparel manufacturers have direct
relationships. The supplier performance is assumed to be a key component for efficient
SCM. Recently, suppliers' delivery performance has received more attention from buyers
(Artz, 1999). The size and type of retailers with whom apparel manufacturers do the
business are found to be related to the manufacturers' managerial decisions in the
literature analysis (Kincade & Cassill, 1993; Ko, 1995). And, traditional relationships
such as short-term and low-cost oriented adversarial relationships are barriers to

56
implementing the SCM activities. More collaborative partnership-like relationships, with
suppliers or retailers are hypothesized to be found in the apparel manufacturers who show
high level of SCM activities. Overall, Hypothesis Two suggests that levels of the SCM
activities identified in Hypothesis One be related to the selected apparel manufacturers'
characteristics (i.e., product characteristics, production system, fabric suppliers, retail
customers).

H3. Apparel manufacturers' company characteristics (i.e., product characteristics,

production system, fabric suppliers, retail customers) are related to inventory
performance (i.e., level of inventory within the type).
As a company tries to adjust company characteristics to the changing environment
when adopting innovation, the company's performance should be distinguished from non-
adopter companies' performances. In this study, inventory level within the type will be
used as an indicator of a company's performance. The level of inventory held in an
apparel manufacturing plant is related to the decisions on the volume and timing of raw
material purchasing, production, and delivery (McPherson, 1987).

First of all, product characteristics (i.e., fashion goods, basic goods) have a
relationship with raw material purchasing, production, and delivery (Fisher, Hammond,
Obermeyer, & Raman, 1997). Raw materials for highly seasonal and fashionable goods
may need to be ordered once in case the supplier discontinues production of the raw
material. Raw materials for basic goods may be ordered continuously because the
demand for the basic goods is stable and certain (Tersine, 1988). Suppliers of raw
materials of basic goods can provide the materials on a continuous basis. Planned
production volume will be different according to the product characteristics. According
to U. S. Bureau of the Census's (1996) report on value of manufacturers' inventories by
type, average finished goods inventory accounted for 42.8% and average raw material
inventory for 36.1% of the total average inventory in women's wear (SIC 233, 234).
Average finished goods inventory and raw material inventory of men's wear (SIC 231,
232) accounted for 55.9% and 25.7% of the total average inventory, respectively in 1996.
Considering, women's wear represents higher fashion level than men's wear, fashion

57
goods may be stored as raw materials and produced as close to the selling season as
possible, and basic goods are produced and stored as finished goods until delivered.

The level of WIP is closely related to the production system. For example, the
bundle system has the highest level of WIP, compared to PBS or modular system (Bailey,
1993). Therefore, production system may be one factor that determines the level of
inventory in the apparel manufacturing.

Fabric suppliers' cost offering, delivery performance and characteristics of

supplier-buyer relationship will affect the apparel manufacturers' inventory level. For
example, if buyers are more cost-oriented, they may place an order in a bulk, which
implies high levels of raw material inventory. When suppliers' are not reliable in product
quality, timely delivery, and fill rate, apparel manufacturers may need to build higher
inventory levels to avoid the potential shortage of raw material. In addition, if the
relationship with apparel manufacturers is long-term and collaborative based on trust, and
QR-oriented, the apparel manufacturers' inventory performance will be enhanced.

The size of the retail customer may affect inventory levels for an apparel
manufacturer. If retailers, who are a big account to apparel manufacturers, want to place
more frequent and small-lot orders, apparel manufacturers need to tune their delivery
volume and timing to the will of the retailers. The partnership-like relationships between
retailers and apparel manufacturers are also considered to have positive a impact on the
company's performance in the supply chain (Harrington, 1999). In addition, the type of
retailer is found to influence apparel manufacturers' decision on adopting QR programs
(Kincade & Cassill, 1993; Ko, 1995). Therefore, Hypothesis Three states that the level
within type of inventory held in apparel manufacturers' sites differ based on apparel
manufacturers' characteristics (i.e., product characteristics, production system, fabric
supplier, retail customers).

58
H4. Apparel manufacturer' level of SCM activities is related to inventory
performance (i.e., level within the type)
This study also focuses on the level of inventory within the type in apparel
manufacturing as a measure of inventory performance and an effect of SCM. Although a
high level of SCM activities are assumed to reduce the level of inventory, according to
the effect of the company's organizational characteristics, the inventory performance
among apparel manufacturers may differ. Therefore, how the level of SCM activities
alone is related to the level within type of inventory will clarify the effect of the four
company characteristics on the inventory performance of the apparel industry.

Assumptions

This study was done based on the following assumptions:

1. Apparel manufacturers have some level of involvement with SCM activities.
2. Apparel manufacturers have knowledge of product characteristics, production system,
and have a business relationship with at least one fabric supplier and retailer.
3. Apparel manufacturers have knowledge of their inventory status.

Research Design

This study employed non-experimental quantitative research. Specifically, the

design falls into the single cross-sectional design using the mail survey method, which is
the most frequently used descriptive research design. Previous studies of similar subjects
to this study used the same research design (Kincade, 1988; Ko, 1993; Jones, 1999). This
research design requires one sample of respondents and quantitative data analyses with a
large and representative sample. The descriptive research was used to test specific
hypotheses and make careful descriptions or explore the possible relationships between
variables unlike the causal research that is designed to determine cause and effect
relationships (Malhotra, 1996).

To increase the internal and external validity, the sampling procedure in this study
applied the stratified random sampling technique. Internal validity refers to whether the

59
variation in dependent variables is really caused by the variation in independent variables
(Malhotra, 1996). The extraneous variables that are not included in the study but might
affect the dependent variables should be controlled to increase the internal validity.
External validity refers to the ability to generalize the findings of the study.
Randomization and stratification in sampling is one method of controlling extraneous
variables and of increasing generalizability beyond the sample.

Sample

This study focuses on U.S. apparel manufacturers. The population consists of the
apparel manufacturers whose products can be classified as fashion goods or basic goods.
Apparel manufacturer classified under SIC 235, 237, 238, and 239 were excluded
because the items, which include hats, caps, fur goods, and miscellaneous apparel
accessories, are considered not to represent both seasonality and fashionability as well as
men's and women's. The sample mainly consists of SIC 231, 232, 233, and 234. The SIC
codes 231, 232, 233, and 234 are assigned mainly to manufacturers of men's and women's
garments which account for approximately 48.3% in the number of establishments and
60.7% of value of manufacturers' inventories in the apparel manufacturers under SIC
code 23 (US Census Bureau, 1996).

The sample companies were selected from purchased directories: The National
Register of Apparel Manufacturers, Women & Children's Wear (Marche Publishing,
1998) and The National Register of Apparel Manufacturers, Men & Boys (Marche
Publishing, 1997). The source of the population was selected because of its large
coverage in terms of the number of apparel manufacturers listed in the directories, which
represents approximately 6,500 out of 11,731 U.S. apparel manufacturing companies.
The main features of the directories include apparel manufacturers' address, phone/fax
numbers, sales volume, employment, and key personnel, listed by state and city, activity,
and product category. This information is important in sampling the population and
contacting the sample. However, the use of a general directory could result in sampling
bias. Listed companies' demographic characteristics may not be representative of the

60
overall U. S. population of apparel manufacturers in terms of company size and product
category. The characteristics of the sample were compared to those of U.S. population to
reveal the bias in results due to sample-specific characteristics.

From the data compiled from County Business Patterns (1996), the total number
of establishments in the target population of apparel manufacturers is estimated at 11,731.
According to Krejcie and Morgan (1970), the suggested sample size for a population of
10,000 is 370 or 3.7 % of the population. In previous studies on apparel manufacturers
where a similar data collection method was used to this study, sample sizes ranged from
118 of Priyadarshi's (1996) study to 246 of Lin, Kincade, and Warfield's (1995) study.
Response rates ranges from 32.5% (ZuHone & Morganozky, 1995) to 48% (Kincade,
Cassill, & Williamson, 1993). These previous studies reported that intensive follow up
contacts with respondents increased the response rates. Krejcie and Morgan's (1970)
suggestion on the sample size shows a big difference from other previous studies, which
focused on one size strata or one state. Sample size for this study was based on Krejcie
and Morgan's (1970) suggestion and the predicted response rate was based on the
previous studies' experience.

From the list, a stratified random sample of 1,195 apparel manufacturers without
location limitation was selected. The list was stratified into four groups by a cross
tabulation of size (i.e.,number of employees) and SIC code (i.e., SIC 231/232, SIC
233/234) (see Figure 3-1). Stratified sampling is a sampling technique that follows two-
steps: partitioning the population into subpopulations and selecting the elements by a
random procedure (Malhotra, 1996).

In this study, operational definitions of small and large apparel manufacturers are
from data compiled from County Business patterns (US Census Bureau, 1996). Small
apparel manufacturer is defined as US apparel manufacturer classified under SIC 231 to
234, with 1 to 19 employees, which accounts for 59.7% of the population. Mediums to
large apparel manufacturers are US apparel manufactures who are classified under SIC
231 to 234, with over 20 employees. Sample was stratified by the company size because

61
apparel manufacturing company is assumed to have differentiating characteristics such as
marketing and manufacturing strategies, the level of usage of advanced technologies, and
the level of QR adoption (Jones, 1999; Priyadarshi, 1996; Sullivan & Kang, 1999).

The proportion of SIC code stratum to the total population is 21% and 79%
respectively (U.S. Census Bureau, 1996). Product category that complies with the
different SIC code is known to affect numerous characteristics or manufacturers. For
example, according to U.S. Bureau of the Census (1996), capital expenditure for plant
and equipment does not parallel the proportion of the establishment between men's wear
and women's wear. Apparel manufacturers under SIC 231/232 expend more capital for
plant and equipment compared to those under SIC 233/234. Besides, product category
may be an indirect variable that is associated with the fashion level. Therefore, to control
the effect of extraneous variables and increase the internal validity, stratified sampling
was used. Figure III-1 shows the description and the proportion to the total population of
each stratum.

Stratum 1 Stratum 2

Small SIC 231 & 232 Medium to large SIC 231 & 232
7.6% 12.9%

Stratum 3 Stratum 4

Small SIC 233 & 234 Medium to large SIC 233 & 234
52.1% 27.4%

Figure III-1. Sample Stratum Description

62
 Instrument

A mail questionnaire was designed to document the relationship between SCM

activities, apparel manufacturers' characteristics, and the level within type of inventory
management in U.S. apparel manufacturers. The questionnaire consists of questions
regarding the following variables: the SCM activities of six dimensions, apparel
manufacturers' characteristics (i.e., product characteristics, production system, fabric
suppliers, retail customers), and inventory performance (i.e., selected performance
indicators for raw material purchasing, production, delivery). The questions used in the
questionnaire were adopted and compiled from the previous study of SCM, QR,
inventory management, and apparel manufacturing (Artz, 1999; Droge & Germain, 1998;
Germain & Droge, 1998; Jones, 1999; Kanakadurga, 1994; Kincade, 1988; Ko, 1993).
The example of the instrument is attached in the Appendix E.

SCM Activities

Respondents were assessed for their level of agreement with 26 statements that
characterize the SCM activities in six dimensions (Question I). Those six dimensions are
partnership, information technology, operation flexibility, performance measurement,
management commitment and leadership, and demand characterization. Each dimension
consists of four to five activity statements with the total of 26 questions identified
through literature analysis. For each activity, six-point items with endpoints of not at all
(0) and very high (5) were used. In compiling the contents into a specific scale,
examples of activities specific to the apparel industry were adopted from Ko and
Kincade's (1998) study of QR technologies. Table III-1 summarizes the contents and
sources of the items under six dimensions.

63
Table III-1
Measurements Contents of SCM Activities
Dimension Measurement Activities Source
Partnership • Consolidation in supplier base Cooper & Ellram(1993),
• Collaboration with suppliers and customers in Higginson & Alam(1997),
forecasting and production planning Johnson(1998),
• Active information and relevant technology sharing Jones & Riley(1985),
• Conversion to long-term relationship with suppliers Kincade & Cassill (1993)
and customers
Information • Use of computer-to-computer communication Alber & Walker(1997),
technology • Use of EDI Copacino(1998),
• Use of POS data communication Higginson & Alam(1997),
• Use of bar-coding Jones & Riley(1985),
Kincade & Vass (1998),
Sabath (1995)

Operation • Ability of volume flexibility Alber & Walker(1997),

flexibility • Ability of product flexibility Cooper & Ellram(1993),
• Ability of delivery flexibility Johnson(1998)
• JIT practice

Performance • Establishment of performance measurements with Johnson(1998),

measurement customers and suppliers Sabath (1995),
Vass (1995)

Management • Top management's awareness of benefits Higginson & Alam(1997),

commitment • Top management's desire to change Vass (1995)
and leadership
Demand • Knowledge of demand pattern Alber & Walker(1997),
characterization • Aligning supply capabilities with demand cycles Copacino(1998),
• Applying the uncertainty in demand to operational Priyadarshi (1996)
decisions

Company Characteristics

This section measured the four company characteristics to examine the

relationship with the SCM activities and with inventory performance. Those four
characteristics are product characteristics, production system, fabric suppliers, and retail
customers.

64
 Product characteristics. This study used the product line characteristics construct
for the product characteristics. One open-ended question is used for respondents to
divide their product line into two groups, fashion goods and basic goods, and describe the
proportion of each type of goods to the total production in percentage (Question II-1).
This question was devised based on the recent trend that apparel manufacturers are
dealing with multiple product lines of fashion goods and basic goods. In Priyadarshi's
study (1996), demand uncertainty was used to measure product characteristics. Fashion
goods and basic goods used in this study were defined to be different in ease of
forecasting, fashion level and seasonality in demand pattern, and frequency of style
change based on Priyadarshi's research instrument (1996). In detail, fashion goods refer
to products of which the demand is hard to forecast due to high fashion level and
seasonality and have quite varied style changes season by season. Basic goods refer to
products of which the demand is relatively easy to forecast due to low level of fashion
level and seasonality and have the basic garment style that repeats frequently.

Production system. One multiple-choice question was presented to respondents to

select the major production system they use (i.e., bundle system, PBS, UPS, modular
system) (Question IV-2). This question was compiled from Kanakadurga's (1994)
research instrument and an additional choice of UPS was added to the original question.

Fabric suppliers. Respondents were asked to evaluate the suppliers' cost- offering
and delivery performance (Question V-4). Delivery performance questions were formed
by adopting four questions from Artz's (1999) study. These questions were asked to
address the quality of products, satisfaction with the delivery performance of the supplier,
on-time delivery, and defect rate of the product on six point scale with endpoints of not at
all (0) and very high (5). Also, respondents were asked to indicate the most appropriate
description of the nature of their relationship with fabric suppliers on a six-point scale
(Question V-5). These questions were based on the definition of partnership and
compiled from previous studies (Cooper & Ellram, 1990; droge & Germain, 1998).

65
 Retail customers. Major retail customers' type was asked with a multiple-choice
question, which was adopted from a previous study (Ko, 1993) (Question IV-3). Retailer
types are department stores, specialty stores, mass merchandisers/discounters, small
independent stores, direct mail, and others. Direct mail was added by the author because
this type of retailer is growing fast (Diamond, 1993). The extent of cost orientation and
ownership by the respondent, and size relative to the respondent's company are asked to
address the retail customers' characteristics (Question IV-4, V-1). These items were
included because the typical classification of retailers' type is not mutually exclusive.
Extent of the attributes mentioned in this questionnaire has been found to be different
according to the type or size of retailers. The same questions that address the nature of
relationship with fabric suppliers were asked again to describe the relationship with retail
customers as well (Question V-2).

Inventory Performance

In this section, questions were developed to obtain the level of inventory within
the type. The level of raw material inventory was measured on four open-ended scales by
asking the following: (1) how many weeks of raw materials inventory were kept on hand
(Question III-3), and (2) what is the typical order lead-time for raw material from order
placement to order receipt (Question III-8). The former question was adopted from
Sterling and Lambert's (1979) study of methodology for assessing logistics operating
systems and the latter question was from Droge and Germain's (1998) study of JIT
impact on inventory level.

Weeks of raw material is a traditionally used inventory measurement throughout

many industries, called weeks of supply (WOS), or equivalently, days of supply (Tersine,
1991). WOS indicates the expected time to exhaust a raw material's current on-hand
inventory (Question III-3). To calculate the weeks of inventory, the following formula is
used: weeks of inventory = on-hand inventory/weekly requirements rate (Pena,
Beaumarigage, & Nelson, 1997). Raw materials for the apparel products include fabrics,
threads, and a variety of components such as zippers, embroidery, and binding. In this

66
study fabric was used as a representative of the raw material because fabric costs account
for about 80% of material costs (Tyler, 1991).

The level of WIP was measured on open-end scales by asking the average
production lot size in pieces (Question III-4). In addition, in Question III-8, respondents
were asked to indicate shop lead-time in days (time span from shop order release to the
order completion). The production lot size and the rate of operation overlapping are
directly connected to the production system (i.e., bundle system, PBS, UPS, modular
system) that is distinguishable by the WIP level (Kanakadurga, 1994). And reducing the
WIP level is associated with the reduction of shop lead-time (Tersine, 1991).

The level of finished goods inventory was measured on open-ended scales by

asking the following: (1) the ratio of excess and surplus finished goods inventory to total
production of finished goods (Question III-5), (2) service level in percentage (Question
III-6), (3) annual inventory turnover ratio (Question III-7), and (4) delivery lead-time to
retail customers (Question III-8). According to Tersine (1991), excess and surplus
inventories refer to the inventories that are produced over the required amount and of
which demand is discontinued respectively. These types of inventory are known to be
costly to hold and should be removed from the system. Therefore, the level of
unnecessary inventories such as excess and surplus inventories can be used to compare
the efficiency of the inventory management among manufacturing companies. On-time
delivery rate was measured to assess respondents' delivery performance to retailers' order
by asking them write their performance in percentage. Inventory turnover ratio is
obtained by dividing the sales by the average inventory for a given period (Branam,
1984). Usually, a high turnover ratio is associated with the low level of inventory and
recommended to be desirable (Tersine, 1991).

Three lead-time questions were asked by presenting the illustration that depicts
the inventory flow from material acquisition to product shipment (Question III-8). This
illustration was adopted and modified from Tersine (1991, p.16). The same questions
regarding the level of inventory by the type weree asked for fashion goods and basic

67
goods respectively in the assumption that most apparel manufacturers deal with both
product line categories and different inventory performance would be found in ordering
fabrics, producing and delivering finished garments.

Besides the detailed inventory performance measures specified above,

respondents were assessed their expectation level of improvement in inventory
performance as a result of SCM activity. To be more specific, expectation measures were
measured by averaging the five expectation scores which assessed their expectation level
of improvement in five inventory performance measures (i.e., fill rate, order lead-time,
on-time delivery rate, turnover ratio, overall inventory level) (Question I-2). Basic
assumption of this measure is that respondents who are aware of the benefits of SCM
might be more active in implementing SCM activities.

For demographics information, two open-ended questions were used to ask

companies' size and annual total production volume (Question IV-4). In addition, SIC
code was asked to describe the profile of the respondents (Question IV-1). A multiple-
choice question was presented.

Reliability and Validity of the Instrument

To check the construct validity of the SCM activity items, factor analysis with a
varimax rotation was done to test whether all 26 items load on six dimensions as
specified. A priori determination was used to determine the factor numbers. Eigen value
of each factor was compared with the results of a priori labels. Items that loaded higher
than .6 (i.e., correlation value between variables and factors) on one factor and lower than
.4 on other factors were accepted (Jones, 1999; Kincade & Vass, 1998 Malhotra, 1996).
The most commonly used Cronbach's alpha (internal consistency method) were
addressed to check reliability of the scales for the SCM activities (see Question I-1 to I-5)
and the other Likert-type statements used to measure the relationship with fabric supplier
and retail customers, respectively (see Question V-2, V-3). Cronbach's alpha value of 0.7
is known to be adequate for internal consistency (Leedy, 1997).

68
 Overall, developing the questionnaire from questions previously used in industry
research and based on literature analysis enhances construct validity of this study. In
addition, to increase face validity, major terms (i.e., fashion gods, basic goods) in the
question was accompanied with the definition in the instrument (e.g., Question II). In the
pilot test, the expert panel including college professors of clothing and textiles and
management science departments and six apparel manufacturers who were excluded from
the final sample reviewed the questionnaire and recommended changes to clarify and
eliminate ambiguous meanings in instructions and questions (Fowler, Jr., 1993; Jones,
1999; Lin, Kincade, & Warfiled, 1995).

Data Collection

Data were collected following the self-administered mail survey method as

suggested by Dillman and Christenson (1974) and Hoinville and Jowell (1977). Self-
administered mail survey has the advantages of relatively low cost and easy access to
widely dispersed samples (Fowler, 1993). Questionnaires were mailed to a stratified
random sample of 1,195 apparel manufacturers listed in the purchased directories, The
National Register of Apparel Manufacturers, Women's and Children's Wear (Marche
Publishing, 1998) and The National Register of Apparel Manufacturers, Men's and Boys
Wear (Marche Publishing, 1997). The questionnaire was addressed to the plant manager
or inventory planner who was assumed to be in charge of all the activities of developing,
executing, and delivering the product line. These people were also assumed to be aware
of the general characteristics of the company.

Data collection procedures were adopted and slightly modified from previous
research on apparel manufacturers (Jones, 1999; Ko, 1993; Priyadarshi, 1996). In the first
mailing, the questionnaire was sent with (1) the cover letter that informed an overview of
the aim of the survey, assurance of the confidentiality, identification of the researchers,
and the sponsoring university, (2) a self-addressed stamped return envelope, and (3) a
courtesy card that respondents could receive the summary of the findings of the study and

69
they could participate in the incentive program. For the incentive of participating in the
survey, a drawing was done. Out of respondents who replied, two respondents had a
chance to win $200. Intensive follow-ups and the incentive program were used to
increase the response rate (Hoinville & Jowell, 1978). After the sending the
questionnaire, a follow-up postcard was sent to the sample one week later, thanking those
participants who replied and stressing the importance of cooperation to non-respondents.
To non-respondents, the follow-up letter emphasized both the urgency and the
importance of a response with a replacement questionnaire, another return envelope and
courtesy card. To decrease the non-response bias, telephone-interview was done with a
random sample of non-respondents. The characteristics of non-respondents were
described to compare with those of respondents and the reason for non-response were
identified so that the survey results could be adjusted to account for non-response. All
data collection materials were attached in the Appendix.

Data Analysis

To test the model in Figure 1-1, four research hypotheses were built. Based on
the research hypotheses, detailed null hypotheses are presented in this section. Statistical
methods such as cluster analysis, factor analysis, ANOVA, chi-square, correlation, and
regression were used to test the null hypotheses that state no relationship between
variables. Based on the results of statistical analyses the null hypotheses were rejected or
failed to reject. A probability of p>0.05 was used to determine significance for each
testing. Data were analyzed using the SPSS 8.0 program. If at least one variable was
found to be significant, the hypothesis was rejected.

Before testing the null hypotheses, descriptive statistics were used for all variables
to describe the mean, frequency, percentage, and correlation of each value of the response
as well as to profile the respondents' demographic characteristics (i.e., firm size,
manufacturer type) and to explore the preliminary relationships between variables.

70
H10. Levels of SCM activities among apparel manufacturers are not different.
To test this null hypothesis, hierarchical cluster analysis and factor analysis were
employed. First, agglomerative hierarchical cluster analysis were used to cluster apparel
manufacturers into k number of relatively homogenous groups based on levels of SCM
activities. Squared Euclidean distance measure, the most commonly used measure of
similarity (Malhotra, 1996), and Ward's method, one type of clustering procedures to
minimize the within-cluster variance, were employed. The group number, k, was
determined based on the relative sizes of the clusters. Another hierarchical cluster
analyses with a reduced sample that consisted of randomly selected respondents were
done to assess the reliability and validity of the results of the cluster analysis as Malhotra
(1996) suggested. This cluster analysis method was based on previous research in
creating typologies (Jones, 1999; Shim & Kotsiopulos, 1993).

Next, a factor analysis was used to compare the SCM activity level among the
clusters. Factor analysis could abstract the reliable items that could significantly explain
the variance of each factor. After the SCM dimensions were abstracted, to explain
significant characteristics of each classified cluster in terms of SCM dimensions,
ANOVA and a post hoc test (i.e., Tukey test) were used by identifying the mean
differences of the SCM dimensions identified by factor analysis. According to Howell
(1997), Tukey test is the favorite pair-wise test to compare mean differences used in
many studies. Each SCM dimension score was obtained by dividing the summation of
the scores assigned to each activity question by the number of the questions under the
corresponding dimension.

H2a0. The levels of SCM activities do not differ based on apparel manufacturers'
product characteristics.
Product characteristic was measured on the continuous scale that indicates the
ratio of fashion goods and basic goods to the total products produced by apparel
manufacturers respectively. ANOVA and Tukey test were employed to see the mean
difference of product line characteristics between SCM activity levels.

71
H2b0. The levels of SCM activities do not differ based on apparel manufacturers'
production system.
Production system was measured on the categorical scale (i.e., bundle system,
PBS, UPS, modular system, others). Chi-square, ANOVA, and Tukey test were used to
evaluate whether SCM activity levels are different among apparel manufacturers who
adopt different production systems.

H2c0. The levels of SCM activities do not differ based on apparel manufacturers'
fabric suppliers.
Fabric suppliers in this study were measured based on three variables: cost
offering, delivery performance, and the nature of the relationship with the fabric
suppliers. Four Likert-type questions were adopted from Artz (1999) study. The
delivery performance score was obtained by dividing the sum of three scores by three.
The nature of the relationships (i.e., adversarial vs. partnership-like, short-term vs. long-
term, QR-oriented vs. non-QR-oriented) with the fabric supplier were the average formed
by dividing the sum of three question scores by three. For these two aspects (i.e.,
delivery performance, relationship), ANOVA, Tukey test, and correlation analysis were
used to identify any significant relationships with SCM activity levels.

H2d0. The levels of SCM activities do not differ based on apparel manufacturers'
retail customers.
Retail customers in this study were measured based on four aspects: relative size
of the retailers, nature of the relationship with the retailers, and type. The size and the
nature of the relationship with the retailers were measured on six-point Likert-type scale.
The nature of the relationships (i.e., adversarial vs. partnership-like, short-term vs. long-
term, QR-oriented vs. non-QR-oriented) with the retailer will be averages by dividing the
sum of three question scores by three. For these four aspects (i.e., retailer size,
relationship with retailers), ANOVA,Tukey test, and correlation analysis were used to
identify any significant relationship with SCM activity levels. The type of retailers (i.e.,
department store, specialty chain store, mass merchants/discounters, small independent
stores, others) was measured on the categorical scale. Chi-square , ANOVA, and Tukey

72
test were used to evaluate the relationships with SCM activity levels. Three additional
Likert-type questions were used to describe the retailer characteristics. In the current
study, the retailer type was defined based on the merchandise mix. The type of retailer is
sometimes determined by cost orientation and ownership by the manufacturer
(Donnellan, 1996; Jarnow & Dickerson, 1997; Rabolt & Miller, 1997). Therefore, cost
orientation of retail customers and ownership by the manufacturer were additionally
analyzed to fully describe the relationship with the type of retail customers

H3a0. Apparel manufacturers' product characteristics are not related to their

inventory performance of fashion goods and basic goods in terms of level
within type.
A pair of eight detailed inventory performance questions and one expectation
measure to profile the inventory levels by the type were asked on the continuous scale
(i.e., two for raw materials, two for WIP, and the remaining four for finished goods
inventory). The relationship between the four apparel manufacturers' characteristics and
inventory levels within the type of each product category were evaluated using
correlation analysis, ANOVA, Tukey test, and multiple regression analyses. Correlation
analysis was used to examine the relationship between product characteristics and
inventory performance measures.

A pair of detailed inventory performance questions was asked based on the

product characteristics (i.e., fashion goods vs. basic goods). Matched sample t-test was
used to evaluate the mean differences of inventory levels within type between product
category. The results of the matched sample t-test were expected to reveal how
differently apparel manufacturers deal with the product characteristics in planning raw
material purchasing, production, and shipment within a company.

H3b0. Apparel manufacturers' production system, fabric suppliers, fabric suppliers,

retail customers) are not related to their inventory performance of fashion
goods and basic goods in terms of level within type.

73
 Chi-square analysis was used to see the relationship between SCM activity
clusters and production system. Both variables were measured on categorical scale.
ANOVA and Tukey test were used to examine mean differences of inventory
performance measures between production system.

H3c0. Apparel manufacturers' fabric supplier characteristics are not related to their
inventory performance of fashion goods and basic goods in terms of level
within type.

H3d0. Apparel manufacturers' retail customer characteristics are not related to their
inventory performance of fashion goods and basic goods in terms of level
within type.
A multiple-regression was used to identify the relationship between chain partners
(i.e., fabric suppliers, retail customers) and inventory performance measures. In the
regression, predictors are fabric suppliers, and retailer customers, and criterion is eight-
inventory performance measures and one expectation measure. To diagnose the effect of
multicollinearity, VIF (Variance Inflation Factor) was calculated. As the variance
inflation factor increases, so does the variance of the regression coefficient, making it an
unstable estimate. Large VIF values are an indicator of multicollinearity.

H40. Apparel manufacturers' level of SCM activities is not related to their inventory
performance of fashion goods and basic goods in terms of level within type.
ANOVA, Tukey test, and multiple regression were used 17 times to evaluate the
direct relationship between the SCM activity level and inventory performance for all 17
inventory questions. ANOVA and Tukey test were used to examine the mean difference
of 17 inventory performance measures among SCM activity clusters identified by
through the cluster analysis. Multiple regression was used to see the relationship with
SCM activity level and inventory performance measures. In this analysis, predictor is
SCM activity level. Regression analysis was used to reveal which specific SCM activity
dimension, would influence the inventory performance in the apparel industry.

74