.v u Jm -
wit? um warm“... x f.§fi@fl
I I I

 

 

 

 

 

 

t.

, .3
W. . 9

find“ . }
.luann
$9.1",

‘ '5.-

hr 3 Ii:-

: n...

gt“. Maul";

dflzsdvfl.t&1n.fl t
I! I}.

"\.Jl-n
v." .59 ~
tits—RA?!"
‘1’. \Q,‘~v‘lxlrtu| 2
iiiufigui
n. l. .
.fuisxut...‘

in
tn...

0’ 2

II: :1
Inlinlvaol‘lo’itiv-
‘9?!- s
tit‘ix a.

S
IQtIII...‘
I 31!... 10.12».
:‘ailivfifuulc:
‘ I...‘ I. .
.«tfiVrfluu‘
.3 ii...“ I.
‘I‘Ail‘i‘ttvz
fiihviffth

it

u
.11! $90.. 5‘
.3...

5"
fit. A.

t...

a!
n.

 

 

f”) UBRARY
Mich 'r. State

I :4“!
UIIIVCIDIly

 

 

 

This is to certify that the
dissertation entitled

MEASUREMENT OF ART AND CULTURAL CONSUMPTION
BEHAVIORS FOR DEVELOPING STRATEGIC
COMMUNICATIONS '

presented by

Doc Syen Kang

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Media and lnforrnation Studies

 

 

7 A 5km W

Major Professor” 5 Signature

 

IZ°L0.07 iz-zooofi

 

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProjIAcc&Pres/CIRCIDaIeDue.indd

 

MEASUREMENT OF ART AND CULTURAL CONSUMPTION
BEHAVIORS FOR DEVELOPING STRATEGIC
COMMUNICATIONS

By

Doo Syen Kang

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Media and Information Studies

2009

ABSTRACT

MEASUREMENT OF ART AND CULTURAL CONSUMPTION BEHAVIORS FOR
DEVELOPING STRATEGIC COMMUNICATIONS

By

Doo Syen Kang

Art and cultural organizations are in dire need of promotion. The two main

target audiences include the public at large and legislators at all levels of government.

The industry is struggling due to unstable economic conditions, a lack of systematic

support from law makers, and a largely apathetic general public. To encourage the

economic development of the sector, organizations need to identify how people are

engaged in art and cultural consumption, how those behaviors are related, and what

affects their behavioral patterns. Information such as this will be useful for developing

the efficient and successful messages targeted organizations’ primary audiences.

To obtain the valuable and practical data for the creation of successful

messages, this research investigates three research questions: 1) Are there any

variables which explain patterns in art and cultural consumption behaviors?; 2) Are

there any relationships among those variables?; and 3) Can people be categorized by

relationships found among the variables? Based on the research questions, this

research is informed by two theories that provide the conceptual frame for developing

the questionnaires for the survey; Theory of Planned Behavior (TPB) and Social

Cognitive Theory (SCT).

The general public in the Great Lakes States was asked to participate in the
data-driven Web survey. The survey questions were divided into several sections
based on the theories that guided this study. Normal Ogive Harmonic Analysis Robust
Method was used to check the validity of using a multiple dimensional analysis. As a
next step, Multiple Item Response Theory examined the valid items, determined their
associations, and categorized the items as constructs of measurement.

The results indicated that people who attend art performances have a high
probability to participate in activities related to the performing arts and vice versa. In
addition, once people become exposed to art and culture, regardless of whether they
were engaged in voluntarily or not, they seem to have a more positive attitude toward
participating in art and cultural activities later. Those who enjoy traveling are more
likely to participate in a variety of other cultural activities, such as attendance at
different performances and even donation to cultural organizations.

This study identifies how people behave differently and determines which
items are valid for segmenting people based on the engagement levels. By interpreting
the results in accordance with the goals of campaigns, communication planners can
develop more efficient and target-oriented messages and strategies. The persuasive
messages targeting a specific group of people will boost more participation in the art

and cultural sectors and play a role to enlighten people and policy makers.

ACKNOWLEDGEMENTS

All of the things I have done at Michigan State University have had a great

impact on my personal and scholarly growth. First of all, I would like to sincerely

thank my committee co chairs. Dr. Timothy Levine aided in my development of a

robust commitment to research and a strong need to know what is going on behind

simple statistical tests. He opened another world to me which led to my challenge to

pursue another master’s degree in statistics. Dr. Lucinda Davenport was always

supportive of my decision and made me recognize how important balanced

perspectives are to a researcher.

I would like to express appreciation to my committee members. Dr. Teresa

Mastin acted as a mentor to me by standing by my side to share my stress and

difficulty as a doctoral student, instructor and researcher. I. don't believe many

graduate students are given the opportunity to develop a personal and intimate

relationship with any of their committee members. I thought it would be difficult to

get her guidance and support when she decided to leave Michigan State University for

DePaul University. Now I am so glad that I can continue our relationship by working

at DePaul University together. And Dr. Edward Mahoney in the Department of

Community, Agriculture. Recreation and Resources Studies (CARRS) offered me a

precious chance to work with the various local groups and to be engaged in a giant

project funded by the Michigan government. I am also grateful to the Department of

CARRS for allowing me to use the data collected through the project for my

dissertation.

In addition I would like to thank Dr. Mark Reckcase in the Department of

Measurement and Quantitative Method for his input in my dissertation, especially in

the method part, valuable knowledge and accessibility. He was the one who was

knowledgeable of the new method which I used for my dissertation, but was very new

to the area of communication. I would like to thank Dr. Alexander von Eye in the

Department of Psychology. He showed me unconditional support for and trust to my

ability as a researcher. He taught me positivity, flexibility and skepticism as well.

Finally, and most importantly, I would like to thank both my friends and

family, especially to Victoria Fitton. She always listened to my complaints as a mother,

made me recognize the importance of contributing to a society as a scholar and

conveyed an enthusiasm and excitement about teaching as a professor. And I owe my

deepest gratitude to my perfect husband, Young Ho Kim. His quiet support and

patience made this dissertation possible. No matter how hard I try to express thanks to

him, it would not be enough. Without his warm care and help, it would have been

impossible for me to go through this challenging journey.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................. vii
LIST OF FIGURES ........................................................................... viii
CHAPTER 1
INTRODUCTION ............................................................................. 1
Problem Statement ........................................................................ 1
Research Objectives ...................................................................... 5
Research Scope and Questions .......................................................... 7
Conclusion ................................................................................ 10
CHAPTER 2
THEORETICAL FRAMEWORK AND LITERATURE REVIEW .................... 12
Audience Research ....................................................................... l3
Attitude and Behavior Linkage ......................................................... 20
CHAPTER 3
METHOD ....................................................................................... 33
Sampling ................................................................................... 33
Survey Method: Web-based Survey .................................................... 37
Measurement .............................................................................. 45
Analysis Method: Multidimensional Item Response Theory (MIRT) ............. 59
CHAPTER 4
ANALYSIS OF RESULTS ................................................................... 70
NOHARM ................................................................................. 7O
TESTFACT ................................................................................ 73
CHAPTER 5
FINDINGS AND IMPLICATIONS ........................................................ 90
Results and Discussion .................................................................. 90
Future Studies ............................................................................ 107
Implications ............................................................................... 109
Conclusion ................................................................................ 1 15
APPENDIX .................................................................................. 119
REFERENCES .............................................................................. 125

vi

LIST OF TABLES

Table 1 Use and value of research measures among public relations practitioners...15

Table 2 Demographics of the respondents ................................................ 36

Table 3 Principles for the design of Web surveys and their relationship to traditional

sources of survey error ........................................................................ 43
Table 4 Unique variance of 107 items ...................................................... 71
Table 5 Factor loadings ...................................................................... 72
Table 6 Item statistics of 107 items .......................................................... 73
Table 7 Item statistics of 30 items .......................................................... 77
Table 8 Chi-square and degrees of freedom ............................................... 79
Table 9 NROOT largest latent roots of the correlation matrix ........................... 80
Table 10 Item parameters and descriptive statistics in the four dimensions .......... 81
Table 11 Standardized difficulty, communality and factor ................................ 83
Table 12 PROMAX rotated factor correlations ............................................ 84
Table 13 Inter-item correlation matrixes .................................................... 85
Table 14 Items to define the three dimensions ............................................ 109
Table 15 Eight clusters based on the three dimensions ................................... 112

vii

LIST OF FIGURES

Figure 1 Diagram of Theory .................................................................. 27
Figure 2 Sampling .............................................................................. 35
Figure 3 Perceived path model of the variables ............................................ 51
Figure 4 Relationship between difficulty and facility ..................................... 75
Figure 5 Histogram of 30 items ............................................................... 78
Figure 6 Dendrogram for clustering of 30 items (initial correlation) ................... 88
Figure 7 Dendrogram for clustering of 30 items (smoothed correlation) .............. 89

viii

Chapter 1. Introduction
This chapter explains the motivation for the current research. The research
problem is identified and the reason this research is important is explained. This
preview also contains specific research questions corresponding to the research
objectives. The case is made that these research questions address theoretically
interesting issues of practical importance. This introduction begins with a problem

statement.

Problem Statement

The unstable economic and social environment highlights a critical challenge
facing art and cultural organizations: promoting the entire industry to two primary
target audiences, the public and policy makers. With the low awareness of the
significance of the art and cultural industry, organizations suffer from dwindling
engagement from the public and depleted support from legislators. To boost the
economic development of the art and cultural sectors, organizations need to collect
scientific data for verifying the value of art and culture and for developing efficient
messages to get across to both target audiences. Unfortunately it is difficult for art and
cultural organizations to work collectively, because they are all struggling to survive

and their associations cannot afford extensive research projects.

The art and cultural sector is believed to be an industry with great potential
for the development and vibrancy of a society, despite indefinable outcomes that are
not seen now, but which might be substantial later (Hall, 2000; Markusen & Schrock,
2006). Without immediate consequences, however, policy makers doubt the
importance of the sector in regard to economic development, and the public has a
tendency to be less involved in art and culture-related activities than in any other
consumption behaviors.

Moreover, the art and cultural sector is often not identified as a top area of
concern with respect to resource allocation and community support, as the educational
benefits and economic impacts of the industry have been in controversy. One recent
example reflecting the underestimated value of the arts, culture and art education is
Governor Jennifer Granholm’s executive order to close the Michigan Department of
History, Arts and Libraries and split its functions among seven other departments and
some outside groups, including possibly Michigan State University.

Though it is evident that art and culture are the essence of human history, it is
hard to expect both policy makers and consumers to put emphasis on those disciplines
when they struggle with a sluggish economy and consider belt-tightening in
consumption. What needs to be done is not always identical to what people are doing.

Felkins (2001) summarizes several ways to remove these kinds of contradictory and

inconsistent qualities that are called social dilemmas: l) privatizing to the extent

possible; 2) directly tying the return to the cost; 3) depending on small groups; and 4)

using coercion and other outside inducements. In the case of art and culture, the fourth

one appears most appropriate.

Art and culture-related organizations are in dire need of governmental support

as a back up when voluntary public cooperation falls short. For example, when

environmental groups expect little chance of aggressive and voluntary environmental

action from the population, they try to get protective legislation passed. A similar case

can be made in regard to funding of the arts. When art and cultural organizations and

educators know that the public will not voluntarily help out, they need to persuade

policy makers to move in favor of them for giving the public what they need. Their

needs include government subsidies which will help art educators teach art and

cultural classes, help art and cultural producers such as artists and business owners

keep doing their jobs for a living and help organizations increase public awareness

and ultimate sales by planning community events and communication campaigns.

Many types of organizations can play a substantial role in advocating art and

cultural producers and encouraging cooperation from the public. News media can

provide educational campaigns effective in bringing both the public and policy

makers on board with what needs to be done. Strategic communication campaigns can

be operative to enhance the public’s awareness of the importance of art and cultural

activities and also to encourage policy makers to address the potential of the art and

cultural industry in boosting the economy.

To develop persuasive and efficient messages for both audiences,

organizations need to collect scientific information on how people behave and think.

Persuasion is much easier if the message is compatible with a person’s general

disposition toward a subject (Wilcox & Cameron, 2009). For example, if legislators

are presented with data that supports the theory that art education at an early period of

life is closely related to art consumption at a latter period in life, they will be more

confident with keeping or even increasing the budget for art education. And if

organizations know what target audiences want to hear, they will be better able to

establish successful communication plans.

The reality, however, is that public arts organizations are mostly small- and

medium-sized organizations that cannot afford to collect massive and systematic data

from the public. Each organization spends its effort and limited resources for

implementing the limited data collection of its own consumers or artists. These efforts

do not help organizations work together for the shared causes. Primarily based on

their experience and intuition, organizations claim the connection between art

education and art consumption, in addition to the contribution of the art and cultural

industry to economic development. Unfortunately, it is difficult to convince policy

makers without supporting data.

Therefore, both organizations and policy makers will benefit from the current

research, which will provide data for organizations to use to persuade the public and

policy makers. Both will get robust rationale for their stances and important data for

developing better messages appealing to the public.

Research Objectives

The most important component for organizations to identify when developing

strategic communication plans appealing to target audiences is how people think and

behave. Specifically, it is pivotal to know how people are engaged in art and cultural

consumption, how those behaviors are related, and what affects their behavioral

patterns. The data collected will be conducive to the development of more persuasive

and target-oriented messages.

This research aims to help organizations develop and implement strategic

communication campaigns that target both policy makers and the general public that

will eventually contribute to advocating for art and cultural producers and educators,

and the discipline itself, and enhancing cultural economic development. A

communicator wants to inform and or persuade; a recipient wants to be entertained,

informed, or alerted to opportunities that can fulfill individual needs. Audiences come

to messages for very different reasons: 1) surveillance of the environment to find out

what is happening that has some impact on them; 2) entertainment and diversion; 3)

reinforcement of their opinions and predispositions; and 4) decision making about

buying a product or service (Wilcox & Cameron, 2009). This explains why the

communicators must tailor messages that focus on getting the target audiences’

attention. It is when messages match target’s needs and expectations that messages are

most likely to be effective.

Although communication campaigns come in many varieties and cannot

always be placed into mutually exclusive categories, the field is increasingly making a

distinction between at least two main types of campaigns based on their primary

purpose, or what the campaigns are ultimately trying to achieve: individual behavior

change versus policy change campaigns (Dorfman, Ervice & Woodruff, 2002; Henry

& Rivera, 1998). Individual behavior change campaigns try to promote behaviors that

lead to an improved individual or social well-being. Policy change campaigns attempt

to mobilize public and decision maker support for policy support or change (Coffman,

2002; Dorfman et al., 2002; Weiss & Tschirhart, 1994). In some ways, campaigns for

both purposes look similar, and especially the art- and culture-related campaigns

should target both sides at the same time for better performance.

The collection of information assists in shaping these types of campaigns and
is similar to the practical marketing research approaches and methods that commercial
marketers use (Balch & Sutton, 1997). It aids in defining the scope of the problem,
identifies possible campaign strategies, provides information about the target audience,
finds what messages work best and how they should be framed, determines the most
credible messengers, and identifies the factors that can help or hinder the campaign
(Valente, 2001).

Therefore, the current study identifies and understands how the market
generally works, which factors have great impact on public consumption behaviors,
how those factors are correlated and how the target audience might be segmented by
the indexes of their involvement in art and culture-related activities. It will contribute
to developing efficient messages for the public and policy makers to be more involved
in art and cultural activities and to recognize the importance of the industry from

long-term and comprehensive perspectives.

Research Scope and Questions
The research objectives present a clear idea of where and how the research
should begin. This research incorporates three questions corresponding to the research

objectives: 1) Are there any variables that explain patterns in art and cultural

consumption behaviors?; 2) If those variables exists, are there any relationships

among those variables?; and 3) Can people be categorized by relationships found

among the variables?

According to Grunig and Repper’s (1992) situational theory of publics,

people uninterested or uninvolved in a topic will not process messages deeply, but

those more interested will be more likely to elaborate and to think more carefully

about the message. The most effective campaign will combine various strategies to

address the needs and interests possessed by different target publics or to address the

challenges presented at different steps in the persuasion process. For example, a

campaign might use an affective strategy to pique the public’s interest in an issue and

follow that with a more logically based message to deepen the public’s understanding

of issues.

The target population of this research must be the general public without a

priori segmentation. With the data derived from the research, the campaign planners

can categorize and cluster the public by their beliefs, attitudes, and behaviors in the

cultural sectors for establishing a customized campaign plan. Their attitudes will be

measured and their behavior will be classified to assess meaningful patterns. With this

information, the campaign strategy can be more target-oriented and the campaign

message will be more target-specific. In addition, the theoretical frame or identified

patterns might be utilized in the future to evaluate the effect of campaigns and to

gauge changes in behavior.

Research about how people behave and think in a specific area, so—called

audience research, is becoming an increasingly important tool for arts administration

and policy making. Audience studies have been used to assess public attitudes

towards the arts, to determine the composition of audiences for particular institutions,

to provide input into decision making, to provide baseline data for market

development programs, and to estimate the economic impact of arts activities on

specific communities and states (DiMaggio & Useem, 1979). Information gathered

can be further used in identifying marketing opportunities and evaluating the

contribution and performance of the cultural economic development strategy over

time. It might even be used to articulate the market and non-market values of arts and

culture.

To satisfy these purposes of audience research, the following research

questions are asked: 1) What kind of factors are valid for measuring public

consumption behavior in the art and cultural sectors?; 2) Are there any associations

between the valid factors?; and 3) Can those associations be used for distinguishing

publics and for explaining their different behavioral patterns and unobserved

characteristics?

This study will explore how people are engaged in art and cultural activities,

reveal any stereotyped patterns and salient differences in cultural consumption, and

analyze them through a valid clustering of audiences. It will contribute to identifying,

understanding, interpreting, and predicting audience behaviors in the art and cultural

industry, and will enhance planning a more efficient and target-oriented

communication campaign. The current research does not include an economic impact

study that focuses on obtaining the estimated values, such as the number of employees

and sales amount, by plugging some self-administered numbers into the econometrics

formulas.

Conclusion

Despite the dire need to boost the art and cultural sectors, organizations suffer

from low awareness of the industry from their main audiences, the general public and

legislators. To make the industry more vibrant, there is a need for active engagement

from the public and solid support from law makers. There is a need to encourage

people to participate in art and cultural consumption and pass legislations favorable to

the arts.

Given this situation, organizations need reliable data for developing

persuasive messages that appeal to the general public and policy makers.

Organizations can move people better with messages developed from the reliable data

of people’s behavior and interests. Legislators like to have scientific data of how the

whole industry works and what is conducive to the vibrancy of the art and cultural

sectors as rationale for passing laws designed to support art and cultural organizations.

For example, if there is a positive connection between art education in childhood and

art consumption as an adult, law makers will be more likely to support the passage of

art education-related laws.

In this context, the current research aims to identify how people behave as art

and cultural consumers, to determine the factors which influence consumers’ attitudes,

beliefs and behaviors, and to categorize people based on their behavioral patterns. To

obtain these objectives, three research questions are asked about factors that can

measure consumption behavior in the art and cultural industry, the associations of the

factors, and the usability of the associations as categories to represent different

behavioral patterns and unobserved characteristics. The results will help organizations

develop successful campaigns, of which the ultimate goal is individual and public

behavioral change.

Chapter 2. Theoretical Framework and Literature Review

This section explains the theoretical background for the current research. The
rationale for the approach taken by this study comes from audience research and
attitude and behavior linkage theories respectively. With respect to audience research,
the general focus-of the research and how it is developed is introduced, while two
specific theories are selected for explaining the linkage between attitude and behavior:
Theory of Planned Behavior and Social Cognitive Theory. Even though the survey
questions are developed based on both theories, the purpose of this study is not to test
any specific theory. The theories in the current undertaking play a role as schemas to
account for the findings of this research.

Audience research represents the importance of knowing audience attitude
and behavior for customizing messages to the needs of target audiences. The objective
of this study is to identify and understand behavioral patterns in art and cultural
consumption for developing efficient messages for public and policy makers to be
more engaged in the art and cultural activities and recognize the importance of the
industry itself. Therefore the history and development of audience research will be
helpful for better understanding why the research is important and how the results can
be utilized.

Linkage of attitude and behavior is described by the two widely used theories:

the Theory of Planned Behavior (TPB) and the Social Cognitive Theory (SCT). The

comparison of the main concepts highlights the similarity and the difference of the

two theories and how much researchers are interested in advancing the linkage of

attitude and behavior. With this long history and popularity of the two theories, the

association between attitude and behavior is regarded as the appropriate theoretical

framework for developing the survey questionnaire, as there is no existing study of

behavioral patterns in the art and cultural industry.

Audience Research

Knowing your audience and tailoring messages to your audience are twin

rules of effective persuasion. Therefore, knowledge of audience characteristics such

as beliefs, attitudes, concerns and lifestyles is an essential part of persuasion. The

information helps the communicator tailor messages that are salient, answer a felt

need, and provide a logical course of action (Wilcox & Cameron, 2009). Analysis data

give more information of the target public’s buying habits and ways of spending

leisure time than demographic information such as age, gender and income.

Campaign organizers execute an audience segment study to ensure the

consistency, integration, and proper timing of all campaign activities directed to a

given target audience. The messages, tactics, and strategies of the campaign differ

13

according to the audience and may change over time. Tailoring the campaign's

messages to respond to specific audience segments based on their behavioral patterns

requires use of a variety of messages and channels. Research must be conducted to

ensure that the campaign is properly positioned in the minds of the campaign’s

diverse target audiences ("Integrated communication plan," 2002). Apart from making

it easier to develop more specific, relevant, and appealing messages, segmenting the

target audience also helps to choose the appropriate communication tools, determine

the best ways to distribute communication products, and develop adequate incentives

to effectively promote the desired behavior ("Outreach campaign planning - target

audience," n.d.).

According to a national survey of 300 professionals assessing their research,

public relations practitioners recognize the values of more fundamental information

than media management for achieving their campaign goals and bringing out audience

change in their actual behavior (Austin & Pinkleton, 2006). They adopted the easier

items to quantify such as media coverage and volume of media pick ups, instead of

implementing research to collect more useful information that could be applied for a

variety of usage such as measurement of change in attitude, knowledge and ultimately

behavior (see Table 1). Without scientific data to compare pre- and post-

communication campaigns, it is impossible to implement strategic communication

plans and to claim that audience change is derived from the success of a

communication plan and the messages directed to the target audience. Therefore,

communicators’ frustration appears to lie in the discrepancy between communication

practices which focus on getting more messages covered by media and what needs to

be done with conducting and applying research.

Table 1 Use and value of research measures among public relations practitioners

 

Mean for Use (Rank) Mean for Perceived Value (Rank)
Advertising equivalency 3.7 1 (9) 337(9)
Volume of media pickups 543(2) 457(8)
Favorable media treatment 535(3) 4.96(7)
Media coverage 547(1) 519(5)
Changes in awareness 495(4) 564(4)
Changes in knowledge 485(5) 590(2)
Changes in attitudes 474(6) 577(3)
Changes in sales 407(8) 508(6)
Changes in behavior 464(7) 602(1)

 

 

Note. All measures on 7-point scales. with I indicating less use or value and 7 indicating more use or

value.

The idea of audience research has been characterized according to its

emphasis on message receivers. This concept has been applied to several areas of

research including marketing, advertising, and communication under a variety of

names, such as consumer, market segment or audience research.

In the field of mass communications, there has been a widespread and

enthusiastic call for convergence in theoretical approach and research traditions

(Radway, 1984; Seiter, Borchers, Kreutzner & Warth, 1989; Billig, 1991; van Dijk,

1991; Silverstone & Hirsch, 1992). This call for convergence has centered on the

audience of mass communication and generated a body of research which drew upon

diverse theoretical and methodological traditions (Livingstone, 1993).

A revisionist movement has emerged in media and cultural studies. One of

the key notions in the movement was a re-conceptualization of the audience as an

active producer of meaning. Meaning was constructed through the interaction

between text and the social and discourse positions of audiences (Curran, 1990).

Morley (1980) showed that divergent groups responded in very different ways to

certain media and that these differences reflected the different discourses and

institutions in which people were situated. He illustrated the importance of different

sub-cultural formations within the same class in generating different audience

responses. From this perspective, one can no longer talk about the effects of a

message on a homogeneous mass audience who is expected to be affected in the same

way.

Another aspect of the perspective that audiences perceived mass-

communicated meanings differently was the tendency of people to seek out media

content that reinforces what they think and to avoid content that challenges their

beliefs (Curran, 1990). Existing research supported that subcultural formations within

the audience influenced the extent to which media representations were accepted or

not (Brigham & Giesbrecht, 1976; Vidmar & Milton, 1974). A study revealed a more

complex picture in which groups with different clusters of attitudes — whether

ethnocentn'c, authoritarian or traditionalist — responded to the series in partially

different ways (Wilhoit & de Bock, 1976). Meyer (1976) produced an integration of

some of the themes of the revisionism that different types of audiences, bringing

different beliefs, attitudes, and values to the viewing of the show as a result of

different socialization processes, are affected in distinctly different ways. The mass

audience has been shown to be significantly heterogeneous, particularly in relation to

gender, class, culture, and age, but also in relation to cognitions, involvement, and

styles of viewing, as both reception and ethnographic research have demonstrated

(Livingstone, 1993).

Katz and Liebes (1985) described exposure to and perception of media

content as highly selective, governed by the attempt of the individual to avoid

cognitive dissonance. Consequently, the intended or anticipated audience response

could be resisted or twisted if the message challenged the preconceived opinions of

the audience (Schroder, 1987).

In the field of marketing, which is well known for advanced approaches to the

market or audience, the marketing concept is an underlying frame to explain why

market research is important. The marketing concept calls for most of the effort to be

spent on discovering the wants of a target market and then creating the goods and

services to satisfy them (Kotler & Zaltman, 1971). This concept is the key to

achieving organizational goals and consists of determining the needs and wants of

target markets (Kotler, 1980). An organization takes the time, effort, and financial

expense of gathering information about present or prospective exchange partners. If

marketers strive to understand exchange partners and tailor offerings for them,

information is a necessary preparatory step to developing that proper blend. Added

information can be used to reduce organizations’ effort in making exchanges. This

value is directly related to the effort required to make exchanges (Houston, 1986).

The marketing concept does not urge marketers to depend solely on

marketing research for guidance. Dependence on customers’ expressions of their own

needs and wants suggests that some marketers fail to take a long—run view of the

marketing concept. Consumers are not necessarily good sources of information about

their needs. They don’t necessarily know how they will react under different

environmental conditions. Sometimes consumers have to learn about new

technologies, beliefs, and ways of behaving. Anticipating future needs and wants is

consistent with the marketing concept (Houston, 1986). This doesn’t mean that

marketers reject the marketing concept, but customer information must not be limited

to the current, expressed needs and wants.

18

Duncan and Moriarty (1998) found common elements in marketing and

communication: relationship exchange, feedback, and information signs/signals. The

notion that communication is a central integrative process in marketing is

demonstrated in the evolving theories of integrated and relationship marketing.

Zinkhan and colleagues, in a special session on Integrated Marketing

Communications (IMC) and relationship marketing at Emory University's 1996

relationship marketing conference, argued that these two disciplines are

complementary metaphors.

There are also important points of intersection between communication and

marketing theories that support the argument for a communication-based model of

relationship marketing. Both marketing and communication are in the midst of

fundamental changes that are similar in origin, impact, and direction (Duncan &

Moriarty, 1998). Parallel paradigm shifts move both fields from a functional,

mechanistic, production—oriented model to a more humanistic, relationship-based

model. Therefore, comprehensive and various frames to view the target audience or

consumers are important.

As researchers advocate that there are interchangeable notions in the two

disciplines, marketing and communication/public relations, communication campaign

designers are advised to integrate the different, but applicable ideas and practices from

marketing field into communication planning and evaluation. Undeniably, marketing

has made a great progress in detecting, measuring, and predicting the public in the

name of the consumer. The application of advanced techniques in marketing to the

communication field will contribute to finding more accurate and realistic

measurement and evaluation methods in communication campaign planning and

implementation.

In this context, in-depth audience research should be designed to make an

inference of unobserved characteristics from the explicit attitudes and behavior. This

research made an effort to find other dynamics to moderate and affect the relationship

between the attitude and behavior of the audience participating in arts- and culture-

related activities. Categories to segment audience were not limited to consumer

demographics examples, like age, gender, occupation, income, marital status,

presence of children in the household, and geographic location of residence.

Comprehensive study would be conducive to tailor future communication campaigns

to the specific target audience, to create a mutual understanding of the importance of

that market, and to draw collaborative endeavors from the public.

Attitude and Behavior Linkage

The theoretical structure between attitudes and behavior is a basis for

20

justifying the goals and strategies for public relations campaigns. McGuire (1989)

wrote that successful communication campaigns depend on a good understanding of

two types of theories: those that explain “how someone will process and respond to a

message” and those that explain “why someone will or will not respond to a message

in desirable ways.” The connection between attitudes and behavior is assumed to

provide an accountable infrastructure for the “how” and “why” questions mentioned

above. Organizers believe they are able to influence, predict, and change people

through strategic media campaigns. The media are seen as potential agents of remote

control, possessing the ability to persuade individuals to change their attitudes and

behavior (Schroder, 1987). This also appears to be applied to establishing the

successful marketing plans.

The American Marketing Association defines consumer behavior as the

dynamic interaction of affect, cognition, behavior, and the environment (Bennett,

1995). Marketing managers have recognized the importance of consumers and have

sophisticated approaches and detailed data from which to develop successful

marketing strategies (Peter & Olson, 2005). In order to change or influence consumer

behavior, marketers need to understand why people behave the way they do (Fishbein

et al., 2001 ). In other words, their strategies and plans need to be based on underlying

dynamics. The conceptual models tell practitioners and managers what to do and in

21

what sequence.

Frameworks provide a basis for both research and strategy (Andrease, 1995).
Just as theory is important for marketing strategy, it is also important for evaluation of
marketing strategy. Providing the strategy with a theoretical base can both support its
development and serve as a basis for its implementation and evaluation (Valente,
2001). Strategic marketing planning does not need to rely on just one theory. Theories
can be used in combination, and many strategies have integrated more than one of
these into their designs and planning.

Since LaPiere’s study (1934), which is probably the most widely cited study
in attitude-behavior relationship research (Fazio & Roskos-Ewoldsen, 1994), there
have been a considerable number of studies executed to investigate the link between
attitudes and behaviors. As Campbell (1963) concluded, the degree of correspondence
between attitude and behavior is, for the most of part, yet to be discovered. Allan
Wicker’s conclusion (1969) that it is more likely that attitudes will be unrelated, or
only slightly related, to overt behavior than that attitudes will be closely related to
actions, led researchers to look at the attitude-behavior relationship from either a
methodological or a moderator variable perspective (Berger & Mitchell, 1989). This
dichotomous perspective is still predominant in recent studies. Researchers focus on a

method-analytic review including meta-analysis (Baggozi & Bumkrant, 1979; Dillon

22

& Kumar, 1985; Hausenblas, Carron & Mack, 1997; Armitage & Conner, 2001),

while others place emphasis on addition or substitution of moderators to increase

accountability and predictability (White, Ten'y & Hogg, I994; Wiesner & Tesch-

Romer, 1996; Van Ryn, Lytle & Kirscht, 1996).

Theory of Planned Behavior

In an effort to find a more accountable relationship between attitudes and

behaviors, Ajzen and Fishbein created the Theory of Planned Behavior (Ajzen, 1985;

Ajzen, 2001; Ajzen & Fishbein, 2000), which is one of the most popular theories

shared in public relations, advertising, and marketing ("Theory clusters," n.d.). The

Theory of Planned Behavior (TPB) is extended from the Theory of Reasoned Action

(TRA) (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975), which suggests

performance of a given behavior is primarily determined by the intention to perform

that behavior. Many public relations campaign studies have used the intention factor

as a predictor of future behavior. Thus, to predict whether people will be engaged in a

specific behavior, researchers measure their intention to do or not to do it.

Two major factors influence those intentions; l) a person‘s attitude toward the

behavior and 2) a person’s subjective norms about the behavior, which is the belief

that the people important to the person think he or she should or should not perform

the behavior. In turn, attitudes and subjective norms are influenced by behavioral and

normative beliefs (Peter & Olson, 2005). Compared to the theory of reasoned action,

the model of planned behavior includes the additional measures of control beliefs (the

perceived frequency of occurrence of salient facilitating or inhibiting factors) and

perceives behavioral control and the additional causal relationships between the

variables (McElreath & Blamphin, 1994).

Studies in tourism, one of the primary sectors in the art and cultural industry,

integrated the variable of past behavior in the TPB (Lam & Hsu, 2004, 2006). Ajzen

(1991) criticized the use of past behavior on grounds that it offers no explanatory

content, but previous empirical tests of this theory found that past behavior can

enhance prediction of intention and/or future behavior (Triandis, 1977; Berkowitz,

1982; Doney & Cannon, 1997; Quellette&Wood, 1998; Abraham & Sheeran, 2003).

Researchers try to find more applicable variables to predict intention to travel to a

specific region, e.g., specific country, wine tourism, and rural tourism.

Lam and Hsu (2004, 2006) tested TPB with potential travelers from Mainland

China to Hong Kong. The questions in their study included: Will travelers‘ past

experiences at a destination affect their choice of the destination again in the future?

Are travelers influenced by personal expectations and experiences? Will expectations

of others affect travelers’ choice of a destination? What is the role of social influence

on travelers’ decisions? Will facilitating or inhibiting factors affect the decisions?

24

Results showed that data fit the TPB model moderately well and explained

respondents’ traveling intention. Attitude, perceived behavioral control, and past

behavior were found to be related to traveling intention.

Another study in tourism (Sparks, 2007) tested TPB’s ability to predict tourist

intention to take a wine-based vacation using perceived control with past attitude.

Wine/food involvement, normative influences, and the three wine expectancy-value

dimensions (core wine experience, destination experience, and personal development)

contributed to the intention to take a vacation to a wine region. This study also

included past experience with wine and destination in the model.

Maestro et a1. (2007) examined tourists’ attitudes toward rural tourism. The

findings confirmed the positive effect of attitude toward rural tourism on perceived

quality and the moderating role of familiarity. Experience moderates the relationship

between attitude toward rural tourism and satisfaction and perceptions of quality. As a

result, the more experienced the tourists, the less altered the evaluation result, i.e., the

more accurate the predictive measure. In tourism research, researchers put value on

past experience as a powerful moderator to enhance the correspondence between

attitudes and behaviors of potential tourists.

Social Cognitive Theory ( SC T)

Another popular theory that is frequently compared to TPB and prevalent in

25

health communication campaigns research is Social Cognitive Theory (SCT). This

theory suggests that self-efficacy — the belief that one has the skills and abilities

necessary to perform the behavior under various circumstances — and motivation to

perform the behavior, are necessary for behavior change (Bandura, 1997). Many

researchers have striven to find the similarity and/or difference between PBC in TPB

and self-efficacy in SCT, which is summarized in the later section that compares TPB

and SCT. Self-efficacy and PBC are perceived as very similar concepts. And self-

efficacy appears closely related to prior behavior, like prior experience, exposure and

knowledge, which is found to be one of the most important factors affecting

behavioral change in tourism.

Figure 1 provides a graphical explanation of the theory. A person must

believe he or she can perform the behavior in various circumstances and have an

incentive (positive or negative) to do it (Parraga, 1990). This mostly factors into one

of the important public relations campaigns, health-related campaigns and research,

such as anti-alcohol, condom use, and anti-drug programs, to have a direct

relationship with behavioral intentions for future health-related behaviors. One of the

goals of health-related campaigns is to encourage people to do what organizations

want people to do. Therefore it appears to be appropriate to take self-efficacy into

account for knowing what affects behavioral changes in art and cultural consumption.

26

P

B4 #13

 

P: Personal factors, B: Behavior, E: Environmental

Figure 1 Diagram of Theory (Bandura, 1986)

Bandura’s (1997) social cognitive theory explains how people acquire and
maintain certain behavioral patterns, while also providing the basis for intervention
strategies. He believes the way people perceive their ability to cope with a situation
and to control its outcome (perceived self-efficacy) influences how people behave in a
given situation.

Comparison between TPB and SC T

In general, behavior is viewed as the result of a set of beliefs about specific
products or services and a set of affective responses to the behavior. The beliefs are
represented by behavioral beliefs and outcome evaluations in TPB and by outcome
expectations in SCT. Together they are referred to as the net benefits (realized or
expected) accruing from use of the system (Seddon, 1997). Affective responses are
typically measured by attitudes toward use and an individual's evaluation of the
behavior as either positive or negative. These commonalities in the models reflect a

belief in the cognitive basis of behavior.

27

SCT and TPB include other beliefs that might influence behavior independent
of perceived outcomes (Compeau, Higgins & Huff, 1999). The TPB model
incorporates the notion of perceived behavioral control as an independent influence
on behavior, recognizing that there are circumstances in which a behavior might be
expected to result in positive consequences (or net benefits) yet not be undertaken due
to a perceived lack of ability to control the execution of the behavior (Ditsa, 2002).
Perceived behavioral control encompasses facilitating conditions, such as perceptions
of resources and technology (Thompson, Higgins & Howell, 1991), as well as
perceptions of ability or self-efficacy (Taylor & Todd, 1995). SCT gives prominence
to the concept of self-efficacy, defined as beliefs about one's ability to perform a
specific behavior, recognizing that our expectations of positive outcomes of a
behavior will be meaningless if we doubt our capability to successfully execute the
behavior (Compeau et al., 1999).

Since Ajzen (1991) argued that the PBC and self-efficacy constructs were
interchangeable, several researchers have suggested that the two variables are
different notions and have compared how they work differently in predicting a
specific variable. Terry (1993) suggested that self-efficacy and PBC are not entirely
synonymous. Researchers advocated the use of measures of self-efficacy, as opposed

to PBC, in the prediction of intentions and behavior (Armitage & Conner, 1999). De

28

Vries, Dijkstra, and Kuhlman (1988) tested the model through attitude and behavior
toward smoking. Attitudes and subjective norms were significant predictors of the
intention (not) to smoke, but self—efficacy expectations added significantly to the
prediction of the intention. Self-efficacy also had a direct effect on behavior after
controlling for intention. In this study, self-efficacy highly reflected the actual control
or the skills of the adolescents. Furthermore, it appeared that non-smoking
adolescents had higher self-efficacy expectations towards non-smoking than smokers.
Dzewaltowski, Noble and Shaw (1990) also found that self-efficacy directly
impacted behavior. Social cognitive theory and the theories of reasoned action and
planned behavior were examined in the prediction of four weeks of physical activity
participation. The social cognitive theory variables significantly predicted physical
activity participation, with self-efficacy and self-evaluation of the behavior
significantly contributing to this prediction. The greater the confidence in
participating in physical activity and the greater the satisfaction with present physical
activity, the more physical activity was performed. Perceived control and intentions
did not account for any unique variation in physical activity participation over self-
efficacy. Therefore, the social cognitive theory constructs were better predictors of
physical activity than those from the theories of reasoned action and planned behavior.

On the other hand, there are studies which focus on the conceptual and

29

empirical distinction between PBC and self-efficacy in the attitude—behavior

relationship, instead of obvious opposition to PBC or strong support for self-efficacy.

White, Terry and Hogg (1994) reported that perceived behavioral control only

affected behavioral measures of discussing the use of condoms with any new partner

for safer sex behaviors, while self-efficacy had a strong effect on the intentions to

discuss and the intentions to use condoms. A study was undertaken in the context of

HIV—preventive behaviors of sexually experienced heterosexual undergraduates. The

study employed a longitudinal design, with the proposed predictors of performing the

behaviors under consideration assessed (using a condom every time you have sexual

intercourse during the next month and discussing whether to use a condom with any

new partners during the next month) prior to the measures of reported behavior. A

group norm emerged as a distinctive predictor (in addition to attitude and subjective

norm) of intentions to practice both safer sex behaviors. The data also revealed some

support for the distinction among the different measures of control (self-efficacy,

perceived behavioral control, and planning).

In contrast, Terry and O’Leary (1995) found that self-efficacy only predicts

intentions, while PBC predicts exercise behavior. Their study was concerned with the

prediction of intentions to engage in regular exercise (for at least 20 minutes, three

times a week for a fortnight) and actual exercise behavior. Undergraduate subjects

30

participating in the study revealed support for the view that separate measures of self-

efficacy and perceived behavioral control should be employed in the theory of

planned behavior. Efficacy expectancies were found to influence behavioral intentions,

but not actual behavior. In contrast, levels of perceived behavioral control appeared to

have no effect on behavioral intentions, but emerged as a significant (positive)

predictor of actual behavior. There was also evidence that the effects of intentions on

behavior are moderated by the level of perceived behavioral control.

Povey, Conner, Sparks, James and Shepherd (2000) also examined the

application of TPB on two dietary behaviors, particularly focusing on the roles of

perceived control and self-efficacy as two components of the perceived behavioral

control construct in the TPB. First, the general public was asked whether they eat five

portions of fruit and vegetables per day or eat a low—fat diet, and then they were asked

for actual eating behavior one month later. In addition, the individual components of

perceived behavioral control (perceived control and self-efficacy) and their

determinant beliefs were examined. For each behavior, the TPB variables were found

to be good predictors of intentions, although less good at predicting behavior, with

self-efficacy being consistently more predictive than perceived control.

As a further step, Armitage and Conner (1999) suggested the need for multi-

method assessments of behavior. Their study consisted of four parts: (1) to apply TPB

31

to eating a low-fat diet; (2) to consider differences between self-efficacy and PBC; (3)
to examine self-identity as a potential addition to the model; and (4) to identify beliefs
which may provide useful targets for interventions. The findings supported the
efficacy of the TPB as a predictor of a low-fat diet. Evidence for the distinction
between self-efficacy and PBC was provided through the finding that self-identity
independently suggested intention and might prove a useful addition to the TPB.

Based on these comparative studies, this research integrates both theories into
the conceptual model for developing the questionnaires for the survey. The theoretical
background provides a guideline of which items should be asked; attitude, prior
experience and behavior. As this is an exploratory study which is not designed to test
and/or confirm the model or theory, however, TPB and SCT play a significant role not
as theories for comparison to identify which one fits better in explaining art and
cultural behaviors, but as a theoretical frame for developing the relevant questions to
be asked. The theory-based questions also will be useful for analyzing the findings

later based on the assumed connection among the latent characteristics.

32

Chapter 3. Method

This chapter introduces sampling, survey method, survey questions, and the
statistical methods used to collect and analyze the data. The research sampled the
general public in the Great Lakes States, using a panel list randomly selected by the
professional sampling company. Those sampled were asked to participate in the on-
line survey, which was developed by a data-driven system. People can skip, stop or
continue to answer the questions based on their previous answers. The survey
questions were divided into several sections based on the theories introduced in
Chapter 2. To analyze the data, multiple item response theory was used for taking the

interaction between people and items into account.

Sampling

Sampling is a powerful tool that helps researchers obtain accurate information at a
reasonable cost (Austin & Pinkleton, 2006), whereas a bad sampling frame results in
fatal damage to the generality of the research findings. A perfect sample would be a
scaled-down version of the population, mirroring every characteristic of the whole
population. Of course, no such perfect sample exists for a complex population.
However, a good sample will reproduce the characteristics of interest in the

population as closely as possible. Sampling details of the study follow.

33

Target population - General Public in Great Lakes States

An on-line survey used random sampling of households in Michigan and
other Great Lakes states including Ohio, Wisconsin, Illinois, Indiana, and Minnesota,
who had access to the lntemet, specifically e-mail account owners.
Sampling method -— Disproportionate Stratified Random Sampling

Sampling was conducted based on “Product Multinomial,” which stopped the
survey once the response reached the total of at least 6,000; 2,000 from Michigan and
4,000 evenly distributed from other Great Lakes states. It was implemented under the
assumption of the homogeneity of multinomial distribution and avoided a cell count
under 5 in any category of each multinomial distribution (Rice, 2007). Michigan
residents were oversampled for a further comparison study.
Sampling frame — Panelists

The professional sampling company, Survey Sampling International (SSI)',
recruited panelists through thousands of Web sites, which were designed to optimize
the probability that the panel reflected the overall composition of the reachable
population online.
Sampled population

Through an invitation letter and two reminders sent by e-mail, 6,110 out of

 

For more rntonnatron on sampling process, please vrsrt http://ir'wri:surr-reysanrplrrm.com

34

the listed individuals completed the survey. Figure 2 gives a graphical idea of who

were sampled for the survey. Demographics of the respondents were compared to

those of the lntemet users, not the US. population, as the study included lntemet

usage as one of the indicators in the model. Table 2 presents the specific composition

of the respondents.

Non-res nse Non-cove e
p0 Sampled - rag
- Not reachable — No access to Internet
Population
- Refuse to
Response

Figure 2 Sampling

The collected data contained more White—Caucasian (90%), more female
(63%), more educated (98% had at least high school diploma) and wealthier people
(44% got more than $50,000 a year) than the population as a whole. This is a general
phenomenon in academic study-based “volunteers” and “Internet users.” The
respondents were somewhat biased from the population standpoint, but were deemed
suitable for the purpose of this study, which was to examine variables closely related
to the lntemet and cultural consumption. The observation unit was the individual
lntemet-user. The pre-set goal, 6,000 responses was achieved in 10 days, from Nov.

17 to Nov. 27, 2006.

35

Table 2 Demographics of the respondents

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

All Michigan Other
Res ndents Great Lakes
Gender __
Male 36.80% 26.50% 42.40%
Female 63.20% 73.50% 57.60%
Family Status __ _
Single without children 23.00% 21.20% 23.70%
Single with children 7.70% 9.00% 6.80%
Single with children not at home 8.10% 7.80% 8.10%
Married without children 12.90% 1 1.20% 13.90%
Married with children 29.50% 31.00% 29.00%
Married with children not at home 18.90% 19.70% 18.50%
State of Residence
Michigan 34.90%
M 65-10%
Education _ _
Less than high school 0.40% 0.50% 0.40%
Completed some high school 2.00% 2.10% 2.00%
Completed high school 19.50% 19.90% 19.50%
Some college 33.50% 34.90% 32.80%
2-year college degree 1 1.40% 12.20% 1 1.00%
4-year college degree 18.40% 17.10% 19.20%
Completed some graduate courses 5.30% 4.50% 5.50%
Master‘s degree 7.60% 7.30% 7.60%
Doctoral degree 1.90% 1.60% 2.00%
Sexual Orientation
Straight/heterosexual 94.00%
Gay/lesbian 2.50%
Other 3.50%
Age gMean) 45.4 44.9 45.6
Employment _
Full time outside the home 40.20% 37.00% 42.10%
Part time outside the home 14.00% 15.30% 13.30%
Self employed outside the home 3.00% 2.90% 3.10%
Self employed at home 6.30% 6.40% 6.30%
Home maker 13.00% 15.60% 1 1.90%
Unemployed 9.10% 9.70% 8.80%
Retired 16.90% 16.00% 17.50%
Student 7.@% 8.30% 6.30%
Race
Caucasian or White 89.80% 89.80% 90.80%
African American or Black 5.70% 6.20% 5.30%
Asian or Pacific Islander 1.20% 1.00% 1.20%
Native American 1.10% 1.10% 1.20%
Income
Less than $25,000 20.10% 22.10% 19.00%
$25,000 - $29,999 10.00% 10.20% 9.80%
$30,000 - $39,999 14.30% 14.10% 14.50%
$40,000 - $49,999 12.10% 11.70% 12.20%
$50,000 - $59,999 10.30% 10.20% 10.40%
$60,000 - $74,999 12.40% 1 1.50% 13.00%
$75,000 - $99,999 1 1.00% 10.60% 1 1.20%
$100,000 - $124,999 5.00% 5.00% 5.00%
$125,000 - $149,999 2.00% 1.80% 2.20%
$150,000 - $174,999 1.10% 1.00% 1.20%
$175,000 - $199,999 0.70% 0.80% 0.70%
$200,000 or more 1.00% 1.10% 0.90%

 

 

 

36

Survey Method: Web-based survey

Today’s research methods are fast developing and mutually incorporated, in

terms of technique and content, in tune with the fast-paced outside environment and

its complicated needs. As a result, it is difficult to single out one method that is sure to

collect reliable and valid data to explicate, explain, and predict the target audience.

Many people have access to the lntemet, and the lntemet is fast becoming the

communication method of choice for many Americans (Duffy, 2000). With the market

growth of user-friendly programming software and an increasing population with

access to the lntemet, the Web survey is emerging as one of the frequently used

research methods (Couper, 2000). Researchers from many disciplines are starting to

see the benefits of collecting data using the lntemet, and journals are publishing data

that have been collected online (Schleyer & Forrest, 2000).

Cost effectiveness

Given the same number of completed surveys, Web surveys can collect data at

dramatically lower costs than traditional methods, such as mail or telephone surveys.

The time it takes to collect data from Web surveys is shorter than any other method. In

this study, using the professional sampling company cost as high as postage, printing,

and stationery costs in mail survey, but when considering the processing cost, like

data entry that is extremely expensive and time-consuming, it appears extremely cost

37

effective and also eliminates potential errors in rekeying data (Granello & Wheaton,

2004).

Great number of respondents

The power of Web surveys is that they make survey data collection available

to the masses. Researchers can get access to numbers of respondents at dramatically

lower costs and in shorter time than traditional survey methods (Farmer, 1998; Lazar

& Preece, 1999).

Source accessibility

Any researcher can put survey questions on dedicated sites, offering low-cost

services, and collect data from potentially thousands of people. User-friendly

programming democratizes the data-collecting process (Couper, 2000).

High-technique

Unlike on paper, the appearance of a survey can vary according to a

researcher’s purposes. Design may be much more flexible for Web surveys and

increase readability, because there are more tools available to the designer (color,

sound, images, animation, etc.; Couper et al., 2001). Design should be used based on

survey purpose and target population. The survey necessitates data-driven

programming in which respondents had only to answer the questions tailored by their

previous answers, due to the multi-layered sets of questionnaires.

38

With the growing population of lntemet users, the Web-based surveys are
becoming more frequently used as a method to collect the data. The potential risk of
Web surveys is that with the proliferation of such surveys, it will become increasingly
difficult for people to distinguish between good and bad surveys. Nowadays, people
are so bombarded with survey requests that they either tune out completely or base
their participation decisions on the content, topic, entertainment value, or other
features of the survey. Four traditional sources of survey error- sampling, coverage,
measurement, and non-response-that exist with any other survey methods, including
mail and telephone surveys were considered carefully as follows:

Coverage error

Coverage error is the biggest hindrance of Web surveys due to the limited
accessibility (Couper, 2000; Dillman & Bowker, 2001). Yet when researcher try to
study the population who uses the lntemet when searching for information and
purchasing products, the Web survey is a good method for making the research more
inclusive of the desired target population (Couper, 2000).

According to statistics updated in June 2009, by lntemet World Statsz, there
are 251,735,500 lntemet users in the United States; 73.9 % of the population. In

principle, about 27% of the population cannot be reached when researchers use the

 

2 www.intemetworldstats.com/stats2.htm#north

39

Web-based survey (Fig.2). It is still far behind the penetration rate of telephones,

94.5%, but appears large enough to be considered a powerful medium to reach the

public. In addition, the error might be mitigated by setting up a clear limit of

generalization. The outcome will not be generalized the entire population, but limited

to lntemet users, who are actually the primary target audience of the communication

campaign.

Sampling error

Before sampling, the researcher sets the target number of the complete

sample (N=6,000) to increase the sample size for lower sampling errors and to

compensate for coverage and non—response problems that might exist. Statistical

simulation as well as mathematical proof shows that an increase in sample sizes

contributes to a decrease in sampling errors (Lohr, 1999).

In case N=6,000, the margin of sampling errors is 10.013 for 95% confidence

 

interval by the formula: CI = .4: Z(o/2)\l{p*(l-p))/(N-l) In addition to large numbers,
a traditional indicator of a high quality survey, it puts emphasis on the need to define
survey populations and to select probability samples by using the directory listing of
the professional sampling company.

Measurement error

Measurement error is the deviation of the answers of respondents from their

40

true values on the measure. Measurement errors in self-administered surveys can arise

from the respondent (lack of motivation, comprehension problems, deliberate

distortion, etc.) or from the instrument (poor wording or design, technical flaws, etc.).

To minimize measurement error, the survey instrument was made easy to understand

and to complete, and poor wording or design, comprehension problems, deliberate

distortion (Dillman & Bowker, 2001), and even technical flaws were checked by pre-

test to Michigan State University students, staff and professors (N=298) from October

25 to November 10, 2006, by snowballing and convenient sampling.

Non-response errors

Several studies have compared response rates from Web studies to those from

mail surveys of the same populations and have found that the Web surveys did not

reach the response rate levels of the mail surveys(Crawford, Couper & Lamias, 2001;

Nicholas & Sedivi, 1998; Tse et al., 1995). To compensate, there are findings that Web

surveys have significant advantages over mail surveys in terms of response rate

(Bachmann, Elfrink & Vazzana, 1999; Cobanoglu, Warde & Moreo, 2001; Kiesler &

Sproull, 1986; Parker, 1992). To increase the response rate, this study put credible

names (Michigan State University and the Michigan government) on each page of the

Web survey, placed confidentiality guidelines survey providers should comply with on

the front page, and provided attractive incentives (Singer, Groves, & Coming, 1999;

41

Singer, 2002), such as a cash lottery.

Blumberg, Davis, Khare and Martinez (2005) examined the effect of
extended calling protocols and refusal conversion effort on the data quality. They
found that respondents who were converted tended to be more like other cooperators
than the entrenched non-responders or non-covered. There will be no valid way to
integrate people who do not want to participate in surveys, and these people cannot be
captured by any kind of sampling frame into respondents. The additional effort to
increase response rates may also cause another question about “randomness,” which
means the equal probability to be included in samples. As a result, this study
compared the samples’ demographic files to census—based data of population to check
non-response error (Couper, Traugott & Lamias, 2001).

The list of the principles (Dillman & Bowker, 2001) includes:

1) Introduce the Web questionnaire with a welcome screen that is motivational,
emphasizes the ease of responding, and instructs respondents on the action needed for
proceeding to the next page.

2) Provide a PIN number for limiting access only to people in the sample.

3) Choose for the first question an item that is likely to be interesting to most
respondents, easily answered, and fully visible on the first screen of the questionnaire.

4) Present each question in a conventional format similar to that normally used on

42

paper, self-administered questionnaires.

Table 3 Principles for the design of Web surveys and their relationship to traditional
sources of survey error

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

List Type of Error Checked in
Sampling Coverage Measurement Nonresponse this SlUd)’
1. Introduction J J
2. PIN number J J J
3. First Question J J
4. Format 1 J J J
5. Color/Design J J
6. Convertibility J J J J
7. Instruction J J
8. Format 2 J J
9. Voluntary J J
10. Skip J J
11. Combined J J J
12. Programming J J
13. Process Inf. J J
14. Restraint J J J

 

 

 

 

 

5) Restrain the use of color so that figure/ground consistency and readability are

maintained, navigational flow is unimpeded, and measurement properties of questions

are maintained.

6) Avoid differences in the visual appearance of questions that result from different

screen configurations, operating systems, browsers, partial screen displays, and wrap-

around text.

7) Provide specific instructions on how to take each necessary computer action for

responding to the questionnaire and other necessary instructions at the point where

43

they are needed.

8) Use drop-down boxes sparingly, consider the mode implications, and identify each
with a “click here” instruction.

9) Do not require respondents to provide an answer to each question before being
allowed to answer any subsequent ones.

10) Provide skip directions in a way that encourages marking of answers and being
able to click to the next applicable question.

11) Construct Web questionnaires so respondents scroll from question to question
unless order effects are a major concern, and/or telephone and Web survey results are
being combined.

12) When the number of answer choices exceeds the number that can be displayed in
a single column on one screen, consider double-banking with an appropriate grouping
device to link them together.

13) Use graphical symbols or words that convey a sense of where the respondent is in
the completion process, but avoid ones that require significant increases in computer
memory.

14) Exercise restraint in the use of question structures that have known measurement
problems on paper questionnaires, e.g., check-all-that-apply and open-ended questions.

Among the principles shown above, only two suggestions were not met by

this study, # 11 & # 13. In regard to # 11, it was also recommended by other

researchers to use a mixed-mode strategy (e.g., mail surveys and Web surveys)

(Kaplowitz, Hadlock & Levine, 2004), but the study did not implement the combined

data-collecting techniques. Regarding #13, as the survey was designed under the

computer adaptive system, in which respondents had different questions next based on

their previous answers, the time of completing the survey varied.

Measurement

The pool of the preliminary questions was derived from the objectives of the

study, which were to identify how the market generally works, which factors have

great impact on public consumption behaviors, how those factors are correlated and

how the target audience might be segmented by the indexes of their involvement in art

and culture-related activities.

Unfortunately, it was hard to find any population studies in the art and

culture-related fields. Few organizations uploaded their annual survey reports, of

which samples were limited to their member agencies or artists. Therefore, the survey

questions were finalized with the help of significant input from a wide range of

Michigan arts and cultural organizations.

The process of designing the survey began in January 2006 with a review of

previous arts and cultural surveys, regardless of academic publication and industrial

45

reports, to identify the shared types of questions. Over the course of one year, there

were 59 Cultural Economic Development Online Tool (CEDOT) planning meetings

(phone and in—person) between Michigan State University (MSU) and the state of

Michigan Department of History, Art and Library (HAL). Eighteen CEDOT public

planning meetings also were convened by MSU and HAL staff to engage

representatives of Michigan‘s cultural sector, involving more than 150 participants. In

addition, CEDOT presentations were made at two statewide conferences and at a

national online e-conference involving state arts agencies representing 39 states.

Some of the participating organizations included ArtServe Michigan,

Michigan Libraries Association, Michigan Association of Community Arts Agencies,

Michigan Museums Association, Michigan Festivals and Events Association, The

Henry Ford Museum, Arts League of Michigan, Wayne State University Center for

Art and Public Policy, Michigan Department of Education, State Historic Preservation

Office, Michigan Historic Preservation Network, Michigan Council for Arts and

Cultural Affairs, Lawrence Technological University, Michigan Humanities Council,

National Assembly of State Arts Agencies, National Endowment for the Arts, National

Conference of State Legislators and more.

A series of meetings were arranged to determine the nature of the

participation and market information available, as well as to learn more about the

46

information needs of the participating organizations. Drafts of different survey
instruments were re-examined by HAL staff. Functionality of the Web survey and
understandability of the questions were pre-tested before the survey was implemented
in November. The results of the pre-test were used to fashion a final instrument,
which was reviewed by Michigan State University’s Social Science/Behavioral
/Education Institutional Review Board (SIRB) to insure it met all requirements (e. g.,
confidentiality) for protecting human subjects.

The survey questionnaires were developed through a number of meetings
with different organizations which were discretionally selected by the project client,
Department of History, Art and Libraries. Despite the seemingly low connection
between libraries and art and cultural consumption behaviors, libraries were actively
involved in the meetings and decision making processes. As shown in the results,
library usage was not closely related to the three dimensions which explicitly created
unique characteristics among the items under each dimension and was hard to be
labeled as one separate dimension that explained any patterns in art and cultural
activities.

In addition, the items for the analyses in this study were selected randomly
based on their level of variables, not based on theories. Even if MIRT is a very

conservative method which means there is no way to fix the problematic data, the

47

results might have limitation in generalization and the questions might have

difficulties with replication. Yet this would be regarded as a good starting point to

develop more research and study in the art and cultural sectors.

The survey was categorized into fourteen sections, which contained multiple

item indicators, to make it appropriate for a data-driven type of survey: 1)

demographics, 2) art experience in childhood, 3) donation, 4) membership, 5)

volunteering, 6) attendance, 7) participation, 8) purchase, 9) cultural tourism, 10)

library use, 11) lntemet usage, 12) information source. 13) tour-related decision

making, and 14) image of Michigan. The questionnaire consisted of screening

questions that came first in each part to filter respondents and subsequent questions

explicitly measured the specific art-related consumption only for those who answered

“yes” to the screening questions. Sub—questions under each category were asked to be

used for multidimensional and in-depth analyses when developing more detailed

communication campaigns with emphasis on a specific topic.

The survey included over 500 questions, but the data-driven system allowed

people to skip questions based on their previous answers. Each section started with a

few lead questions, like “have you ever attended any theatrical performances within

the last 12 months?” Each section aimed to dig into details of the respondents’ simple

and basic answers to “have you” questions by building up the questions to make

48

inferences to “how” and “why.”

The lead questions, which all of the respondents were asked to answer,
included: 1) Demographics: Age, Gender, Education; 2) Art education and exposure to
art in childhood and current art education; 3) Donation, membership, & volunteering
during the last 12 months; 4) Attendance at theatrical performance, music concerts,
and dance performance during the last 12 months; 5) Participation in performing arts,
visual arts, and art-related disciplines (for example, genealogy, archaeology and
architectural activities) during the last 12 months; 6) Purchase of visual art/craft and
CD, DVD, books, and etc. during the last 12 months; 7) Any pleasure trip during the
last 12 months, attendance at theatrical performance, visit to museum, festival,
historical sites, zoo, and library; and 8) Library card and library visits during the last
12 months.

In total, 107 items were tested by the MIRT method. The following factors
were considered in the process of selecting the 107 items for this research.

1) Research type: this research was an explanatory research which aimed at
finding reliable measurements to segment people. This was not a confirmatory
research, of which objective was to test a specific theory. In this context, this study
needed to include items as many as possible to identify the right questions to be used

for measuring people in terms of their art and cultural consumption.

49

2) Variable level: To analyze data with MIRT, dichotomous variables were
selected. Unlike the regression analysis, MIRT was developed from logistic formula.
Thus, it was necessary to pick one variable level for analysis. As the lead questions
were the nominal levels of variables, it appeared to be appropriate to choose "yes" or
"no" questions to get statistically meaningful results. Besides the above—mentioned
filtering questions, other consumer behaviors were included to gauge the activeness in
art-related consumption, for example, “have you purchased tickets?” and “were any of
these live theater performances in a different state or country other than where your
permanent residence is located?” were selected for analyzing their art and culture-
related behavior patterns.

3) Sample size of affirmative answers: as people were asked to answer many
questions based on their previous responses, quite a few people said "yes" to the
questions which were asked at the last of each section. The survey was data-driven, it
was difficult for people to have a chance to see the last parts of each section. This
meant that the most of people stopped answering the questions in each section at the
very beginning of each section or at most in the middle of each section and moved to
next section. Even though those in-depth questions would be useful for further
analyses, this study included only the questions, to which at least 10 percent of the

total respondents (at least 600 persons) said "yes." This was justified by the objectives

50

of this study, which were to identify valid items and utilize them for segmentation.

4) Theoretical framework: TPB and SCT played a role as a big frame for
selecting the items for analysis. Even though the research was not designed to test a
theory, it would be more meaningful to determine the reasonable items to be included
and to explain the results from the theoretical perspectives. The path model of the
variables represented the rationale of the selection process of the latent variables for

analysis and helps provide an understanding of those variables.

 

 

 

 

 

 

 

 

 

   

 

 

 

 

  
     
 
 
   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Art Current
Education in Eduézlittion * __;
Childhood \ Attendance
if ‘r ,
5x101???” \ Membership E
childhood \
Tl Donation : Partrcrpatron
If
vr r ' F ‘
/ ”Mr W...
1‘ A11
Cultural
Attitudes _ , T,°“""“ _
& Prior Experience ie— Behavroral Attitudes ——> Consumption Behavror

Self-efficacy & Self-evaluation ————F

Figure 3 Perceived path model of the variables

As a result, other continuous items, like lntemet usage and Michigan’s image

are excluded. Demographic variables, like age and gender, might play a significant

role in finding the differences among the consumers, but for the initial analysis they

51

are not included. They will be used for future analyses as criteria variables to split

people into more specific categories.

The brief summary of the 107 items (see Appendix) are

Attendance at theatrical performance (4 items)

1) Attendance (lead question)

2) In a different state or country or not

3) Purchase of tickets

4) Purchase of subscriptions

Attendance at dance performance (4 items)

5) Attendance (lead question)

6) In a different state or country or not

7) Purchase of tickets

8) Purchase of subscriptions

Attendance at music performance (4 items)

9) Attendance (lead question)

10) In a different state or country or not

1 1) Purchase of tickets

12) Purchase of subscriptions

Participation in any performing arts (10 items)

52

l3)

14)

15)

16)

17)

18)

19)

20)

21)

22)

Engagement (lead question)

Playing a musical instrument

Playing in band or orchestra

Solo singing

Group or choir singing

Dance

Acting

Comedy, magic and mime

Storytelling and readings

Other performances

Participation in any visual arts ( 11 items)

23)

24)

25)

26)

27)

28)

29)

30)

Engagement (lead question)

Drawing

Painting

Printmaking

Sculpture

Textile arts

Photography

Crafts

53

31) Folk or Traditional art

32) Media arts

33) Writing

Participation in any art-related or cultural disciplines (10 items)

34) Engagement (lead question)

35) Architecture

36) Landscape architecture

37) Fashion design

38) Graphic design

39) Interior design

40) Industrial design

41) Historic preservation

42) Genealogy

43) Archaeology

Professional artists (5 items)

44) Considered as a professional artist

45) Formal education

46) Engagement in any art-related professional development activities

47) Member of an art-related guild, union or association/organization

54

48)

Financial robustness

Purchase of Art (2 items)

49)

50)

Any visual arts

Any books, videotapes, DVDs, CDs, records, tapes or music downloads

Hiring of artists (22 items)

51)

52)

53)

54)

55)

56)

57)

58)

59)

60)

61)

62)

63)

64)

Hiring any performing artists (lead question)
Any solo musicians

Any musical bands/groups

Any solo singers

Any singing groups

Any actors or theater companies

Any comedians, magicians, or mime artists
Any performance artists

Any dancers or dance companies

Any DJs

Hiring any architects or designers (lead question)
Any architects

Any landscape architects

Any graphic designers

55

65) Any interior designers

66) Any fashion designers

67) Any industrial designers

68) Hiring any historic preservation/restoration, archaeological, or

genealogical services (lead question)

69) Any historic preservation services

70) Any archaeological services

71) Any genealogical services

72) Any home/building restoration services

Cultural tourism (8 items)

73) Trip to Michigan

74) Any pleasure trips

75) Attendance at a theatrical performance

76) Visit to a museum: art, history, science, or special topics museum

77) Attendance at a festival: art, crafts, music, food, wine, ethnic, antiques, or

car show

78) Visit to an historical site, attraction or community: heritage parks,

lighthouses, monuments, landmarks, archaeological sites, cultural centers

or historical communities

56

79) Visit to a zoo, aquarium, botanical garden or arboretum

80) Visit to a library: national, state, local community, or university/college

Library use (2 items)

81) Having a library card

82) Visit to a library

Behavior as a producer (3 items)

83) Any membership

84) Any donation

85) Any volunteering job

Prior experience (7 items)

86) Exposure to art in childhood: 1 items

87) Having education in childhood

88) Visual arts

89) Performing arts

90) Music (vocal/instruments)

91) Literary arts

92) Crafts

Art education now (7 items)

93) Having education now

57

94) Visual arts

95) History/Appraisal

96) Performing arts

97) Music (vocal/instrumental)

98) Literary arts

99) Applied arts (e.g., architecture, design, genealogy and archaeology)

- Art education for children (8 items)

100) Having children take any art classes

101) Visual arts

102) History/Appraisal

103) Performing arts

104) Music (vocal/instruments)

105) Literary arts

106) Applied arts (e.g., architecture, design, and genealogy and archaeology)

107) Crafts

These items initially incorporated in the statistical analysis were dichotomous

variables, to which people answered “yes” or “no.” The wide spectrum of the items

appeared to play an important role in finding better items to identify how differently

people behave and how they can be clustered.

58

Analysis Method: Multidimensional Item Response Theory (MIRT)

Tukey (1977) describes exploratory data analysis as detective work. A

detective investigating a crime needs both tools and understanding. If he has no

fingerprint powder, he will fail to find fingerprints on most surfaces. If he does not

understand where the criminal is likely to have put his fingers, he will not look in the

right places. Similarly, the data analyst or researcher of data needs both tools and

understanding. It doesn’t seem to be limited only to the case of exploratory data

analysis. Despite the ambiguous definition of statistical analysis, Tukey’s description

can be extended to all statistical analyses. He demonstrates how to affect graphic and

numerical analyses of the data at hand so as to understand them and suggests that the

emphasis on inference in modern statistics results in a loss of flexibility in data

analysis.

With a fast developing statistical approach, researchers seem to put more

value on learning something new than on guaranteeing something basic first. They

believe confirmatory statistics are more critical for making an inference of the

population. Without verifying the basic assumptions and conditions for running

software and calculating statistic parameters, the result will not have any meaning

academically, practically, and statistically.

In an effort to find a descriptive but comprehensive method,

59

multidimensional item response theory (MIRT) is employed for analyzing the

complex behavioral patterns. Depending on a psychometrician’s background, MIRT

can be considered either as a special case of factor analysis (FA) or structural equation

modeling, or as an extension of unidimensional item response theory (IRT) (Reckase,

1997). Relative to MIRT, however, factor analysis and [RT lack interest in the

characteristics of the variables.

There were many researchers who came close to a direct representation of

MIRT through factor analysis models. Horst (1965) tried to partial out the effects of

variation in item difficulty, but did not take the step of actually estimating item

parameters other than factor loading and of modeling probabilities of the correct

responses rather than the actual responses. Christoffersson (1975) and Muthen (1978)

presented the factor loadings and threshold estimates much as they were in MIRT, but

focused on modeling the hypothetical continuous items trait rather than the

probability of correct response. And the probability of a correct response was not

presented as a conditional function of item parameters and vectors of person

parameters, but rather the probabilities were modeled as population statistics. The

feature that was missing in their formulations was the conditional probability of a

correct response to each item as a function of a person’s location in the 0 space

(Reckase, 1997).

60

Additionally, McDonald (1967) indicated the relationship of the regression

function to the conditional probability of a correct response, but used a polynomial

model that could result in values for the dichotomous variable beyond the 0 to 1 range

and did not provide any interpretation for the characteristics of the item variables.

Finally, Bock and Aitkin (1981) provided a two-dimensional extension of the two-

parameter normal ogive model along with an MIRT-type parameterization of the item

characteristics, but still placed emphasis on defining factors rather than investigating

the interaction of persons and items.

In short, factor analysis and MIRT share the basic concept of defining

hypothetical scales that can be used to reproduce the data and of defining scales that

have an arbitrary origin and unit of measurement. Nonetheless, a clear difference

exists between two methodologies. The principal component of factor analysis is the

resolution of a set of variables linearly in terms of a small number of categories or

factors, and this can be accomplished by the analysis of the correlation matrices

(Harman, 1976). A correlation matrix is a data source that ignores all differences in

the means and standard deviations of the variables. MIRT differs from most factor

analysis representations in that the varying characteristics of the input variables are

considered to be of importance and worthy of study.

Another method with similarities to MIRT is Item Response Theory (IRT). It

61

has a number of potential advantages over classical test theory in assessing self-

reported attitude and behavior outcomes as psychometrics (Hays, Morales, & Reise,

2000). Lord (1980) summarizes that a test consists of separate items, and the test

score is a sum of item scores. In this case, however, statistics describing the test

scores of a certain group of examinees can be expressed algebraically in terms of

statistics describing the individual item scores for the same group. It cannot predict

how individuals will respond to items unless the items have previously been

administered to similar individuals. Researchers need to describe the items using item

parameters and the examinees using examinee parameters in such a way that they can

predict probabilistically the response of any examinee to an item, even if similar

examinees have never taken similar items before. This involves making predictions

about things beyond the control of the psychometrician-predictions about how people

will behave in the real world.

With IRT, individuals’ attitudes and behaviors can be scored and used to

categorize them by their activeness and responsiveness, according to each

measurement. And those outcomes can be grouped under one construct with

convergent characteristics. IRT analyzes both attitudes and behaviors toward each

item indicator and the relationship between total items under one construct and one

item indicator. So the individuals with the same score do not come up with the

62

conclusion that they are the same in attitude and behavior.

Yet, people are likely to bring more than a single dimension to bear when

responding to a particular question that seems to require numerous experiences and

abilities to determine their characteristics and behavior. This is especially true for

measures of attitudes and behavior in complex areas. Thus, although unidimensional

IRT models have proven to be useful under certain conditions, there is a need for

more complex IRT models that more accurately reflect the complexity of the

interactions between respondents and measured items (Reckase 2009).

Since the early developments of MIRT, many researchers have developed

descriptive statistics to assist in the interpretation of MIRT analyses (Reckase, 1985;

Reckase & McKinley, 1991), and multiple applications have been made. Miller and

Hirsch (1992) identified small but replicable clusters of items with clear substantive

meaning. Although the percent of variance accounted for by these items was less than

other factor analysis studies, the clustering might be important in identifying sub-

samples of the respondent population, such as specific target market. The MIRT

conception of dimensions of item sensitivity and locations of subgroups in a

multidimensional latent space has provided a natural means for describing differential

item functioning of test items. Davey and Oshima’s study (1994) started to equate

method for use on tests that assess multiple skills.

63

Researchers developed various mathematical expressions for the accuracy of
modeling the interaction between a person and a test item. MIRT models for test items
with two score categories, for example, yes or no in this study, can be grouped into
two categories: compensatory and partially compensatory models. Compensatory
models consist of the two-parameter logistic, the three-parameter logistics, the Rasch,
and the normal ogive models. These models have a relatively long history in
psychometric literature, and there is more experience with their application (Reckase,
2009).

The form of the multidimensional extension of the two-parameter logistic
(M2PL) model (Eq.1) has one thing different from the two-parameter logistic (2PL)
model (Eq. 2) Bimbaum (1968). The exponent of the form a(0-b) results in aO-ab by
multiplication. If —ab is replaced by -d, the expression is in what is called
slope/intercept form, a0 + (1. Thus easier items have higher values for d (DeMars,
2005). M2PL is very similar to the three-parameter logistic model (Eq.3) except for

the guessing parameter, c.

I

aiO. +d.

. 6 J l
U. =11 0. ,d. = . .
H U Pal I) 1+eai9j+di (1)

 

Where uij is the response on Item i by Person j, q is the trait level for Person j, al‘ is

a parameter related to the maximum slope of the item characteristic curve (ICC)

@(gj—bi)

 

P(l]lj:1|6j’ai’bi)=1+eai(gj‘bi) (2)

The M3PL model appears to be appropriate for accounting for observed empirical
data such as that provided in Lord (1980) which shows that respondents with low
capabilities have a non-zero probability of responding correctly to multiple-choice
items (Reckcase, 2009). The probability of selecting a correct response for individuals

with low capabilities does not seem to be related to the scores tested by the items and

it can be interpreted as a single lower asymptote, or pseudo-guessing, parameter, Ci-

ai0j +d,-

 

6

P(Ul.j =1|0j,ai,ci,di)=ci+(1—ci) 0,“!
1 + ear j i (3)

The multidimensional extension of the Rasch model is simply the M2PL model with

all of the a-parameters set to 1.0. Rasch (1960) proposed the simplest model that met

the required assumptions as follows. The original form is given by

 

AB,
Par,j =11Aj,Bl.)=1 :1 B
+ . . (4)
j l

where A J is the single person parameter now generally labeled 0j, and Bi is the

single item parameter now generally labeled bi. For the dichotomous case, using the

same notation as the other models, the multidimensional Rasch model is

65

ai0- +di
6 .1

 

Pa], =1|ai,d,,0j) =

3in +d,- (5)

1+e

where ai is a vector such that ai = bik, and di is a scalar value equal to aik'§ when

g is a vector of item difficulty parameters. The multidimensional extension of the

normal ogive model (Bock and Schilling 2003; McDonald 1999; Samejima 1974) is
expressed in Eq. 6 where Zi (0j) = aiBj' + di. If cl' = O, the result is the normal ogive
version of the two-parameter logistic model. This model defines the probability of
correct response for an item as the area under a standard normal distribution from —-
25(9j).

P(Ul.j =1| 0j,ai,ci,di) =q +(l—q)—1—— je7dt

2.71. "’4' (03') (6)

The other category of the multidimensional item response models is the
partially compensatory model. The partially compensatory product models are
consistent with the hypotheses that test items have different parts related to different
skills or knowledge and overall success requires success on each part. It means each
construct is independent of one another and the individual probability of selecting a
correct response, i.e., “yes,” is the multiplication of the total scores of each construct.

Unlike this assumption, the compensatory models are more consistent with a more

66

holistic view of the interaction of persons and test items. The usefulness of the models

will be determined by how accurately they represent the responses from actual test

items. Bolt and Lall (2003) found that the compensatory model fit the partially

compensatory data almost as well as the partially compensatory model. The partially

compensatory model, on the other hand, did not fit the compensatory data very well.

Thus this study selected the compensatory models for analysis.

MIRT depends on the quality of parameter estimates that are obtained from

the estimation computer software. The programs vary in the types of models that they

support and whether they estimate the person parameter vectors. According to some

parameter recovery studies (Beguin & Glas, 2001; Gosz & Walker, 2002; Maydeu-

Olivares, 2001; Miller, 1991; Stone & Yeh, 2006), Normal Ogive Harmonic Analysis

Robust Method (NOHARM) and TESTFACT software have been available much

longer than the others and neither of them is clearly superior to the other when

considering the recovery of item parameters. There are practical differences in the two

software packages.

Normal Ogive Harmonic Analysis Robust Method (NOHARM) (Fraser,

1998) is a program for estimating the item parameters of the normal ogive form of the

compensatory MIRT using the person-by-item matrix of item scores. It uses a

procedure for estimating the parameters of test items for MIRT models. NOHARM

67

does not provide an option for estimating the elements of the 0 -vector. Therefore it is

impossible to directly estimate the locations of individuals or the correlation between

0 —estimates (Reckase, 2009).

TESTFACT is a computer program for performing factor analysis of inter-

item tetrachoric correlations based on item response theory (Bock, Gibbons, Schilling,

Muraki, Wilson & Wood, 2002). It includes the estimation of the item and person

parameters for the multidimensional extension of the two-parameter normal ogive

model. The two-parameter normal ogive model contains the same parameter structure

as the two-parameter logistic model (Embretson & Reise, 2000). TESTFACT was

programmed assuming a standard multivariate normal distribution of coordinates with

zero intercorrelations. That assumption places constraints on the estimation of the

model parameters.

NOHARM does not provide estimates of the 0-vectors, which is a clear

limitation. TESTFACT provides estimates of the 0-vectors, but it is limited in the

number of dimensions that can be estimated because of the way the full information

aspect of the program uses computer storage. Both TESTFACT and NOHARM have

difficulty separating the dimension when the Os on those dimensions are highly

correlated.

In this context, this study uses both NOHARM and TESTFACT, which seem

68

to be the programs of choice if the data consists of dichotomous item responses
(Reckase, 2009). NOHARM is a free program that gives quick estimates of item
parameters, and TESTFACT has the advantage of having updated versions and more

complete documentation as a paid program.

69

Chapter 4. Analysis of Results
This chapter recaps the results identified through the statistical analyses. As a
preliminary analysis method, Normal Ogive Harmonic Analysis Robust Method
(NOHARM) is used to check the validity of using a multiple dimensional analysis. As
a next step, Multiple Item Response Theory (MIRT) examines the valid items,
determines their associations, and categorizes the items as constructs for measuring

unobserved characteristics of art and cultural consumption behaviors.

NOHARM

NOHARM was run in exploratory mode with the 107 items to see whether
MIRT was the appropriate analytical method for the data. This program does not
estimate the location of individuals, which is the purpose of the study, but can be used
to check whether the data can possibly be analyzed by the MIRT method. One of the
weaknesses the MIRT method contains is that there is no remedy to make data
collected fit in a model, like transformation in regression or bootstrap in structural
equation modeling. Once the results indicate erratic statistical estimates, it is
impossible to repair.

Three estimations were checked: residual, unique variance, and

dimensionality. First, the residual matrix of the items showed all of the values were

70

smaller than 0.038, which was calculated by the equation, dimensions x square root of

one divided by n. Three dimensions were tested and the total number of the responses

was 6,110. Sum of squares of residuals (lower off-diagonals) was 0.17 and root mean

square of residuals was 0.0055. Second, all of the unique variances indicated plus

values as shown in Table 4. Any minus values could not be changed by decreasing

dimension or increasing iteration.

Table 4 Unique variance of 107 items

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 7 8 9
0.128 0.538 0.349 0.581 0.196 0.771 0.622 0.618 0.581
10 1 1 12 13 14 15 l6 17 18
0.736 0.421 0.701 0.752 0.688 0.662 0.722 0.775 0.726
19 20 21 22 23 24 25 26 27
0.707 0.612 0.654 0.637 0.432 0.568 0.603 0.633 0.646
28 29 30 31 32 33 34 35 36
0.662 0.508 0.498 0.624 0.641 0.543 0.694 0.633 0.791
37 38 39 40 41 42 43 44 45
0.609 0.635 0.689 0.585 0.652 0.828 0.591 0.699 0.641
46 47 48 49 50 51 52 53 54
0.571 0.581 0.582 0.614 0.686 0.681 0.591 0.603 0.575
55 56 57 58 59 60 61 62 63
0.578 0.574 0.575 0.574 0.573 0.586 0.656 0.596 0.582
64 65 66 67 68 69 70 71 72
0.574 0.581 0.571 0.571 0.602 0.572 0.572 0.582 0.571
73 74 75 76 77 78 79 80 81
0.907 0.562 0.501 0.470 0.574 0.540 0.646 0.645 0.858
82 83 84 85 86 87 88 89 90
0.794 0.692 0.602 0.640 0.592 0.343 0.561 0.693 0.443
91 92 93 94 95 96 97 98 99
0.641 0.621 0.615 0.608 0.573 0.610 0.600 0.578 0.579
100 101 102 103 104 105 106 107
0.996 0.914 0.582 0.941 0.939 0.606 0.590 0.923

Lastly, the dimensionality was tested by giving a closer look at the factor

loadings under the randomly picked three dimensions. Table 5 illustrated that the

items could be separable based on the multi-dimensions as each item showed high

loadings in different dimensions. There were several items that measured

71

predominantly along with each factor and a few items that measured an equally
weighted combination. As a result, three check points demonstrated there was no

explicit problem with using the MIRT method for the analysis of this data.

Table 5 Factor loadings

Factor Factor

 

72

TESTFACT

The analysis of NOHARM found it problematic to include all of the 107
items in the model, as the data had quite a few items with almost no information in
them, like item 20 with the similar factor loadings under the three different
dimensions. Accordingly it appeared to be impossible to extract 6 or even 7 factors
from these data due to the lack of convergence. To drop the information items which
were not useful for measuring audiences’ behavior, basic items statistics were
estimated in Table 6.

Table 6 Item statistics of 107 items

ITEM

 

S.D. F DIFF IS

73

Table 6 (cont’d)

 

74

Table 6 (cont’d)

N

 

*MEAN: the mean score on the survey for the entire group
RMEAN: the mean score on the survey for those subjects who get the item correct which means
answering yes to the item.

Table 6 shows how each item contributes to the total scores through a
measure of item discriminating power. How much people can be separated by each
item is calculated based on the relationships between difficulty (DIFF) and facility
(FACILITY), which means the proportion that gets the item “yes.” As shown in
Figure 4, they are generically negative-correlated. Intuitively the more difficulty in

saying yes to the item, the smaller the proportion of people saying yes to the item.

 

 

 

 

30.00-
Lines show Mean
25.00-
% 20.00-
15.00-
10.0 i fi F l I
.oo .20 .40 .60 so
Facflny

Figure 4 Relationship between difficulty and facility

75

TESTFACT provides two classical indices, the point biserial (PBIS) and the

biserial correlations (BIS). Both are used for the calculation of correlation between

the score (1 or 0) on the item and the score on the survey as a whole. The higher the

correlation between these two scores, the more effective the item is in separating the

scores of the respondents. The biserial correlation has an advantage to the PBIS, as it

is a measure of association between response on the item and response on the survey.

The formula for calculating the BIS is (RMEAN — MEAN) / S.D. x (FACILITY/

h(FAClLITY)). The BIS is less influenced by item difficulty (DIFF) and tends to be

invariant from one testing situation to another.

Based on the discriminating power, the items with less than 0.1 (=10%

correct answers) of facility and less than 0.5 of BIS were dropped for convergence.

Only 30 items which met the two conditions were summarized in Table 7 with basic

item statistics. Items were labeled with the subjective classification of survey items by

content: “,”A Attendance, “P,” Participation, “B,” Buying, i.e., Purchase, “T,” Tourism,

“,”L Library, “,”D Donation, “,”E Education. This grouping was based on the

similarity of the way in which the survey developers intended and would be used later

to check whether it is the same as the way in which survey participants responded to

the items.

76

Table 7 Item statistics of 30 items

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

No Category Item No. N MEAN S.D. RMEAN FACILITY DIFF BIS
1 A 1 6110 9.33 5.61 13.53 0.374 14.29 0.74
2 A 2 6110 9.33 5.61 14.59 0.141 17.3 0.591
3 A 3 6110 9.33 5.61 12.8 0.442 13.58 0.693
4 A 5 6099 9.33 5.61 14.16 0.297 15.13 0.738
5 A 10 6084 9.31 5.6 14.47 0.102 18.07 0.528
6 A 11 6110 9.33 5.61 13.38 0.323 14.83 0.65
7 P 13 6110 9.33 5.61 13.71 0.131 17.48 0.482
8 P 23 6110 9.33 5.61 14.01 0.24 15.83 0.643
9 P 29 61 10 9.33 5.61 14.47 0.163 16.93 0.607
10 P 30 6110 9.33 5.61 14.8 0.141 17.3 0.615
11 P 33 6110 9.33 5.61 15.53 0.11 17.91 0.647
12 P 34 6110 9.33 5.61 14.36 0.12 17.7 0.538
13 B 49 6110 9.33 5.61 13.88 0.281 15.32 0.676
14 B 50 61 10 9.33 5.61 10.63 0.753 10.26 0.555
15 T 74 6110 9.33 5.61 11.51 0.623 11.75 0.637
16 T 75 6110 9.33 5.61 15.91 0.133 17.45 0.726
17 T 76 6110 9.33 5.61 14.78 0.247 15.73 0.761
18 T 77 6110 9.33 5.61 14.36 0.247 15.73 0.702
19 T 78 6110 9.33 5.61 13.75 0.322 14.85 0.707

20 T 79 61 10 9.33 5.61 14.1 0.225 16.02 0.637

21 T 80 6110 9.33 5.61 15.07 0.145 17.24 0.651

22 L 81 6110 9.33 5.61 10.6 0.687 11.05 0.439

23 L 82 6110 9.33 5.61 11.11 0.611 11.87 0.504

24 D 84 6110 9.33 5.61 15.13 0.146 17.22 0.659

25 E 86 61 10 9.33 5.61 1 1.05 0.69 11.02 0.6

26 E 87 6110 9.33 5.61 11.82 0.577 12.23 0.654

27 E 88 6110 9.33 5.61 14.09 0.19 16.51 0.595

28 E 89 6110 9.33 5.61 13.69 0.207 16.27 0.562

29 E 90 6110 9.33 5.61 12.1 0.46 13.4 0.571

30 E 92 6110 9.33 5.61 13.38 0.205 16.3 0.521

 

 

 

 

 

 

 

 

By obtaining the histogram plot, the overall distribution of the total score of

each respondent was shown in Figure 5. The data had a mean of 9.3 and a standard

deviation of 5.6. The proportion of correct responses, 0.311 was smaller than 0.5 of

the nominal probability of dichotomous variables. Thus, it can be inferred that the

questions were hard for the respondents to say yes. As the items were related to a
special product, i.e., art and culture, this result was predictable. The range of difficulty
enhanced this expectation.

The next step was to determine the number of dimensions needed to
accurately model the relationships of the item scores. Because of the sample-specific
nature of the MIRT methods, it is useful to determine how much the results generalize
to other samples of people or other forms of a survey constructed to the same
specification. This is particularly important when building a general form to measure

audiences’ activeness, or to find audiences’ behavioral patterns in a certain area.

FREQUENCY

**

*****

********

*********

*********

**********
***********
*************
*************
I *e*************
|****************
l*****************
l*******************
I*********************
I*********************
I***********************
It************************

0.0+ ------ + ------- + ------- +- + + +SCORES
0. 5. 10. 15. 20. 25. 30.

____..___..ii

5%

 

Figure 5 Histogram of 30 items

There is no correct answer to the right number of dimensions problem

(Reckcase, 2009). Schilling and Bock (2005) suggest using a chi-square test of the

78

difference in fit of models with m and m+l dimensions to determine the number of

coordinate axes needed to model the relationships in the item response matrix. Tate

(2003) found that these procedures are one of the evaluation methods of research on

model fit. The statistical results to test the null hypothesis that the additional

dimension gives no improvement in fit are provided in Table 8.

Table 8 Chi-square and degrees of freedom

1 2 3 4 5 6
dimension dimensions dimensions dimensions dimensions dimensions

Chi-square 64641.95 55693.5 52208.51 51205.99 45846.91 45836.56
DF 6049 6020 5992 5965 5939 59 14
p value (0.0001 (0.0001 <0.0001 <0.0001 <0.0001 <0.0001

 

 

 

 

 

 

 

 

 

*DF: Degrees of freedom = (the number of total responses — I) — [(dimensions+l) x the number of

total items — dimensions (dimensions-1V2]

The insignificant difference means that adding a dimension doesn’t improve
the model fit and suggests that the smaller of the two compared dimensions is
appropriate for analysis. From the output in Table 8, the chi-square for a l-factor
versus a 2-factor model is 64641.95 - 55693.5 = 8748.45 with 6049 — 6020 = 29
degrees of freedom. As this is highly significant, the null hypothesis that the l-factor
model is better than 2-factor model is rejected. With the repeated chi-square
difference test, the analysis of the real data suggestes that 4 or 5 dimensions will be
needed to model the relationships present in the item score matrix.

For back-up of this idea, NROOT=15 in the FACTOR command was

79

performed. An empirical rule for the selection of the number of factors, m, was to set
m equal to the number of latent roots larger than 1 (Bock et al., 2002). Table 9
presented that 4 or 5 factors was appropriate. The number of factors should be
decided on the basis of both theoretical framework conceming the items and statistical

results.

Table 9 NROOT largest latent roots of the correlation matrix
r|2|3|4j5|6|7l8
8.949588 | 2.462394 | r.47573r j 1.306745 | 1.032778 I 0.505416 l 0.381034 | 0.323848

 

In the path model, the items were theoretically developed by three big

categories; prior experience, behavioral attitude, and consumption, and 30 items

selected for analysis were divided into 7 specific categories by content which can be

changed into 5 categories by merging the categories with only one or two items. In

addition, in the 5-dimension model, item parameter estimates for the data did not

converge even in 700 cycles. Therefore, 4-dimensional space appeared to be the best

for analyzing the current data.

The output of the 4-dimensional model was summarized in Table 10. The

table contains the four a-parameters (m per item) for each item and the d-parameters

(one per item) as well as the multidimensional discrimination estimate A, the

multidimensional difficulty estimate B, and the angles of the direction of best

measurement, a, for each item with the coordinate axes for the space (Bock, Gibbons

80

& Muraki, 1988).

Table 10 Item parameters and descriptive statistics in the four dimensions

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

No Item d a. a2 a3 a. A B 011 012 013 014
No.

l 1 -0.991 1.028 0.132 0.741 0.723 1.46 0.68 46 85 59 61
2 2 - l .460 0.439 0.131 0.301 0.402 0.68 2.15 49 79 64 54
3 -3.007 4.930 -0.106 1.817 1.817 5.56 0.54 27 89 71 71
4 -1.382 1.021 0.086 0.751 0.687 1.44 0.96 45 87 59 61
5 10 -2 .082 0.849 -0.043 0.322 0.219 0.94 2.23 24 87 70 77
6 11 -4013 4.324 -0197 1.532 1.390 4.80 0.84 26 88 71 73
7 l3 -1 .196 0.259 0.228 0.093 0.133 0.38 3.14 47 53 76 70
8 23 -4.066 -2.029 11.829 1.041 4.832 12.98 0.31 81 24 85 68
9 29 -1.889 -0.474 2.240 0.295 0.948 2.50 0.76 80 26 83 68
10 30 -2.1 19 -0.443 2.175 0.292 0.963 2.44 0.87 80 27 83 66
11 33 -2.162 -0124 1.712 0.250 0.838 1.93 1.12 87 27 83 64
12 34 -1.244 0.120 0.326 0.171 0.214 0.44 2.81 74 42 67 61
13 49 -0.785 0.286 0.353 0.462 0.284 0.71 1.11 66 60 49 66
14 50 0.810 0.270 0.325 0.296 0.130 0.53 -1.52 59 52 56 76
15 74 -0.022 2.992 1.358 44.288 -6.546 44.89 0.00 86 88 8 81
16 75 -2.039 0.442 0.133 0.943 0.151 1.06 1.92 65 83 27 82
17 76 -2. 165 0.211 0.320 1.902 -0.045 1.94 1.12 84 81 ll 89
18 77 - l .561 0.190 0.220 1.247 -0.086 1.28 1.22 81 80 14 86
19 78 -1.618 0.133 0.276 1.818 -0.218 1.86 0.87 86 81 11 83
20 79 -1.679 0.068 0.102 1.250 -0.068 1.26 1.33 87 85 8 87
21 80 -1.841 0.156 0.133 0.958 0.014 0.98 1.88 81 82 ll 89
22 81 0.614 0.094 0.230 0.185 0.426 0.53 -1.17 80 64 70 36
23 82 0.361 0.134 0.256 0.263 0.447 0.59 -0.61 77 65 64 41
24 84 -1.440 0.318 0.222 0.503 0.156 0.65 2.20 61 70 40 76
25 86 0.720 0.453 0.524 0.208 0.012 0.72 .100 51 44 73 89
26 87 1 1.144 1 1.046 18.667 -2.605 -8.295 23.37 -0.48 62 37 84 70
27 88 -0.989 0.445 0.907 003.4 -0. 199 1 .03 0.96 65 28 88 79
28 89 -0.955 0.542 0.656 -0.027 -0.223 0.88 1.09 52 41 88 76
29 90 0.353 1.734 3.070 -0.361 -1.801 3.98 -009 64 40 85 63
30 92 -0.875 0.385 0.840 -0.103 -0.309 0.98 0.89 67 31 84 71

The a-parameters and d-parameters are the equivalent of slopes and intercept.

And as explained earlier, the minus d-parameter means harder items than the plus

values. As a result, more respondents answered yes to Items 50, 81, 82, 86, 87 and 90

than any other items. This indicates that more people participated in these activities,

buying books and CDs, having a library card, visiting a library, and having experience

with art and taking an art class in their childhood, especially a music class. The a-

81

parameters that are sensitive to differences based on dimensions are marked in bold
and italics. These values provide a clear sense of which variables are correlated and
which patterns are found in that relationship.

The difficulty parameter, A, indicates the value of 0 corresponding to the
point of steepest slope for the item characteristic curve (ICC). The discrimination
parameter, B, is related to the slope of the ICC where it is steepest. In MIRT case, the
slope of a surface is dependent on the direction of movement along the surface so that
the point of steepest slope depends on the direction. In this context, the corresponding

angles are specified by Eq. 7

(7)

 

This formula is closely related to A and B values as shown in Eq. 8 and Eq. 9.
High positive values of B explain difficult items, which are the opposite interpretation

to d-parameters, as the equation to calculate B has a minus sign in the nominator.

(8)

 

B can be described as -d / A. Thus the values of the two descriptive statistics

have an inverse relationship as shown in Table 10.

82

 

(9)

The variances explained by each dimension are 29.10634, 15.47853, 8.67663,

and 5.84328. Four dimensional spaces capture 60% of the population variance in total.

In Table 11, a communality is computed by squaring that variable's factor loadings

and then by summing these squares. A communality of 0.5 for a variable means that

50% of the variance for that variable is accounted for by all the retained factors. The

negative value of difficulty shows easier items. Item 70 has zero difficulty and 100%

communality. Overall communality exceeds 0.5.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 11 Standardized difficulty, communality and factor
No. Item Diffi- Commu- Principle Factor VARIMAX factor
No. culty nality l 2 3 4 l 2 3 4
1 I 0.558 0.682 -0.683 -0.265 0.277 -0.261 0.284 0.041 0.754 0.176
2 2 1.207 0.316 -0.470 -0.110 0.168 -0.235 0.152 0.029 0.502 0.202
3 3 0.532 0.969 -0.696 0.326 0.591 —0. 170 0.188 0.169 0.950 40.042
4 5 0.787 0.676 -0.673 -0.291 0.278 -0.246 0.296 0.031 0.752 0.147
5 10 1.520 0.467 -0.475 -0.248 0.421 -0.054 0.155 0.143 0.644 -0.086
6 11 0.819 0.958 —0.675 -0.339 0.608 -0.135 0.187 0.180 0.940 -0.085
7 13 1.118 0.127 -0.3 16 0.089 0.124 -0.061 0.055 0.181 0.254 0.164
8 23 0.312 0.994 0.544 0.709 -0.343 -0.279 0.055 0.354 —0.041 0.929
9 29 0.702 0.862 -0.509 0.634 -0.360 -0.266 0.085 0.298 -0.053 0.873
10 30 0.804 0.856 -0.5 12 0.626 -0.353 -0.278 0.083 0.289 -0.038 0.874
11 33 0.996 0.788 -0.554 0.577 -0.253 -0.289 0.074 0.303 0.070 0.828
12 34 1.138 0.163 -0.358 0.135 -0.038 -0.125 0.118 0.132 0.181 0.315
13 49 0.641 0.333 -0.565 0.0 I 5 ~0.038 -0.113 0.317 0.131 0.343 0.314
14 50 -0.715 0.220 -0.466 0.052 0.023 -0.017 0.230 0.211 0.267 0.226
15 74 0.000 1.000 -0.661 -0.533 -0.428 0.310 0.988 -0.0 | 0 0.151 -0.004
16 75 1.399 0.529 -0.642 -0.333 -0.073 0.044 0.592 0.056 0.407 0.097
17 76 0.992 0.790 -0.690 -0.381 -0.368 0.184 0.853 0.035 0.204 0.139
18 77 0.959 0.622 -0.618 40.333 -0.298 0.199 0.757 0.068 0.188 0.095
19 78 0.767 0.775 -0.642 -0.382 -0.389 0.258 0.864 0.054 0.135 0.084
20 79 1.045 0.613 -0.553 -0.388 -0.353 0.176 0.766 -0.023 0.142 0.074
21 80 1.315 0.490 -0.554 -0.321 -0.255 0.121 0.661 0.015 0.206 0.101
22 81 -0.544 0.217 -0.337 0.053 -0.048 -0.3 l 4 0.080 -0.044 0.264 0.373
23 82 -0.310 0.261 -0.404 0.027 -0.054 -0.306 0.138 -0.031 0.303 0.386
24 84 1.205 0.300 -0.533 -0. 125 -0.013 -0.013 0.375 0.122 0.339 0.171
25 86 -0.584 0.343 -0.520 0.200 0.152 0.098 0.166 0.435 0.289 0.209
26 87 ~0.476 0.998 -0.512 0.657 0.310 0.456 -0.007 0.978 0.096 0.182
27 88 0.689 0.515 -0.447 0.489 0.159 0.226 0.029 0.660 0.106 0.260
28 89 0.717 0.436 -0.415 0.361 0.266 0.251 0.024 0.621 0.190 0.115
29 90 -0.086 0.940 -0.458 0.613 0.266 0.533 0.035 0.963 0.022 0.110
30 92 0.625 0.490 —0.362 0.498 0.159 0.291 -0.001 0.672 0.031 0.191

 

 

 

 

 

 

 

 

 

 

 

The most important topic in MIRT will be how many dimensional spaces are
needed to accurately represent the differences in persons. In this study, four 0 vectors
are used as examples along with 30 test items about art-related consumption behavior
that are sensitive to differences in the persons. It also will be important to check the
correlations between factors to evaluate the possibility to merge or separate similar
factors. As shown in Table 12, reliability values are over 0.9.

Table 12 PROMAX rotated factor correlations

 

 

 

 

 

 

 

 

 

0. 02 03 04
9, 1.000 0.078 0.578 0.355
()2 0.078 1.000 0.249 0.442
93 0.578 0.249 1.000 0.424
94 0.355 0.442 0.424 1.000
Reliability 0.9475 0.9702 0.9658 0.8803

To confirm dimensional structure, a cluster analysis of a measure of the
similarity of the constructs is performed. A clustering of items is based on two
decisions: the selection of a measure of similarity between items and the algorithm for
forming clusters. For the similarity measure, one option is the angle between each pair
of item vectors (Miller & Hirsch, 1992). The clustering method that works well when
the angle between items is the similarity measure is Ward’s method, which recovers
the underlying structure of the data more accurately than alternative methods (Child,
2006). This type of analysis does not give conclusive information about the required

number of dimensions. When there is a major difference in the solution, it means that

84

a higher number of dimensions is needed (Reckcase, 2009). In addition, researchers

support that the analysis of angular responses improves the ability to characterize the

variables. Thus, the clustering analysis of the angular data is implemented (Fonseca &

Calder, 2007).

Table 13 Inter-item correlation matrixes

Initial correlation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

No 1 2 3 4 5 6 7 8 9 10 ll 12 13
1 l 48 51 65 71 58 71 80 81 81 77 76 65
2 0.67 1 60 55 74 70 69 80 82 83 78 78 68
3 0.63 0.5 1 53 74 70 73 79 80 81 76 75 68
4 0.42 0.57 0.6 1 72 55 74 82 84 83 78 78 65
5 0.33 0.27 0.27 0.31 1 47 74 80 81 84 83 80 76
6 0.53 0.35 0.34 0.57 0.68 1 80 81 81 84 78 78 70
7 0.33 0.36 0.3 0.27 0.27 0.18 1 76 81 81 74 77 79
8 0.18 0.18 0.19 0.14 0.17 0.15 0.25 l 76 76 75 64 66
9 0.15 0.14 0.18 0.11 0.16 0.16 0.16 0.24 l 36 41 67 69
10 0.15 0.13 0.15 0.13 0.1 0.11 0.16 0.24 0.81 1 43 67 65
l 1 0.23 0.21 0.25 0.20 0.13 0.21 0.27 0.26 0.75 0.73 1 67 67

12 0.24 0.20 0.26 0.21 0.18 0.21 0.23 0.44 0.39 0.39 0.39 1 71

13 0.43 0.37 0.38 0.43 0.24 0.35 0.19 0.40 0.36 0.42 0.39 0.32 l

Smoothed correlation (positive-definite)
Ite 1 2 3 4 5 6 7 8 9 10 ll 12 13
m
l 1.00 63 40 47 59 41 76 81 81 81 77 77 66
2 0.46 1 .00 60 63 71 61 80 79 80 80 77 80 73
3 0.77 0.50 l .00 40 48 16 74 89 89 89 84 78 67
4 0.68 0.45 0.77 1.00 59 41 76 83 83 83 78 77 66
5 0.52 0.33 0.67 0.52 1.00 48 80 87 87 87 88 83 75 .
6 0.76 0.49 0.96 0.76 0.67 1 .00 74 88 87 88 87 79 68
7 0.24 0.17 0.28 0.24 0.18 0.27 1.00 78 79 79 78 83 80
8 0.16 0.19 - 0.13 - - 0.21 1.00 23 23 28 70 70
0.01 0.05 0.04

9 0.15 0.17 - 0.12 - - 0.19 0.92 1.00 31 35 71 71
0.02 0.05 0.06

10 0.16 0.18 - 0.13 — - 0.19 0.92 0.86 1.00 35 71 71
0.01 0.05 0.04

l 1 0.23 0.22 0.10 0.21 0.03 0.06 0.21 0.88 0.82 0.82 1.00 71 70

12 0.23 0.18 0.20 0.22 0.13 0.19 0.13 0.34 0.32 0.32 0.32 1.00 77

13 0.41 0.29 0.39 0.40 0.26 0.38 0.18 0.34 0.32 0.33 0.35 0.22 1

 

 

 

 

 

 

 

85

 

 

 

 

 

 

 

The uncentered version of the Pearson correlation assumes that the mean of
the population is zero. This is equivalent to calculating the cosine of the angle. Thus
the correlation matrices are used to estimate the angles between vectors. The equation
below is similar to the one used for alphas in the descriptive statistics.

I

a1 “2

m m (10)
2011 2027
(=1 (=1

 

 

cos a'l2 =

The lower triangle indicates correlation, and the upper triangle contains the
cosine of the angle between two variables. 1f the cosine of 0° is 1, it means the item
vectors are pointing in the same direction, while 90° illustrates a zero correlation
between the underlying latent variables. Two different matrices are used. The top
table has negative latent roots, but the second one is smoothed and positive definite.
The angles between items that are best at measuring different constructs have much
larger values. For example, the 3rd and 8lh items, i.e., Items 3 and 27 have very high
degrees. These two items are better to be located in the different dimensions. Based
on the two correlation matrices, the cluster analyses dendrograms are shown in
Figures 6 and 7. The horizontal axes are a function of the sum of the squared distance

between the items in one cluster and those in another. The values on the axes cannot

86

be directly converted into an angle measure (Reckcase, 2009). The results show that
there is a clear clustering of the items in the test, and even if the clustering does not
exactly match well with the subjective classification of test items by content, it
provides a useful analysis for understanding the underlying patterns in persons.

The top panel is the output of the initial correlation matrix, and the bottom
panel is based on the smoothed positive definite matrix. The two dendrograms seem
to illustrate the different results in both the order of the items and the appropriate
number of dimensions. However, the items on the clusters are very similar, despite the
discrepancy of a few items, like Items 49 and 50.

The big picture of the classification shows that the two dendrograms have
four or five categories around 10 on the vertical axes. The number of groups with the
clustering analysis is the maximum number of the dimensions to be obtained in MIRT
(Reckase, 2009). Thus, these results support the MIRT analysis that four dimensions

are needed to determine the location of difference in persons.

87

C A S E O 5 10 15 20 25
Label Hum + ————————— + ————————— + --------- + ————————— + _________ +

30 10 -+ ————————————— +
33 11 -+ 1

88 27 -+—+ +--_+

92 30 -+ + ————————— + I 1

89 28 -+-+ l I I

90 29 —+ +_+ 1

49 13 --—+ ————— + | |

50 14 ---+ +---+ + _____________________________ +
23 8 -—-+ ————— + 1

34 12 ---+ l l

86 25 ---+---+ | l

87 26 ---+ +—+ I

13 7 ——————— + I

81 22 -+ ————————————————— +

 

77 18 -+-+

79 20 —+ + ————————————— +

80 21 ——-+ l
S 4 -+—+ |

75 16 -+ +-+ + __________________
1 1 -+ I l 1
3 3 -+—+ +---+ |
2 2 -+ I l 1

84 24 ————— + + ——————— +
10 5 —+-——+ |

11 6 -+ +——-+

74 15 ----- +

Figure 6 Dendrogram for clustering of 30 items (initial correlation)

88

C A S E 0 5 10 15 20

Label 11mm + ————————— + ————————— + ————————— + --------- + ————————— +

3 3 -+

11 6 -+

1 1 -+—+

5 4 -+ + ————————— +

10 5 -+ 1 l

2 2 ---+ 1
49 13 ---+—+ + ----- +
84 24 -——+ +—+ I I
so 14 ----- + 1 I l
81 22 --—+—+ + ————— + l
82 23 —--+ I l 1
34 12 ----- +-+ I

13 7 ----- + + ----------------------------- +

74 15 -+ 1
78 19 -+ |
76 17 -+ I
I
1

77 18 -+
79 20 —+
80 21 -+ ----------------- +

75 16 -+
29 9 —+
30 10 —+
23 8 —+ --------- +
33 11 -+ l
87 26 -+ I

90 29 —+ + ------------------------------------- +

88 27 -+ |
92 30 -+-+ l
89 28 -+ + ------- +
86 25 ——-+

Figure 7 Dendrogram for clustering of 30 items (smoothed correlation)

89

Chapter 5. Findings and Implications
This section summarizes the statistical findings in accordance with the
research questions/objectives which are made explicit in the introduction. It provides
explanation of how the results can be interpreted and used for the future development
of efficient messages and communications. In addition, it suggests ideas of future
studies which will make the study more beneficial to an and cultural organizations, art

educators, producers and law makers.

Results and Discussion

The statistical output supports and explains the research questions. The
objectives of the study corresponding to the three research questions are: 1) to identify
factors for measuring public consumption behavior in the art and cultural sectors; 2)
to determine associations between valid factors; and 3) to utilize the associations for
segmenting the public and for accounting for their behavioral difference in art-related
consumption.

The results support the general intuition related to art and cultural
consumption. For example, those who enjoy traveling are more likely to participate in
a. variety of other cultural activities, such as attendance at different performances and

even donation for cultural organizations. While the widely accepted thoughts are

90

developed from intuitive and empirical suggestions , this study provides scientific and
statistically proved instruments to segment people through the insight of how people
behave differently and, more importantly, which behaviors are more closely related
and influenced by each other.

When trying to persuade publics and legislators, organizations in the art and
cultural industry rely on their knowledge of what is conducive to the success of the
communication plan in accordance with their experience over a long period of time. It
is difficult, however, to regard the empirical generalizations on art and cultural
behaviors as scientific data to support their statements. Therefore, these findings will
provide them with a robust foundation for establishing more reasonable and
trustworthy messages appealing to both target audiences.

The results are categorized by the dimension which represents a construct of
items measuring behavioral patterns. The parameter estimates obtained by the MIRT
method indicate that the 30 items are classified in four different dimensions. The
items under one dimension are closely correlated and define the unique characteristics
about how different people are in their art and cultural consumption. Thus, the items
can be used to separate and cluster people based on the characteristics each dimension
explains. Multidimensionality means people are located in the space of their art and

cultural consumption behaviors according to their responses to the items under each

9l

dimension. In short, a person who answers “yes” to most of the items under
Dimension 1 has different behavioral patterns from a person who answers “no” to
most of the items in terms of the latent characteristic accounted for by Dimension 1.
In addition, a person who has a certain pattern in Dimension 1 can show similar or
different behavioral dynamics in dimensions 2 and 3. As a result, organizations can
apply the differences for implementing more target-oriented communication activities
with the analysis of what factors make people behave in a certain way and how the
difference in their behaviors can be factored into the development of efficient
messages.
Dimension 1: Activities with intangible output and temporal enjoyment

Dimension 1 includes attendance at various performances and an area of
participation in performing arts. Performing arts cover: 1) Playing a musical
instrument, 2) Playing in a band or orchestra, 3) Solo singing, 4) Group or choir
singing, 5) Dance, 6) Acting, 7) Comedy, magic and mime, 8) Storytelling and
readings, and 9) Other performances. This dimension seems to explain temporal
activities. Unfortunately, obtaining tangible outcomes right after implementing
performances is difficult when dealing with temporal activities. By playing musical
instruments and performing arts, people can enjoy the moment and achieve something

in return psychologically, while they cannot touch or see the output.

92

The following seven items are sensitive to the difference in persons on
Dimension 1: 1) Attendance at theatrical performance, 2) Attendance at theatrical
performance in a different state or country, 3) Purchase of tickets of theatrical
performance, 4) Attendance at dance performance, 5) Attendance at music
performance in a different state or country, 6) Purchase of tickets of music
performance, and 7) Participation in any performing arts. According to their responses
to the seven items, people appear to be different in their engagement in art and
cultural activities. Thus, attendees at performances demonstrate different
characteristics as art and cultural consumers, and they can be clustered as different
consumers from people who never attended any performances. As a result,
communicators should come up with different strategic plans based on where they can
find the right target population and what their target prefers doing.

Among the three different categories measuring attendance at performances,
such as theatrical, dance and music, people are more likely to attend a dance
performance with free tickets. Thus, unlike the other two categories, buying tickets
for a dance performance alone is not valid to measure the characteristics of
Dimension 1, as the majority of people do not demonstrate the difference. Simple
attendance appears to be enough for measuring the level of engagement in art and

cultural activities. Based on whether people attend performances, organizations

93

should treat people differently when developing marketing messages.

Purchase of tickets for attending theatrical and music performances has the

highest discriminating power among the items in this dimension. In terms of the

discriminating power, the degree to which people can be separated by each item is

calculated based on the relationships between difficulty and facility, which means the

proportion which gets the item “yes.” Difficulty and facility are negative-correlated.

This means that the more difficult saying “yes” to the item is, the smaller the

proportion of people saying “yes” to the item is. The reason the items with

discriminating power are important is that they considers the relationship between

peOple and items. For example, even if two students have the same scores in a specific

examination, their same scores do not indicate that they have uni-dimensionally the

equal ability or capability. One of the two students may have higher ability in a

specific field, while the other is not in that field. Therefore, people who say “yes” to

the item with higher discriminating power are regarded as very different from others

who say “no” to the same item.

In terms of difficulty, purchase of tickets for attending theatrical and music

performances is identified as comparatively easy questions. They rank first and third

respectively, which represents a high probability for people to say “yes” to those items.

By considering high discriminating power and high difficulty, these two questions are

94

useful for measuring the latent characteristics of Dimension 1. Depending on whether
people say “yes” or “no,” communicators can treat them as different consumers or
audiences.

As fewer people attend dance performances, and music performances in a
different state or country, people who say “yes” to either of those questions can be
categorized into heavy and active consumers in this dimension. Dimension 1 appears
to be a good barometer for determining whether people are more active consumers
and highly engaged in art and cultural sectors.

Compared to attendance at performances, participation in the performing arts
has a relatively low discriminating power. It can be interpreted that even if people are
different in answering this item, it is not enough to determine that people who are not
participating in performing arts are totally different from people who are engaged in
this discipline. The fact, however, that participation in the performing arts is
incorporated into this dimension provides a clear rationale for the connection between
the involvement in attendance at the art-related performances and the participation in
performing arts. The result demonstrates that people who attend theatrical, musical
and dance performances have a high probability to participate in activities related to
the performing arts and vice versa.

Overall, Dimension 1 accounts for how people behave different in temporal

95

activities which do not bring tangible output, such as drawing and photograph.

Playing a musical instrument, singing and dancing, and attendance at various

performances, which are identified as the most useful items to separate people, are

one-time engagements. One does not have to have long-term devotion or commitment

in implementing those activities. If people want to play, they can play. If people want

to see any kind of performance, they can buy tickets and go to the places to enjoy

 

performances. All of the activities in this dimension appear to be temporary and
relatively impromptu. Even if some of the activities such as mime, comedy, and acting
require some skills to execute, the implementation does not last long. Once people
perform them, it is hard to find the physical and tangible outcome. People enjoy and
appreciate the activities at that moment. And people are satisfied with mental and
emotional fulfillment.
Dimension 2: Activities with tangible output and long-term commitment

Dimension 2 contains twelve items as follows: 1) Participation in any visual
arts, 2) Engagement in Photography, 3) Crafts and 4) Writing, 5) Participation in any
art-related or cultural disciplines, 6) Purchase of any books, videotapes, DVDs, CDs,
records, tapes or music downloads, 7) Exposure to art in childhood, 8) Having art
education in childhood, 9) Taking any visual arts classes, IO) any performing arts

classes, 11) any music classes (vocal/instruments), and 12) any crafts classes in

96

childhood. Visual arts include drawing, painting, printmaking, sculpture, textile arts,

photography, crafts, folk or traditional art, media arts, and writing.

The results demonstrate that the three areas, photography, crafts, and writing

can be used as a powerful measurement of art and cultural consumption in Dimension

2. Three items indicate the common characteristics that bring out tangible output of

pouring time, effort, and finance. People can see, touch, and keep photos, drawings,

and books/writings as the results of their endeavors. These activities can be also

labeled as tactile art activities.

In addition, compared to attending various performances, taking art and

cultural classes included in this dimension requires longer commitment. The

precondition to register themselves to those classes will be self-motivation and

relatively long-term commitment. Through their early experience with art and cultural

education, people appear to develop confidence in devoting themselves to the art and

cultural classes when they are grown up.

Purchase of books, CDs, and DVDs, downloading of music, and exposure to

art in childhood indicate low difficulty in parameter estimates. This means that they

are not the best items to distinguish people in terms of their engagement level, as most

of the respondents purchase books, CDs and DVDs, and download music online. The

question about exposure to art in childhood is whether parents or other adults took a

97

person to arts or cultural performances or events when he/she was a child (up to 18
years old).

The fact that most people have experience with purchasing CDs, DVDs and
books, downloading music, and exposing themselves to art activities in childhood
implies that the two questions are good filtering questions to determine whether a
person is actually participating in any kinds of art consumption behaviors or not. It
will serve as the basic information for communication planners to recognize the
existence of the evidently different audience groups and to sort who will be their focal
target population.

The item with the highest discriminating power in this dimension is whether
people had ever taken any art-related classes in their childhood. This means that
people who took art-related classes in their early period of life demonstrate different
consumption behaviors in the art and culture industry. Visual arts, performing arts,
music, and crafts classes are identified as especially useful for segmenting people
based on their engagement level. People who took those classes within the last 12
months have experience with art education in their childhood. And they can be sorted
as a highly involved consumer in Dimension 2. Except for literary classes, all of the
other classes mentioned in the questionnaire demonstrate the power to be used when

segmenting audiences to develop strategic communication plans. They appear to

98

explain the unique characteristics in Dimension 2.

As many educators have been claiming the magnitude of art education for

children, the results support the widely known connection between early education

and future consumption and purchase behavior in a specific area. Once people

become exposed and familiar to art and culture, regardless of whether they are

engaged in classes voluntarily or not, they seem to have a more positive attitude

toward art and culture and are interested in participating in art and cultural activities

later.

The items incorporated into this dimension are the activities which people can

engage in by themselves. When compared to Dimension 1, these items would be

categorized into more self-motivated consumption behaviors. People do not need to

have companions or partners to participate in photography or writing. Their final

decision on whether they will participate in these activities does not seem to be

influenced by outside stimulus. Self-motivation cumulated through prior experience is

more likely to be an important factor in making a final decision of participation.

People will need more time and effort to decide to enroll in the art and cultural classes

in Dimension 2. They are able to obtain the tangible output of their taking classes

which does not disappear right away. Therefore, this dimensional space can be used to

measure a more long-term art and cultural commitment than Dimension l.

99

 

Dimension 3: Socio-psychic investment activities

Dimension 3 includes the items about cultural tourism, purchase of visual arts,
and donation. People are asked about specific activities made on their pleasure trips.
The six activities related to art and cultural consumption are: 1) Attendance at a
theatrical performance on a trip, 2) Visit to a museum: art, history, science, or special
topics museum, 3) Attendance at a festival: art, crafts, music, food, wine, ethnic,
antiques, or car show, 4) Visit to a historical site, attraction or community: heritage
parks, lighthouses, monuments, landmarks, archaeological sites, cultural centers or
historical communities, 5) Visit to a zoo, aquarium, botanical garden or arboretum,
and 6) Visit to a library: national, state, local community, or university/college.

Buying visual arts and giving donations to art organizations have a low
discriminating power. This means that communicators cannot segment people by
asking two questions in terms of their engagement in art and cultural consumption.
The difference in answering the two questions is not a reliable barometer for
distinguishing between people.

People who go to libraries and attend theatrical performances when traveling
are identified as very heavy cultural consumers. Most people visit cultural touristic
attractions in the following order: historical sites, museums, festivals, zoos, libraries,

and theatrical performances. People are more likely to visit historical sites and

 

museum on their trips than a library or a 200. Thus, a visit to a library and attendance
at a theatrical performance can be a measurement to identify the high level of
engagement in art and cultural areas.

One more item included in Dimension 3 is donation. It can be interpreted that
people who are interested in traveling to cultural tourist attractions are more likely to
donate money to art and cultural affiliations. It might be explained by the fact that
travelers have a higher possibility to be more affluent, more educated and more
flexible than people who do not travel. Traveling to cultural spots requires more
money and time compared to engagement in the items included in Dimensions 1 and 2.
People who engage in such activities can be regarded as more experienced consumers
in the art and cultural industry. First-time travelers will visit tourist sites which are
well known to the public, while repeating travelers will try to find more sophisticated
activities in terms of art and cultural consumption, like theatrical performances.

As a result, Dimension 3 explains socio-psychic investment in art and culture.
When it comes to the purposes for participating in the activities in this dimension,
such as traveling, purchase of visual art, and donation, people appear to invest their
money and effort to expand their horizons in art and cultural consumption. Traveling
psychologically enriches their lives. Interaction with local people and extension of

their knowledge related to history and art in the regions that they travel will be a great

l0!

 

asset for their lives. People feel a considerable amount of self-esteem and self-

satisfaction when purchasing visual arts. Visual arts are regarded as not only de’cor for

their physical environment, but psychological refreshment as well. Donation makes

people think of themselves as responsible and contributing member of a society. It

helps people fulfill ethical obligation. People receive something in return: 1) self-

esteem, 2) the opportunity to make a contribution to society, 3) recognition from peers

and the community, 4) a sense of belonging, and 5) ego gratification (Wilcox &

Cameron, 2009). As a result, the fact that people are involved in any of the activities

in Dimension 3 reflects that they are socially, economically, sentimentally, and

psychologically affluent and they are ready to invest their resources in art and cultural

behaviors.

One final cluster: screening questions

One more group is identified separately from the three dimensions. It explains

library usage and contains only two items: having a library card and visiting a library.

A visit to a library appears to measure a different characteristic from a visit to a

library during a cultural trip. Most people seem to have at least one library card and

visit libraries. This cluster accounts for explicitly different latent characteristics from

the other dimensions. It indicates the highest facility, which means the highest percent

of affirmative answers, and the highest discriminating power.

l02

High facility and discriminating power can be interpreted that the library

usage items are useful to filter people in art consumption, like the item of purchasing

CDs, DVDs and books, and downloading music. The basic level of engagement may

depend on the answers to these two questions. People who had never visited a library

and do not have any library card are unlikely to be considered as active consumers in

art and cultural sectors. At the beginning of developing a communication campaign,

these groups will be hard to be a focal segment for persuasion.

According to the correlations among dimensions, Dimension 1 is highly

related to Dimension 3, and Dimensions 1 and 2 are not much related. This indicates

that the contrasting dimensions explain the different latent characteristics. The items

included in these dimensions will aid in identifying people with a certain engagement

level. It suggests that people who attend art and cultural performances are more likely

to participate in art and cultural activities on their trips. Contrary to this, the

dissimilarity between Dimensions 1 and 2 represents that the two dimensions measure

totally different patterns of the respondents’ art and cultural consumption behaviors.

In addition, it is notable that art education in childhood has a great impact on

how actively people are engaged in art and cultural classes in the future, and people

who are interested in traveling to cultural attractions have more tendencies to donate.

This information will be useful for determining where, when, and how

103

 

communicators target the right group of people.

As a result, the associations between items are accounted for by audiences’
behaviors in art and cultural sectors, and these factors can be used for measuring the
unobserved characteristics, like a level of activeness or engagement, and for
clustering people to identify where communicators place their campaigns and how
they will reach the targeted audiences. Therefore, the coordinates to locate people in
the multidimensional spaces can be developed and implied for clustering population
as a unique segment when developing specific messages targeting a specific audience
group.

Each dimension also can be interpreted by the theories, TPB and SCT, which
were primarily used for developing the survey questions. Dimension 1 indicates the
relationship between attendance and participation in performing arts. This construct is
closely related to subjective norm in TPB, in which people care about what their
reference group thinks about their behaviors and how other people’s behavior affects
their behavioral change. In short, their behavior is influenced by others. People
usually attend art-related performances with someone else, unlike going to the movies
sometimes alone.

As a result, attendance at a performance is often a collectivistic behavior and

sometimes people go see a performance without their voluntary choice. It reflects that

people in this dimension take care of others’ interest, and others’ norm has an impact

on their decision-making. This finding is enhanced by the fact that participation in

performing arts is included in this dimension. Performing arts requires audiences,

which means that people engaged in performing arts need someone to work with and

to appreciate their performance. It has the distinctive difference from participation in

writing, craft, and photography, which are incorporated in Dimension 2.

In contrast, Dimension 2 demonstrates that people with earlier experience in

art and cultural activities have a tendency to participate in individualistic consumption

behavior, including taking pictures, writing, and making crafts. It appears to be

connected to self-efficacy in SCT. If people experience art and cultural activities in

their childhood and have good memories about the early exposure, they are more

likely to think that they have the ability to participate in the same activities when they

are grown up. Based on their experience, they can figure out whether they are capable

of dealing with the required skills and they feel self-satisfaction through those

activities.

Self-efficacy may be multi-faceted. Early exposure and early education

appear to be the prominent factors of creating self-efficacy in art and cultural

behaviors. They are likely to contribute to creating self confidence when participating

in art and cultural activities. Taking visual arts, music, and craft classes in childhood

105

 

results in engagement in photography, crafts, and writing. Self-motivated people may

actively participate in art and cultural classes with belief in their own capability of

performance.

Dimension 3 is related to Dimension 1. Dimension 3 can be explained by

PBC in TPB, as perceived behavioral control encompasses resources and technology.

The fact that the activities included in this dimension represent that people with a high

index in Dimension 3 can be assumed that they have resources to execute their

behaviors. Traveling to cultural attractions and attending cultural activities on their

trips requires more confidence in controlling their behaviors. They recognize the

meaningful role of the arts in their lives and they perceive that they can handle all of

the necessities to can'y out their plans and intentions.

Therefore the findings can be interpreted by the two theories, TPB and SCT,

even though the research was largely explanatory, not intended to test any of the

theories. Three concepts, subjective norm, self-efficacy and perceived behavioral

control, appear to be the powerful predictors of actual consumption behavior in art

and cultural sectors. These concepts also aid in understanding the implication of the

results.

How other people behave and think affects engagement in more interactive

activities. In contrast, what people know and experience has impact on more self—

l06

motivated behaviors. The results are supported not only by logical reasoning, but also
by theoretical explanation. The findings will be conducive to developing better
appealing and persuasive messages targeting the general public and law makers, as
they provide organizations with information on how people are engaged in different

activities differently and what affects and influences their behavioral differences.

Future Studies

The results of this study provide suggestions for future studies. The results
showed that early exposure to art and donation to art organizations play a significant
role in segmenting people in this area. Future studies, however, need to deal with the
relationship between these items and the other items which are clustered in the same
dimension. They should include the analyses of which items have impact on the other
items and which items work as moderator or mediator. The correlation is analyzed,
yet the path or route of influence should be determined to provide a better idea for art
and cultural organizations to persuade people toward the desired direction.

In addition, this study presents the simple measurements which can locate
people in the different dimensions of the unobserved characteristics. The research
includes other questions to be used for further relationship studies among the

variables or the constructs, such as age, zip code and sexual orientation. By relating

l07

 

these meaningful segment variables to the findings obtained in this study, more in-

depth and practical results will be explored and employed for better strategic

communication plans.

In terms of methodology, the information functions in MIRT are defined

independently of any specific group of respondents and represent the standard error of

measurement at any given behavioral level (Hambleton & Swaminathan, 1985). Once

a construct is identified with items, the measurement characteristics of the resulting

scale must be questioned: can the full range of vibration in the population of potential

respondents be measured with acceptable precision? Thus it should be examined

whether scores obtained at different occasions in time are stable and consistent by

replication.

Finally, as the MIRT analysis requires complex procedures, additional

graphical analyses may serve several functions. They will provide a visual perspective

that can triangulate or cross-validate traditional quantitative item, test, and differential

item functioning analyses (Ackerman, 1996). They help gain a better conceptual

understanding of the principles of measurement. This study lacks more graphical

explanation, like item response surface and item arrows or angles in multidimensional

space.

l08

 

Implications

The items included in each dimension are summarized in Table 14. The table

indicates which items are similar enough to define the results on each dimension. The

results suggest ideas for how communicators can identify the right audience and right

strategic communication methods including messages, places, and channels. In this

section, implication is developed based in two areas: 1) finding who will be the best

target population by using the result in each dimension and 2) developing

communication plans by combining dimensions and clustering a unique target

segment.

Table 14 Items to define the three dimensions

 

Dimension Items
Dimension 1 0 Attendance at theatrical performance
Activities with intangible output 0 Attendance at music concert
and temporal enjoyment 0 Attendance at dance performance

0 Participation in performing arts

 

 

Dimension 2 0 Art education in childhood
Activities with tangible output and 0 Early exposure to art
long-term commitment . Taking art-related classes
Dimension 3 0 Visit to historical sites, museum, festival,
Socio-psychic investment activities zoo/garden/aquarium, and library and attendance

at theatrical performances on trip
0 Purchase of visual arts
0 Donation

 

Grunig and Hunt ( I984) suggest that communication strategies should be

designed to attract the attention of two kinds of audiences: those who passively

process information (active audiences) and those who actively seek information

(passive audiences). Unlike two audiences, inactive audiences, who usually answer

“no” to the filtering questions, are difficult to be reached and persuaded by

communication campaigns in a short period of time. By considering the items with

high discriminating power, communicators can identify their target audience who is

regarded as highly involved and active consumers in art and cultural consumption

and/or who pay attention to a message only because it is related to self-interest.

Dimension 1 suggests that people who enjoy attending theatrical

performances and music concerts that are held in different states are the most active

audience. These people are already at the certain level of engagement in the art and

cultural industry and seek more sophisticated supplemental information. They will

actively seek more information, so marketers sort them and send them more diverse

and detailed information on performances. As people provide their personal

information, at least their address, when purchasing tickets online or offline,

marketers can identify them. For more passive audiences, like people who attend

various performances, communicators can make them aware of the messages through

brief encounters, such as advertisement on the way to any performances.

According to Dimension 2, children are the best target audiences for the

development of future engagement in the sectors. The finding that early education of

art-related disciplines leads to more active participation in those areas when people

are grown up can be appealing to the policy makers who need to have convincing

llO

evidence and rationale of why educational campaign is ultimately important for

vitalizing the art and cultural industry.

Dimension 3 indicates that people who attend theatrical performances on their

trips are sufficiently different to be segmented as an active audience. This finding is

identical to the findings shown in Dimension 1. To attend performances that are held

in different states, people need to stay in those areas over-night and sometimes plan a

trip that is longer than two days and one night to the perforrnance-hosting cities.

Dimension 3 even demonstrates that this segment is more likely to donate their money

to art and cultural organizations. Performing arts facilities or art and cultural touristic

attractions, such as museums and historical sites, will be best place to implement

fund-raising campaigns.

Further, communicators can develop strategic communication plans by

combining dimensions. With the three dimensions which explain different latent

characteristics of art and cultural consumers, eight different groups can be clustered

(Table 15). A message is more persuasive if environmental factors support the

message or if the message is received within the context of other messages and

situations with which people are familiar. These factors are called timing and context

(Wilcox & Cameron, 2009).

Group 1, who demonstrates active engagement in the three dimensions, can

lll

be reached and enhanced through a variety of communication campaigns targeting the

other six groups with the exception of Group 8, who is an inactive audience. Group 8

can be considered as long-term target audiences for enlightening and educational

communication campaigns, but will be hard to be regarded as target audiences for

marketing and strategic communication plans which expect comparatively immediate

 

 

 

results.
Table 15 Eight groups based on the three dimensions
Group Group Group Group Group Group Group Group

I 2 3 4 5 6 7 8

Dimension H H H H L L L L
l

Dimension H H L L H H L L
2

Dimension H L H L H L H L
3

 

 

 

 

 

 

 

 

* H: high involvement L: low involvement

People in Group 2 indicate low involvement in the items under Dimension 3.

They participate in art and cultural activities with tangible and intangible outcomes,

while they do not travel much. To appeal to this segment, communicators can develop

conjoint marketing strategies with tourism agencies or tourism bureaus of cities. One

suggestion is to select people among attendees at a specific performance and offer

them free coupons, which provide accommodation and transportation. As Dimension

1 is closely related to Dimension 3, people will be attracted to messages about

traveling to different states to attend performances and will be triggered to consider

“2

trips to other cities. If a theatrical performance is associated with a specific city or

state, like “Miss Saigon,” communicators of a specific city or even country’s tourism

bureau can invite an organizer or producer of a performance to a joint promotion or

vice versa. In addition, performing arts producers and organizers can advertise their

performances on pamphlets for each other, when they host performances in different

cities or areas. This will work for active information seekers. These strategies can be

applied for Groups 4 and 7.

In Group 2, another target population is high-involved people in Dimension 2.

They participate actively in photography, writing, and crafts. Thus, communicators

can develop programs such as a trip to the setting of a specific novel with the writer

and a road trip to take scenic photos with a famous photographer. Another idea would

be to host a workshop or small conference to share ideas and exhibit crafts in a place

related to crafts.

For Group 5, the results suggest that the better place and more proper target

public for marketing theatrical, music, and dance performances can be art and cultural

classes and registrants to those classes, especially performing arts classes, such as

acting, playing musical instrument, singing and dancing. Therefore, it might be

effective to create art classes at performing arts centers to make the audience more

loyal and accessible to the products and services they provide.

113

In addition, performing arts centers or organizers of performances can create

more target-specific marketing ideas, such as a group viewing discount for class

participants, and use the associations of private and public art class teachers to

disseminate the information on performances related to their class. Performing arts

facilities can provide places to exhibit their masterpieces or perform their skills for

people who are engaged in performing arts and take visual arts classes at a low price

or for free in case that small space is needed. The same strategy will also work for

Group 6.

For Group 3, art and cultural classes can also be marketed to attendants at

various performances. After attending orchestra concerts, people appear to think more

about taking musical instrument classes. If people are mesmerized by the skillful

gesture and movement of the dancers, their desire to learn dances will be heightened.

Even if not directly related to what they saw, people will have more interest in

learning something related to play, music, and dance. Marketers can provide discounts

on class fees for people who bring specific performance tickets they purchase. This

will be a very efficient communication plan at the point of purchase.

Based on dynamics between Dimensions 1 and 2, Dimensions 1 and 3, and

Dimensions 2 and 3, organizations can adopt one of the suggested communication

strategies or a combined strategy. No matter what kind of strategic plans organizations

ll4

and communicators develop, it will be more efficient and successful when they know

who the target audience is, who will pay more attention to their messages and how

and where they should deliver their messages.

Conclusion

The findings accomplish the objective of this study; to determine the number

of dimensions of items that are sufficiently similar to define the results on the

dimension. The results will be conducive to the development of more target-oriented

and efficient messages and a communication plan. Each dimension contains sets of

items that approximate the coherent characteristics of people as consumers in the art

and cultural sectors and potential target audiences for future communication messages.

Four dimensions are the number of orthogonal coordinate axes that are required to

model the locations of persons measured by the 30 items.

In the study of human behavior including psychology, sociology and market

research, tests and scales are instrumental in making behavioral phenomena accessible

to quantitative analysis (Bock et al., 2002). MIRT method can be used when the

instruments have multiple items which must be combined into a score for each person.

Responses of survey participants are interpreted as an ability, activeness, or

preference for their behaviors in a certain sector.

“5

 

This study explores how people act in art and cultural areas, demonstrates
that evident patterns and salient differences in cultural consumption exist, and
segments people in the multi-dimensional spaces. It contributes identifying how
people behave differently and determining which items are valid for segmenting
people based on the engagement levels. With interpreting and combining the results
suggested in Table IS, strategic communication planners can develop more efficient
and target-oriented communication campaigns and strategies.

There are the difficulties of establishing a coherent body of knowledge for
human behaviors in a specific area. The information on how people create and
demand a high degree of devotion and engagement in art and cultural activities will
aid communicators in predicting their behaviors and identifying where they can be
found and seen and to what kind of messages they will pay more attention. The
predictive scenario can be developed from the analytical measurement. The results of
segmenting population-with the three different dimensions can be utilized as the basic
practical foundation for strategic communications.

The results found in this study can be used by any parties involved in art and
cultural sectors: performing art organizations, non-profit art associations, festival
organizers, schools, educators, galleries, museums, libraries, historic preservation

agents and policy makers/govemment. By combining the simple results with other

116

 

category variables, such as gender, age, education, income and residential area, they

can enhance the value and application of the findings. The implication of the results

will vary depending on who the communicators are, what their focal interests are and

who the target audiences are.

Attention to creative activities as a source of urban revitalization and

economic diversification has increased as traditional manufacturing cities in the north-

eastern US face industrial restructuring (Poon & Lai, 2008). As mentioned above, the

main locomotive of art education in childhood and donation can be utilized for

incubating more engaged cultural consumers. Appealing and persuasive messages

targeting a specific group of people will boost more participation in this cultural

sector and play a role to enlighten people and policy makers. How to describe and

interpret the content of behaviors will further provide communications with a chance

to know what changes have been made in people in the course of time.

As a result, characteristics associated with the potential audience, such as the

audiences’ attitude towards and experience with specific activities can be used in the

design and execution of the communication campaigns. And when decision makers

need a database to help them choose the best of several available alternatives or to

rationalize their final decisions, these results can be properly utilized.

Good research and measurement provide better information which leads to

”7

better decisions. Well-developed measurements will help people predict the

audiences’ willingness and future behavioral patterns in decision making of art and

cultural consumption.

118

APPENDIX
SURVEY QUESTIONS

Theatrical Performances

1) Have you attended any live theater performances (e.g., plays, musicals or operas)
during the last 12 months?

2) Were any of these live theater performances in a different state or country other
than where your permanent residence is located?

3) Have you purchased tickets (not including subscriptions) to live theater
performances during the last 12 months?

4) Have you purchased any subscriptions to live theater performances during the last
12 months?

Dance Performances

5) Have you attended any live dance performances (e.g., ballet, jazz or modern)
during the last 12 months?

6) Were any of these live dance performances in a different state or country than
where your permanent residence is located?

7) Have you purchased tickets (not including subscriptions) to live dance
performances during the last 12 months?

8) Have you purchased any subscriptions to live dance performances during the last
12 months?

Music Concerts

9) Have you attended any live music concerts during the last 12 months?

10) Were any of these live music concerts in a different state or county than where
your permanent residence is located?

11) Have you purchased tickets to live music concerts during the last 12 months?

12) Have you purchased subscriptions to live music concerts during the last 12
months?

Participation in Performing Arts

13) During the last 3 years, have you been engaged in any performing arts, such as
theater, dance, or music?

Which of these performing arts have you been engaged in during the last 3 years?

14) Playing a musical instrument

15) Playing in band or orchestra

119

16) Solo singing

17) Group or choir singing
18) Dance

19) Acting

20) Comedy, magic and mime
21) Storytelling and readings
22) Other performances

Participation in Visual Arts

23) During the last 3 years, have you been engaged in any visual arts, such as drawing,
painting, printmaking, writing, textiles, photography, and/or sculpture?
Which of these visual arts have you been engaged in during the last 3 years?
24) Drawing

25) Painting

26) Printmaking

27) Sculpture

28) Textile arts

29) Photography

30) Crafts

31) Folk or Traditional art

32) Media arts

33) Writing

Participation in Art-related or Cultural Disciplines

34) During the last 3 years, have you engaged in any art-related or cultural
disciplines; architecture; landscape architecture; fashion, graphic, interior or industrial
design; historic preservation or restoration; genealogy; or archaeology?

Which of these art-related or cultural disciplines have you engaged in during the last 3
years?

35) Architecture

36) Landscape architecture

37) Fashion design

38) Graphic design

39) Interior design

40) Industrial design

41) Historic preservation

42) Genealogy

43) Archaeology

120

Professional Artists

44) Do you consider yourself to be a professional artist?

45) Have you received formal education/training in your artistic discipline?

46) During the last 12 months, have you been engaged in any art-related professional
development activities?

47) Are you currently a member of an arts-related guild, union or
association/organization ?

48) Are you able to support yourself/family financially entirely from the revenues that

you earn from your artistic activities?

Purchases of Art

49) Have you purchased any visual arts including drawings, paintings, writings, prints,
graphic designs, crafts or photographs during the last 12 months ?

50) Have you purchased or rented any books, videotapes, DVDs, CDs, records, tapes,
or music downloads during the last 12 months ?

Employment of Performing Artists

51) Have you hired any performing artists including musicians/bands, actors,
comedians, mimes/performance artists during the last 12 months?

52) Have you hired any solo musicians during the last 12 months?

53) Have you hired any musical bands/groups during the last 12 months?

54) Have you hired any solo singers during the last 12 months?

55) Have you hired any singing groups during the last 12 months?

56) Have you hired any actors or theater companies during the last 12 months?

57) Have you hired any comedians, magicians, or mime artists during the last 12
months?

58) Have you hired any performance artists during the last 12 months?

59) Have you hired any dancers or dance companies during the last 12 months?

60) Have you hired any DJs during the last 12 months?

Architect & Design Services

61) Have you hired any architects or designers during the last 12 months?
62) Have you hired any architects during the last 12 months?

63) Have you hired any landscape architects during the last 12 months?
64) Have you hired any graphic designers during the last 12 months?

65) Have you hired any interior designers during the last 12 months?

66) Have you hired any fashion designers during the last 12 months?

67) Have you hired any industrial designers during the last 12 months?

121

Heritage-related Services

68) Have you hired any historic preservation/restoration, archaeological, or
genealogical services during the last 12 months?

69) Have you hired any historic preservation services during the last 12 months?

70) Have you hired any archaeological services during the last 12 months?

71) Have you hired any genealogical services during the last 12 months?

72) Have you hired any home/building restoration services during the last 12 months?

Cultural Tourism

73) Have you ever made any pleasure trips to, or in, Michigan?

74) Have you taken any pleasure trips to destinations more than 50 miles from your
home during the last 12 months?

Have you taken any pleasure trips to destinations more than 50 miles from your home
during the last 12 months on which you participated in any-of the following arts or
cultural activities?

75) Attended a theatrical performance (e.g., musical, play, or opera), dance
performance, or concert

76) Visited a museum: art, history, science, or special topics museum

77) Attended a festival: art, crafts, music, food, wine, ethnic, antiques, or car show
Have you taken any pleasure trips to destinations more than 50 miles from your home
during the last 12 months on which you participated in any of the following arts or
cultural activities?

78) Visited an historical site, attraction or community: heritage parks, lighthouses,
monuments, landmarks, archaeological sites, cultural centers or historical
communities

79) Visited a zoo, aquarium, botanical garden or arboretum

80) Visited a library: national, state, local community, or university/college

Use of Library
81) Do you currently have a valid library card?
82) Have you visited a library during the last 12 months?

Cultural Membership, Donation and Volunteering

83) Are you affiliated (e.g., member, officer, and employee) with a national, state or
local membership-based organization that supports, sponsors, or implements arts,
heritage or cultural activities. This could include artist organizations that fundraise
and sponsor cultural activities and facilities (e.g., concerts, art shows, heritage

preservation, and community libraries)?

122

84) During the last 12 months, have you donated any money to an arts, heritage or
cultural organization (or cause) including but not limited to organizations that you
were/are a member ?

85) During the last 12 months, have you volunteered any time to an arts, heritage or
cultural organization (or cause) including but not limited to organizations that you

were/are a member ?

Art-related Experience in Your Childhood

86) When you were a child (up to 18 yrs old) how often did your parents or other
adults take you to arts or cultural performances or events ?

87) When you were a child did you take any type of visual (e.g., painting and
drawing), performing (e.g., dance, singing, and musical instruments) or literary arts or
crafts lessons or classes?

What type of classes did you take when you were a child?

88) Visual arts

89) Performing arts

90) Music (vocal/instruments)

91) Literary arts

92) Crafts

Your Art Lessons and Classes

93) During the last 12 months have you taken any type of visual (e.g., painting and
drawing), performing (e.g., dance, singing, and musical instruments) or literary arts or
crafts classes or lessons ?

What type of classes did you take?

94) Visual arts

95) History/Appraisal

96) Performing arts

97) Music (vocal/instrumental)

98) Literary arts

99) Applied arts (e. g., architecture, design, genealogy and archaeology)

Art Lessons and Classes of Your Children

100) During the last 12 months, have any children residing in your household taken
any type of visual (e.g., painting or drawing), performing (e.g., dance, singing, or
musical instrument), or literary arts or crafts classes or lessons?

What types of classes did children living in your household take during the last 12
months? Check all that apply.

123

101) Visual arts

102) History/Appraisal

103) Performing arts

104) Music (vocal/instruments)

105) Literary arts

106) Applied arts (e.g., architecture, design, and genealogy and archaeology)
107) Crafts

124

REFERENCES

Abraham, C. & Sheeran, P. (2003). Implications of goal theories for the theories of
reasoned action and planned behavior. Journal Current Psychology, 22(3),
264-280.

Ackerman, T. (1996). Graphical representation of multidimensional item response
theory analyses. Applied Psychological Measurement, 20(4), 311-329.

Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl,
& J. Beskmann (Eds.), Action control: From cognition to behavior (pp. 11—39).
New York: Springer.

 

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and
Human Decision Process, 50(2), 179-211.

Ajzen, I. (2001). Nature and operation of attitudes. Annual Review of Psychology, 52,
27-58.

Ajzen, 1., & Fishbein, M. (1980). Understanding attitudes and predicting social
behavior. Englewood Cliffs, NJ: Prentice-Hall.

Ajzen, I., & Fishbein, M. (2000). Attitudes and the attitude-behavior relation:
Reasoned and automatic processes. In W. Stroebe & M. Hewstone (Eds.),
European Review of Social Psychology (pp. 1-33). John Wiley & Sons.

Andrease, A. R. (1995). Marketing social change. San Francisco: Jossey-Bass.

Armitage, C. J., & Conner, M. (1999). Distinguishing perceptions of control from
self-efficacy: predicting consumption of a low-fat diet using the theory of
planned behavior. Journal of Applied Social Psychology, 29(1), 72-90.

Austin, E. W., & Pinkleton, B. E. (2001). Strategic public relations management:
Planning and managing effective communication programs. Mahwah, New

Jersey: Lawrence Erlbaum Associates, Inc.

Bachmann, D. P., Elfrink, J., & Vazzana, G (1999). E-mail and snail-mail face off in
rematch. Marketing Research, 11(4), 10-15.

125

Bagozzi, R.P. & Brunkrant, RE. (1979). Attitude Organization and the Attitude-
Behavior Relationship. Journal of Personality and Social Psychology, 37,
913-929.

Balch, G, & Sutton, SM. (1997). Keep me posted: A plea for practical evaluation. In
M.E. Goldberg, M. Fishbein, & S.E. Middlestadt (Eds.), Social marketing:
Theoretical and practical perspectives (pp. 61—74). Mahwah, NJ: Lawrence

Erlbaum Associates.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive
theory. Englewood Cliffs: Prentice Hall.

Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: Freeman.

Béguin, A. A., & Glas, C. A. W. (2001). MCMC estimation and some model-fit
analysis of multidimensional IRT models. Psychometrika, 66, 541-561.

Bennett, P. D. (1995). Dictionary of marketing terms (2nd ed.). Chicago, IL:
- American Marketing Association.

Berger, I.E., & Andrew, A. M.(1989). The effect of advertising on attitude
accessibility, attitude confidence, and the attitude-behavior relationship.
Journal of Consumer Research, 16, 269-279.

Berkowitz , L. (1982). Advances in experimental social psychology. New York, NY:

Academic Press.

Billig, M. (1991). Talking of the royal family. London: Routledge.

Bimbaum , A. (1968), Some Latent Trait Models and Their Use in Inferring an
Examinee’s Ability. In F.M Lord and MR. Novick (Ed.), Statistical Theories
of Mental Test Scores, (pp. 397-472), Reading, MA: Addison-Wesley.

Blumberg, S. J., Davis, K. E., Khare, M, & Martinez, M. E. (2005). The effect of
survey folow—up on nonresponse bias: 2002-2003 joint Canada/United States
survey of health. Paper presented at the Annual meeting of the Statistical
Society of Canada, Saskatoon.

126

Bock, R. D., & Aitkin, M. (1981). Marginal maximum likelihood estimation of item
parameters: Application of an EM algorithm. Psychometrika, 46, 443-459.

Bock, R. D., Gibbons, R., & Muraki, E. (1988). Full information item factor analysis.
Applied Psychological Measurement, 12, 261-280.

Bock, R. D., Gibbons, R., Schilling, S. G, Muraki, E., Wilson, D. T., & Wood, R.
(2002). TESTFACT. In M du Toit (Ed) IRT from SS]: BlLOG-MG M ULTILOG.
PARSCALE, TESTFAC T (pp. 410-502). Lincolnwood, IL: Scientific Software

International.

Bolt, D. M., & Lall, V. F. (2003). Estimation of compensatory and noncompensatory
multidimensional item response models using Markov chain Monte Carlo.
Applied Psychological Measurement, 2 7, 395-414.

Brigham, J., & Giesbrecht, L. (1976). All in the family: Racial attitudes. Journal of
Communication, 26(3). 69-74.

Campbell, D. T. (1963). Social attitudes and other acquired behaviroral dispositions.
In S. Koch (Ed.), Psychology: A study of a science (Vol. 6). New York, NY:
McGraw-Hill.

Child, D. (2006). The essentials of factor analysis (3rd ed.). New York, NY:
Continuum.

Christoffersson, A. (1975). Factor analysis of dichotomized variables. Psychometrika,
40, 5-32.

Cobanoglu, C., Warde, B., & Moreo, P. J. (2001). A comparison of mail, fax and Web-
based survey methods. International Journal of Market Research, 43(Quarter
4), 441-452.

Coffman, J. (2002, May). Public communication campaign evaluation: An
environmental scan of challenges, criticisms. practice, and opportunities.

Cambridge, MA: Harvard Family Research Project.

Compeau, D., Higgins, C. A., & Huff, S. (1999). Social cognitive thebry and
individual reactions to computing technology: A longitudinal study. M15

127

Quarterly, 23(2), 145-158.

Couper, M. P. (2000). Web surveys: a review of issues and approaches. Public
Opinion Quarterly, 64(4), 464-494.

Couper, M. P., Traugott, M. W., & Lamias, M. J. (2001). Web survey design and
administration. Public Opinion Quarterly, 65(2), 230-253.

Crawford, S. D., Couper, M. P., & Lamias, M. J. . (2001). Web surveys: Perception of
burden. Social Science Computer Review, 19, 146-162.

Curran, .I. (1990). The new revisionism in mass communication research: A

reappraisal. European Journal of Communication, 5, 135-164.

Davey, T. C., & Oshima, T. C. (1994). Linking multidimentional item calibrations.
Paper presented at the annual meeting of the National Council on
Measurement in Education, New Orleans.

DeMars, C. E. (2005, April). Guessing Parameter Estimates for Multidimensional IRT
Models. Paper presented at the annual meeting of the American Educational

Research Association, Montreal.

de Vries, H., Dijkstra, M., & Kuhlman, P. (1988). Self-efficacy: the third factor
besides attitude and subjective norm as a predictor of behavioural intentions
Health Education Research, 3(3), 273-282.

Dzewaltowski, D. A., Noble, 1. M., & Shaw, J. M. (1990). Physical activity
participation: Social cognitive theory versus the theories of reasoned action
and planned behavior. Journal of Sport and Exercise Psychology, 12, 388-405.

Dillman, D. A., & Bowker, D. K. (2001). The Web questionnaire challenge to survey
methodologists. In M. B. U. D. Reips (Ed.), Dimensions of lntemet science
(pp. 159-178). Lengerich: Pabst Science Publishers.

Dillon, W. R. & Ajith Kumar. (1985). Attitude Organization and the Attitude-Behavior

Relation: A Critique of Bagozzi and Bumkrant's Reanalysis of Fishbein and
Ajzen, Journal of Personality and Social Psychology, 49 (1), 33-46.

128

DiMaggio, P., & Useem, M. (1979). Decentralized applied research: Factors affecting
the use of audience research by arts organizations. The Journal of Applied
Behavioral Science, 15, 79-94.

Ditsa, G (2002). Executive information systems use in organizational contexts:
Uncerstanding the user behavior. Paper presented at the International

Association of Computer Information Systems Conference.

Doney, P. M., & Cannon, J. P. (1997). An examination of the nature of trust in buyer-
seller relationships. The Journal of Marketing, 61 (2), 35-51.

Dorfman, L., Ervice J ., & Woodruff, K. (2002). Voices for change: A taxonomy of
public communications campaigns and their evaluation challenges. Berkeley,
CAzBerkeley Media Studies Group.

Duffy, M. (2000). Web-based research: An innovative method for nursing research.
Canadian Oncology Nursing Journal, 10(2), 45-49.

Duncan, T., & Moriarty, S. E. (1998). A communication-based marketing model for
managing relationships. Journal of Marketing, 62 (2), 1-13. 1

Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychologists.
Mahwah, NJ: Lawrence Erlbaum Association.

Farmer, T. (1998). Understanding this think we call lntemet research. Retrieved
March 23, 2008 from http://infotekonline.com/Info_Online_New/White_
paper_on_intemet.htm.

Fazio, R. H., & Roskos-Ewoldsen, D. R. (1994). Acting as we feel: When and how
attitudes guide behavior. In S. Shavitt & T. C. Brock (Eds.), Persuasion:
Psychological insights and perspectives (pp. 71-93). Needham Heights, MA:
Allyn & Bacon.

Felkins, L. (2001). An introduction to the theory of social dilemmas. Retrieved May
10, 2008 from httpz/lperspicuity.net/sd/sd-l.html

Fishbein, M., Triandis, H.C., Kanfer, F. H., Becker, M., Middlestadt, S. E., & Eichler,
A. (2001). Factors influencing behavior and behavior change. In T. R. A.

129

 

Baum, & J. Singer (Ed.), Handbook of health psychology. Mahwah, NJ:

Lawrence Erlbaum Association.

Fishbein, M., & Ajzen, I. (1975). Belief attitude, intention, and behavior: An
introduction to theory and research. Boston, MA: Addison-Wesley.

Fonseca, L., & Calder, B. (2007, May). Clustering Acoustic Backseatter in the
Angular Response Space. Paper presented at the US Hydrographic Conference,
Norfolk, VA.

Fraser, C. (1998). NOHARM: A Fortran program for fitting unidimensional and
multidimensional normal ogive models in latent trait theory. Armidale,

Australia: The University of New England, Center for Behavioral Studies.

Gosz, J. K., & Walker, C. M. (2002). An empirical comparison of multidimensional
item response theory data using TESTFACT and NOHARM. Paper presented at
the annual meeting of the National Council for Measurement in Education,

New Orleans.

Granello, D. H., & Wheaton, J. E. (2004). Online data collection: strategies for
research. Journal of Counseling and Development, 82(4), 387-393.

Grunig, J ., & Hunt, T. (1984). Managing public relations. New York: Holt, Rinehart
& Winston.

Grunig, J. E., & Repper, F. C. (1992). Strategic management, publics, and issues. In
Gruni g (Ed.), Excellence in public relations and communication management:
Contributions to effective organizations (pp. 117-157). Hillsdale, NJ:
Lawrence Erlbaum Associates.

Hall, P. (2000). Creative cities and economic development, Urban Studies, 3 7, 639—
649.

Hambleton, R. K., & Swaminathan, H. (1985). Item response theory: Principles and
applications. Bonston, MAzKluwer Academic Publishers

Harman, H. H. (1976). Modern factor analysis (3rd ed.). Chicago, IL: University of
Chicago Press.

130

Hausenblas, H.A., Carron, A.V., & Mack, D. E. (1997). Application of the theories of
reasoned action and planned behavior to exercise behavior: A meta-analysis.
Journal of Sport and Exercise Psychology, 19(1), 36-5 1 .

Henry, G. T., & Rivera, M. (1998, October). Public information campaigns and
changing behaviors. Paper presented at the meeting of the Association for

Public Policy Analysis and Management, New York.

Horst, P. (1965). Factor analysis of data matrices. New York, NY: Holt, Rinehart and
Winston.

Houston, F. S. (1986). The marketing concept: what it is and what it is not. Journal of
Marketing, 50(2), 81-87.

Integrated communication plan. (2002). Retrieved March 20, 2008, from
httpzl/www.mediacampaignorg/publicatm/integrjlan/intro-3.html

Kaplowitz, M. D., Hadlock, T. D., & Levine, R. (2004). A comparison of Web and
mail survey response rates. Public Opinion Quarterly, 68(1), 94-101.

Katz, E., & Liebes, T. ( 1985). The export of meaning: Cross cultural readings of
American TV. Paper presented at the Manchester Symposium on Broadcasting.

Kiesler, S., & Sproull, L. S. (1986). Response effects in the electronic survey. Public
Opinion Quarterly, 50(3), 402-413.

Kotler, P. (1980). Principles of marketing. Englewood Cliffs, NJ: Prentice-Hall.

Kotler, P., & Zaltman, G. (1971). Social Marketing: An approach to planned social
change. Journal of Marketing, 35 (July), 3-12.

Lam, T. & Hsu, C. H. C. (2004). Theory of planned behavior: potential travelers from
China. Journal of Hospitality & Tourism Research, 28, 463-482.

Lam, T. & Hsu, C. H. C. (2006). Predicting behavioral intention of choosing a travel
destination. Tourism Management, 27, 589-599.

LaPiere, R. T. (1934). Attitudes vs. actions. Social Forces, I 3, 230-237.

131

Lazar, J., & Preece, I. (1999). Designing and implementing Web-based surveys.
Journal of Computer Information Systems, 39(4), 63-67.

Livingstone, S. (1993). The rise and fall of audience research: An old story with a new

ending. Journal of Communication, 43(4), 5-12.

Lohr, S. L. (1999). Sampling:design and analysis. Pacitif Grove, CA: Brooks/Cole
Publishing Company.

Lord, F. M. (1980). Applications of item response theory to practical testing problems.
Hillsdale, NJ: Lawrence Erlbaum Associates.

Maestro, R. M. H., Gallego, P. A. M., & Requejo, L. S. (2007). The moderating role
of familiarity in rural tourism in Spain. Tourism Management, 28 (3): 951—964.

Markusen, A. & Schrock, G (2006). The artistic dividend: urban artistic specialisation
and economic development implications, Urban Studies, 43, 1661—1686.

Maydeu-Olivares, A. (2001). Multidimensional item response theory modeling of
binary data: large sample properties of NOHARM estimates. Journal of
Educational and Behavioral Statistics, 26, 51-71.

McDonald, R. P. (1967). Nonlinear factor analysis. Psychometric Monographs, 15(1-
167).

McDonald, R. P. (1999). Test theory: a unified treatment. Mahwah, NJ: Lawrence

Erlbaum Associates.

McElreath, M., & Blamphin, J. (1994). Partial answers to priority research questions-
and gaps-found in PRSA's body of knowledge. Journal of Public Relations
Research, 6, 69-104.

McGuire, W. J. (1989). A perspectivist approach to the strategic planning of
programmatic scientific research. In B. Gholson, A. Houts, R. Neimeyer, & W.
R. Shadish (Eds.), The Psychology of Science: Contributions to Metascience
(pp. 214-245). New York: Cambridge Univ. Press.

Meyer, T. P. (1976). The impact of all in the family on children. Journal of

132

Broadcasting, 20(1), 23-33.

Miller, T. R. (1991). Empirical estimation of standard errors of compensatory MIRT
model parameters obtained from the NOHARM estimation program (Research
Report ONR 91-2). Iowa City, IA: American College Testing Program.

Miller, T. R. & Hirsch, T. M. (1992). Cluster analysis of angular data in applications
of multidimensional item-response theory. Applied Measurement in Education,
5(3), 193-211.

Morales, L. S., Reise, S. P., & Hays, R. D. (2000). Evaluating the equivalence of
health care ratings by Whites and Hispanics. Medical Care, 38, 517-527.

Morley, D. (1980). The 'Nationwide" audience. London: British Film Institute.

Muthen, B. (1978). Contributions to factor analysis of dichotomous variables.
Psychometrika, 43, 551-560.

Nicholas, E., & Sedivi, B. (1998). Economic data collection via the Web: A census
bureau case study. Paper presented at the Survey Methods Section, American
Statistical Association Meetings, Dallas, Texas.

Outreach campaign planning - target audience. Retrieved March 10, 2008, from

http://www.wipo.int/imutreach/en/guides/planningltarget audiencehtml

 

Parker, L. (1992). Collecting data in the email way. Training and Development, 46(7),
52.

Parraga, I.M. (1990). Determinants of Food Consumption. Journal of American
Dietetic Association, 90, 661-663.

Peter, J. P., & Olson, J. C. (2005). Consumer behavior & marketing strategy (7th ed.).
New York, NY: McGraw-Hill/Irwin.

Poon, J.P.H. & Lai, C.A. (2008).Why are non-profit performing arts organizations
successful in mid-sized US cities?. Urban Studies, 45(11), 2273-2289.

Povey, R., Conner, M., Sparks, R, James, R., & Shepherd, R. (2000). Application of

133

 

the Theory of Planned Behaviour to two dietary behaviours: Roles of
perceived control and self-efficacy British Journal of Health Psychology, 5(2),
121-139.

Quellette, J. A.,&Wood,W. (1998). Habit and intention in everyday life: The multiple
processes by which past behavior predicts future behavior. Psychological
Bulletin, 124(1), 54-74.

Radway, J. (1984). Reading the romance: Women, patriarchy and popular literature.
Chapel Hill: University of North Carolina Press.

Rasch, G. (1960). Probabilistic models for some intelligence and attainment tests.

Danmarks Paedagogiske Institut, Copenhagen.

Reckase, M. D. (1985). The difficulty of test items that measure more than on ability.
Applied Psychological Measurement, 9, 401-412.

Reckase, M. D. (1997). The past and future of multidimensional item response theory.
Applied Psychological Measurement, 21(1), 25-36.

Reckase, M. D. (2009). Multidimensional item response theory. New York, NY:
Springer.

Reckase, M. D., & McKinley, R. L.‘. (1991). The discriminating power of items that

measure more than one dimension. Applied Psychological Measurement, 15,
361-373.

Rice, J. A. (2007). Mathematical statistics and data analysis. Belmont, CA: Thomson
Higher Education.

Samejima, F. (1974). Normal ogive model on the continuous response level in the
multidimensional space. Psychometrika, 39, 1 1 1-121.

Schilling, S., & Bock, R. D. (2005). High-dimensional maximum marginal likelihood
item factor analysis by adaptive quadrature. Psychometrika, 70, 533-555.

Schleyer, T. K. L., & Forrest, J. 1. (2000). Methods for the design and administration

of Web-based surveys. Journal of the American Medical Informatics

134

 

Association, 7, 416-425.

Schroder, K. C. (1987). Convergence of antagonistic traditions? the case of audience

research. European Journal of Communication, 2, 7-31.

Seddon, P. B. (1997). A respecification and cxtension of the DeLone and McLean
Model of IS Success. Information Systems Research, 8(3), 240-253.

Seiter, E., Borchers, H., Kreutzner, G, & Warth, E-M. (1989). Remote control:

Television audiences and cultural power: London: Routledge.
Silverston, R., & Hirsch, E. (1992). Consuming technologies. London: Routledge.

Singer, E. (2002). The use of incentives to reduce nonresponse in household surveys.
In R. M. Groves, D. A. Dillman, J. L. Eltinge, & J. A. L. Roderick (Ed.),
Survey nonresponse (pp. 163-177). New York, NY: John Wiley & Sons, Inc.

Singer, E., Groves, R. M., & Corning, A. D. (1999). Differential incentives: beliefs
about pratices, perceptions of equity and effects on survey participation.
Public Opinion Quarterly, 63, 251-260.

Sparks, B. (2007). Planning a wine tourism vacation? Factors that help to predict

tourist behavioural intentions. Tourism Management, 28, 1180-1192.

Stone, C. A., & Yeh, C-C. (2006). Assessing the dimensionality and factor structure of
multiple-choice exams: an empirical comparison of methods using the

Multistate Bar Examination. Educational and Psychological Measurement, 66,
193-214.

Tate, R. (2003). A comparison of selected empirical methods for assessing the
structure of response to test items. Applied Psychological Measurement, 27(3),

159-203.

Taylor, S. & Todd, PA. (1996). Understanding information technology usage: a test of
competing models, Information Systems Research, 6(2), 144-176.

Terry, D. J. (1993). Self-efficacy expectancies and the theory of reasoned action. In D.
1. Terry, C. Gallois, & M. McCamish (Ed.), The theory of reasoned action: Its

135

 

application to AIDS-preventive behaviour (pp. 135—151). Oxford: Pergamon.

Terry, D. J ., & O‘Leary, J. E. (1995). The theory of planned behavior: The effects of
perceived behavioral control and self-efficacy. British Journal of Social
Psychology, 34, 199-220.

Theory clusters. (2006). Retrieved May 1, 2008, from http://www.tcw.utwente.nl/
theorieenoverzicht/Theory%20clusters/

‘ Thompson, R. L., Higgins, C. A., Howell, J. M. (1991). Personal computing: toward a
conceptual model of utilization, MIS Quarterly, 15(1), 125-143.

Triandis, H. C. (1977). Interpersonal behavior. Monterey, CA: Brooks/Cole.

Tse, A. C. B., Tse, K. C., Yin, C. H., Ting, C. 8., Yr, K. W., Yee, K. P., et a1. (1995).
Comparing two methods of sending out questionnaires: E-mail versus mail.
Journal of the Market Research Society, 3 7, 441-446.

Tukey, J. W. (1977). Exploratory Data Analysis. Philippines: Addison-Wesley
Publishing Company, Inc.

Valente, T. (2001). Evaluating communications campaigns. In R. E. R. C. K. Atkin
(Ed.), Public communication campaigns (pp. 105-124). Thousand Oaks, CA:
Sage.

van Dijk, T. A. (1991). Racism and the press. London: Routledge.
Van Ryn, M., Lytle, L. A., & Kirscht, J. P. (1996). A test of the theory of planned
behavior for two health-related practices. Journal of Applied Social

Psychology. 26(10), 87 l -883.

Vidmar, N ., & Milton, R. (1974). Archie Bunker's bigotry: A study in selective
perception and exposure. Journal of Communication, 24(2), 240-250.

Weiss, J.A., & Tschirhart, M. (1994). Public information campaigns as policy
instruments. Journal of Policy Analysis and Management, I 3, 82-119.

White, K. M., Terry, D. J., & Hogg, M. A. (1994). Safer sex behavior: The role of

136

 

attitudes, norms, and control factors. Journal of Applied Social Psychology, 24,
2164—2192.

Wicker, A. W. (1969). Attitudes versus actions: the relationship of verbal and overt
behavioral responses to attitude objects. Journal of Social Issues, 25(4), 41-78.

Wiesner, M, & Tesch-Romer, C. (1996). Use of hearing aids by the elderly:
Correlation between intention and behaviour. Z Gerontol Geriat, 29(4), 273-
279.

Wilcox, D.L., & Cameron, GT.(2009). Public Relations: strategies and tactics (9th
ed.). Boston, MA: Pearson Education, Inc.

Wilhoit, GC. & de Bock, H. (Eds.). (1980). Mass Communication Review Yearbook,
Beverley Hills, CA: Sage Publications.

137

Ill”

1

“3
[kilo
l1
3
6
0
3
0
3
9
2
1
3

1111111111111111