An Empirical Study of Consumer Switching from Traditional to Electronic Channel: A Purchase Decision Process Perspective Alok Gupta * ([email protected]) Bo-chiuan Su ([email protected]) Zhiping Walter ([email protected]) ALOK GUPTA ([email protected]) is an Associate Professor of the Department of Information and Decision Sciences, Carlson School of Management, University of Minnesota, USA. He received his PhD in Management Science and Information Systems from the University of Texas at Austin in 1996. His research has been published in various information systems, economics, and computer science journals such as Management Science, ISR, CACM, JMIS, Decision Sciences, Journal of Economics Dynamics and Control, Computational Economics, Decision Support Systems, IEEE Internet Computing, International Journal of Flexible Manufacturing Systems, Information Technology Management, and Journal of Organizational Computing and Electronic Commerce. He received prestigious NSF CAREER award for his research in Online Auction in 2001. His current and teaching interest are in the area of economic modeling and analysis of electronic commerce. He serves on the editorial board of ISR, DSS, and Brazilian Electronic Journal of Economics. BO-CHIUAN SU ([email protected]) is an Assistant Professor of the Department of Information Management at the National Central University, Taiwan (R.O.C.). He received the Ph.D. degree in Business Administration, specialized in information systems, from the School of Business, University of Connecticut, U.S.A. Dr. Su’s research interests include economic issues in electronic commerce, Internet marketing, and Enterprise Resources Planning (ERP). His research will appear in Decision Support Systems. Zhiping Walter ([email protected]) received the Ph.D. degree in Business Administration, specializing in Management Information System from the Simon School of Business, University of Rochester. She is currently an Assistant Professor of Management Information System at the School of Business, University of Colorado at Denver, USA. Dr. Walter’s research interests are in the areas of economics of information systems, Internet marketing, and Information Technology in Healthcare. Her research has been published or has been accepted to publish in Communications of the ACM, European Journal of Operational Research, Decision Support Systems, International Journal of Healthcare Technology Management, Technology Analysis and Strategic Management, ICIS proceedings and HICSS proceedings. * Author names in alphabetical order.
55
Embed
An Empirical Study of Consumer Switching from Traditional to
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
An Empirical Study of Consumer Switching from Traditional to Electronic Channel: A Purchase Decision Process Perspective
ALOK GUPTA ([email protected]) is an Associate Professor of the Department of Information and Decision Sciences, Carlson School of Management, University of Minnesota, USA. He received his PhD in Management Science and Information Systems from the University of Texas at Austin in 1996. His research has been published in various information systems, economics, and computer science journals such as Management Science, ISR, CACM, JMIS, Decision Sciences, Journal of Economics Dynamics and Control, Computational Economics, Decision Support Systems, IEEE Internet Computing, International Journal of Flexible Manufacturing Systems, Information Technology Management, and Journal of Organizational Computing and Electronic Commerce. He received prestigious NSF CAREER award for his research in Online Auction in 2001. His current and teaching interest are in the area of economic modeling and analysis of electronic commerce. He serves on the editorial board of ISR, DSS, and Brazilian Electronic Journal of Economics. BO-CHIUAN SU ([email protected]) is an Assistant Professor of the Department of Information Management at the National Central University, Taiwan (R.O.C.). He received the Ph.D. degree in Business Administration, specialized in information systems, from the School of Business, University of Connecticut, U.S.A. Dr. Su’s research interests include economic issues in electronic commerce, Internet marketing, and Enterprise Resources Planning (ERP). His research will appear in Decision Support Systems. Zhiping Walter ([email protected]) received the Ph.D. degree in Business Administration, specializing in Management Information System from the Simon School of Business, University of Rochester. She is currently an Assistant Professor of Management Information System at the School of Business, University of Colorado at Denver, USA. Dr. Walter’s research interests are in the areas of economics of information systems, Internet marketing, and Information Technology in Healthcare. Her research has been published or has been accepted to publish in Communications of the ACM, European Journal of Operational Research, Decision Support Systems, International Journal of Healthcare Technology Management, Technology Analysis and Strategic Management, ICIS proceedings and HICSS proceedings.
* Author names in alphabetical order.
1
An Empirical Study of Consumer Switching from Traditional to Electronic
Channel: A Purchase Decision Process Perspective
Abstract
This paper examines the relationships between the operating characteristics of the consumer
purchase decision process and the channel switching intentions of consumers. A theoretical
model that explains consumer channel switching intentions is constructed and tested based on a
sample of 337 actual consumers. The analysis indicates that the overall channel-switching
tendency from offline to online is approximately 52% across four product categories, including
books, flight tickets, wine, and stereo systems. The order of switching tendency (flight tickets,
books, stereo systems, wine) is consistent with their search and experience attributes: flight
tickets and books are search goods whereas wine and stereo systems are experience goods. The
logistic regression analysis across product categories shows that consumers’ differences in
channel risk perceptions, price search intentions, evaluation effort, and waiting time between
online and offline channels have significant impacts on their tendency of switching from offline
to online shopping. The results also indicate that those attracted to purchase online perceive
significantly lower channel risk, search effort, evaluation effort, and waiting (delivery) time
online than offline and express significantly higher price search intentions online than offline.
Although consumers attracted to offline channels also perceive lower search cost and higher price
search intentions online than offline, their perceived online search effort and price search
intentions are significantly lower than those attracted to online channels. These results provided
further support to the importance of the factors examined in influencing consumer channel
switching. It also suggests that demographics might not be effective bases for market
segmentation.
2
Keywords and Phrases: Electronic commerce, retail channel switching, risk perceptions,
purchase decisions
3
1. Introduction
Many researchers and practitioners have recognized the emergence of online shopping as a new
retail format. Online retailing sales reached $51.3 billion in 2001 and were expected to reach
$72.1 billion in 2003 [13]. In a time-constrained world, online stores allow consumers to shop
from the convenience of remote locations. However, online retailing sales just accounted for one
percent of total retail spending in 2001 and most of online stores were losing money [83, 89].
An International Council of Shopping Centers’ report [44] also forecasted that in the U.S. only
4.7% of retail sales would be online by 2005 and 5.3% by 2010. Currently the dollar amount of
products purchased via the Internet in the U.S. is about $45 billion [89]. But forecasters expect
this amount to grow to more than $100 billion, and possibly exceed $200 billion, by 2004 [35,
36]. These forecasts represent average annual growth rates of 40-80%. While this would still
amount to under 5% of all retail sales in 2004, it nonetheless represents a dramatic increase in
Internet retailing.
To investigate this phenomenon, it is critical to study what drive or inhibit the consumer’s
intentions to shop online. Specifically, an issue of particular interest to both practitioners and
academics is in understanding the consumer’s channel switching behavior (from offline to online)
and identifying the factors influencing such behavior. Based on the consumer purchase decision
process, this study identifies five major factors that potentially influence a consumer’s switching
tendency from shopping offline to online. These factors are: channel risk perceptions, price
search intentions, search effort, evaluation effort, and delivery time. The anticipated effects of
these factors are incorporated into a model, which is tested with data of four products possessing
various product characteristics both aggregately and separately: books, flight tickets, wine, and
stereo systems. Wine and stereo systems are experience goods whereas books and flight tickets
4
are search goods.1 Information economics theory distinguishes between search and experience
goods typically in terms of consumers’ ability to know quality before and after buying [23].
Search goods are those products of which quality can be assessed prior to purchase, whereas the
quality of experience goods cannot be assessed without being ascertained by product use.2 It is
clear that the taste of a bottle of wine and the sound of a stereo system are difficult to examine
and compare online because of their sensory attributes.3 One of the primary objectives of this
study then is to determine whether the impacts of these factors on consumer channel switching
are similar across search and experience goods.
While many researchers have focused on consumer demographics, the technological
characteristics of virtual stores, and the unique capabilities of the Internet medium to provide
interactive and personalized online shopping experiences [18, 47, 54, 55, 56, 74, 81], no prior
studies have focused on how the factors associated with the consumer purchase decision process
can affect their channel switching tendency differentially. Further, there is an increased interest
in understanding the attitude differences between consumers who prefer online shopping versus
traditional stores. The central questions are: are there attitude and perceptual differences between
the consumers attracted to shop online and those attracted to traditional stores, and if so, what are
1 Darby and Karni [23] add a third category, namely “credence” goods. The quality of such products cannot be determined reliably even after usage. They attribute wine to be one of the examples. However, researchers in information economics generally still consider wine as experience goods. 2 We recognize that all goods have some combination of search and experience attributes. A search good is simply one for which the consumption benefits most important to consumers are predicted reliably by attribute information available to them before buying. This reasoning implies that the same product can be a search or experience good, depending on the benefits that are important to consumers and the inferences consumers make about how well those benefits are predicted by information available prior to purchase. In the present study, we still adopt the general definition by information economics for search and experience goods. 3 In some cases, experiential attribute information could still be conveyed effectively electronically. For example, customers or chat community post their own reviews, with positive word of mouth clearly facilitating evaluating sensory or experiential product attributes. Assurance from the expertise or certification of third parties can also facilitate evaluating experiential attributes prior to purchase, such as Bizrate.com and Epinions.com. However, we do not expect such business to be commonly used by consumers and in some occasions consumers have to subscribe by paying fees.
5
such differences? Understanding the effect of such differences on consumers’ choice of
channels should be useful in devising marketing and channel strategies, since if the consumers
attracted to shopping online are different on these attitude differences from those attracted to
shopping in traditional stores, the design and marketing strategies for the two environments
should be tailored to fit the profiles of their target consumers. However, many online marketing
decisions regarding product assortment, pricing and promotional strategies rely only on what are
observed in the online environment, without knowing the exact causal explanation for the
consumer’s channel choice behavior. Consequently, our research is an important first step.
Overall, to fill these research gaps, this research has two main objectives:
to analyze consumer channel switching behavior and identify which operating
characteristics of the consumer purchase decision process lead them to switch
channel;
to investigate what kinds of consumers are more likely to be attracted to
shopping online instead of to traditional stores and their attitude differences.
This article is organized as follows. Next section provides the theoretical background for
explaining the positive relationship between purchase intentions and actual purchase behavior. It
then follows that the difference in purchase intentions between online and offline channels is a
reasonable indicator of channel switching from offline to online. Section 3 identifies and justifies
why the factors identified from the consumer purchase decision process are relevant to consumer
channel switching. Then, the research hypotheses and model are proposed accordingly. Data
collection procedures and measures are explained in Section 4. After presenting and discussing
the results of analysis in Section 5, Section 6 concludes the paper.
6
2. Theoretical Background
2.1 Purchase Intentions and Actual Purchase Behavior
In this study, the difference in purchase intentions between online and offline channels is used as
an indicator of channel switching tendency. A consumer tends to switch to online channels if his
online purchase intentions are higher than offline purchase intentions. Essentially, consumers
express preferences (stated purchase intentions) based on utility maximization in terms of the
costs and benefits of the retail formats presented to them [2]. It implies that a consumer’s utility
obtained from online shopping needs to exceed the utility provided by the traditional format to
cause the consumer to switch to an online environment.
The theory of reasoned action (TRA) asserts that behavior is influenced by behavioral
intentions [3]. Research in social psychology suggests that intentions should be the best predictor
of an individual’s behavior because they allow each individual to independently incorporate all
the relevant factors that may influence his or her actual behavior [34]. Several studies have
examined the relationship between purchase intentions and actual purchase behavior for durable
goods [1, 19, 32, 39, 62, 72] and for nondurable goods [37, 46, 88, 92]. The observed
relationship between intentions and purchase is generally positive and significant. Since Internet
shopping behavior shares the volitional nature of the phenomena that TRA tries to explain and
predict [48], the degree to which people express their intentions to purchase should therefore be a
reasonable predictor of their actual purchase behavior. It then follows that consumers’
differences in purchase intentions between online and offline channels should be a reasonable
indicator of their tendency to switch from offline to online channels. That is, a consumer’s
offline purchase intensions are used as a point of reference for assessing his tendency to switch to
online channels.
7
2.2 Factors Relevant to Purchase Decision Process
Based on the consumer purchase decision process, five factors that potentially affect consumers’
intentions to shop online and offline are identified. These factors are: channel risk perceptions,
price search intentions, search effort, evaluation effort, and delivery time. Again, using
consumers’ offline perceptions on these factors as points of reference, the differences in these
factors between online and offline channels are incorporated into a model for explaining their
channel switching tendency.
It is well established in marketing and consumer behavior literature that the consumer
purchase decision process includes five stages: problem recognition, information search,
evaluation of product options, purchase decision, and post-purchase support [29, 53]. The
purchase process starts when a consumer recognizes a problem or a need. Since the desire to buy
a product/service is largely subconscious (e.g., thirst, hunger, or admiration of a neighbor’s new
car) and the utility from consuming the product/service itself is the same no matter whether the
consumer obtains this product/service from a physical store or from an online store, it should play
a very minor, if any, role in driving the consumer to purchase online.
The next stage is information search. Information search (including price and product
information) usually incurs search effort. When purchasing a product from a brick-and-mortar
store, a consumer has to spend time browsing the aisles. If the consumer cannot find a suitable
product at the store (e.g., high prices and/or no favorable product attributes), he must keep
spending effort on additional searches. In contrast, online shopping can dramatically reduce
search effort for price and product information with just a few clicks. Specifically, the relative
ease of online search for better prices motivates consumers to shop online. Consequently,
8
consumers who have stronger price search intentions may find online shopping more attractive
than offline shopping.
The third stage of the purchase decision process is the evaluation of product options,
which incurs evaluation effort. It involves examining and comparing product attributes such as
price, brand, quality, and others. Despite the reduced search costs for price information,
consumers may feel troubled in evaluating non-price attributes online. The color and style of a
product may not be exactly as it appears when displayed on the computer screen. Product quality
may be hard to evaluate online as well. It is especially true for the “feel and touch” product
categories. For example, consumers may be apprehensive about buying something without
touching or feeling it because of quality uncertainty [9, 33]. Therefore, the online medium can
facilitate information search but impede evaluation of product options in terms of non-price
attributes.
During the evaluation stage, consumers will also evaluate their perceived risk associated
with online shopping. Risk perceptions are considered to influence consumers’ evaluation and
choice behavior [28, 79]. Research has shown that a consumer’s decision to modify, postpone, or
avoid a purchase decision is heavily influenced by his perceived risk [7, 87]. Online shopping
might be perceived to be riskier, thus reducing the overall utility that a consumer can obtain from
shopping online. However, a consumer perceiving a certain amount of online shopping risk may
or may not avoid the risk. Researchers define perceived risk in terms of uncertainty and
consequences [8, 10, 70, 79]; these two components of risk, uncertainty and consequences, have
been found in research on risk perceptions in non-marketing contexts as well [84, 85]. According
to risk theory, perceived risk increases with a higher level of uncertainty and/or a greater chance
associated with negative consequences [66]. For example, if a consumer is considering buying
an unfamiliar bottle of wine for a dinner party, the perceived risk associated with that purpose
9
could arise because he does not know how the wine will taste (uncertainty) and is worried about
his guests’ reactions if it is not a good wine (negative consequences). Thus, whether a consumer
is willing to bear a particular risk depends on his perceptions on the likelihood of the occurrence
of the risk and on the importance, or severity of the possible negative consequences, of such risk.
Consequently, this study conceptualizes a consumer’s channel risk perceptions as the interaction
effect of the likelihood and the importance of a risky situation perceived by the consumer when
buying through that channel.
After a purchase decision has been made, the product still has to be physically delivered
(except digitized products/services, of course) if the purchase is made online. Since consumers
tend to maximize utility subject to time constraints [8], the efficiency of delivery becomes a real
concern to both consumers and online retailers. Online retailers often experience low customer
satisfaction due to their poor fulfillment of on-time delivery [50]. Since different consumers
value the speed of delivery differently, time-sensitive consumers may favor a traditional channel
simply because it saves delivery time. To account for the effects of waiting problem associated
with delivery on channel preference, delivery time is included in our model.
3. Research Hypotheses
As discussed earlier, our analysis is derived from the consumer purchase decision process, and
five constructs potentially influencing the consumer’s channel switching tendency are identified.
This section analyzes the relationships between these constructs and the consumer’s tendency to
switch from offline to online channels, and then the research hypotheses are proposed
accordingly.
10
3.1 Channel Risk Perceptions
While the Internet allows consumers to shop at the convenience of remote locations, they may be
apprehensive about making purchase online if they perceive risk associated with online shopping.
Interestingly, some models of online buying behavior have excluded perceived risk as a predictor
of online shopping [58, 59, 58, 61], although others see risk reduction as a key to increase
consumers’ participation in e-commerce [11, 68]. Consumers’ perceived risk associated with
online shopping has received limited attention despite its potentially important implications for
online shopping. Some early research suggests that risk perceptions may play a minor role in the
adoption of online shopping [47], but several recent industry and government-related studies
have nevertheless considered consumers’ risk perceptions to be a primary obstacle to the future
growth of e-commerce [22, 30, 31].
Five components of perceived risk have been proposed in the literature: financial,
performance, physical, psychological, and social [45, 51, 78]. Financial risk stems from paying
more for a product than being necessary or not getting sufficient value for the money spent [76].
Consumers generally address this problem by shopping around for a more satisfactory price.
Performance risk, sometimes referred to as quality risk, is based on the belief that a product will
not perform as well as expected or will not provide the benefits desired [8]. Physical risk
involves the potential threat to consumer safety or physical health and well-being. Psychological
risk arises from the likelihood that the purchase fails to reflect one’s personality or self-image.
Social risk is concerned with an individual’s ego and the effect that the consumption is
observable by others and has on the opinions of reference groups. For shopping, the effect of
physical risk is minimum, since shopping activities usually do not involve physical risk. Note
that perceived security of online transactions and concerns for privacy should be included as
11
elements of performance risk in online shopping. Thus, except for physical risk, the other four
types of risk represent the four critical sources of channel risk perceived by consumers.
Of course, different individuals have different levels of risk perceptions toward online
shopping. Here, these risk perceptions are associated with the Internet as a purchasing medium
rather than the consequences of purchasing a particular product. Research has indicated that
perceptions of risk can be extended beyond the product to the shopping medium itself [20, 86].
Such concerns are likely to affect consumer behavior on the Internet and may help to explain why
most consumers still use the Internet for browsing rather than buying [12, 93]. The amount of
risk perceived has been suggested to be a major factor in deciding whether a consumer would
shop via a certain retail channel [20, 49, 71, 86]. Based on this perceptive, consumers who
perceive lower risk online than offline are more likely to switch to online channels than those
risk-laden consumers, resulting in the following hypothesis:
H1 (Channel Risk Perceptions): Consumers who perceive lower risk in conducting purchases
online than offline are more likely to switch to online channels.
3.2 Price Search Intentions
With the advent of the Internet, consumers expect to find lower prices more easily in the online
environment than in the offline environment. Search engines and agent technologies would
dramatically reduce search costs, prices would plummet, and deep discounting is prevalent online
[6, 9, 15]. Because of this heightened expectation for lower prices online, consumers would
demonstrate higher price search intentions over the Internet than when shopping in traditional
stores. This would have a positive impact on consumers’ tendency to switch from offline to
12
online channels, because the Internet provides a single source for consumers to evaluate their
price consideration sets, instead of searching for prices in many traditional stores. Thus, it is
held that consumers with greater positive perceived differences in price search intentions between
online and offline channels would exhibit a higher tendency to switch to online channels.
H2 (Price Search Intentions): Consumers who perceive higher price search intentions in
conducting purchases online than offline are more likely to switch to online channels.
3.3 Search and Evaluation Efforts
On the Internet, search effort (for price and non-price information) is dramatically reduced. This
should have a positive impact on consumers’ intentions to switch to online shopping. The
widespread availability of information on the Internet is one of the reasons why many buyers
view search and purchase on the Internet as a utilitarian activity [94]. Many online buyers revel
in the fact that they can get information directly without having to go through a salesperson who
usually has very limited information compared to a web site [94]. On the other hand, it may be
fairly difficult to evaluate certain types of product information online, thus impeding consumer
channel switching. This is especially true for shopping “look and feel” products [12, 33, 61].
Pictures and animation certainly help but only to a very limited extent. Alternative technologies
such as online customization tools will also assist, but, for example, unless the consumer feels the
swatch of the fabric for the suit, a purchase decision is difficult [33]. Thus, it is proposed that
when consumers perceive increasingly less search and evaluation efforts required for shopping
online than offline, they also exhibit a higher tendency to switch to online shopping.
13
H3 (Search Effort): Consumers who perceive lower search effort in conducting purchases online
than offline are more likely to switch to online channels.
H4 (Evaluation Effort): Consumers who perceive lower evaluation effort in conducting
purchases online than offline are more likely to switch to online channels.
3.4 Delivery Time
Most online transactions still involve physical product delivery, and the efficiency of delivery can
become a real burden for both consumers and online retailers.4 Among the dot-coms, their
inexperience in real-world operations, marketing, and administration has left many online
shoppers complaining about late or nonexistent deliveries. Online retailers often experience low
customer satisfaction due to their poor fulfillment of on-time delivery [50]. The speed at which
ordered items are delivered is important. If timing is so important and one of the major benefits
offered by e-commerce is its “convenience” (time-related) for shopping, then shortening delivery
time should increase the utility (benefits) of consumers (especially for time-sensitive consumers)
and thereby motivate them to purchase online. Thus, it is expected that consumers with less
concern about delivery time will be more inclined to switch to shopping online, as suggested by
the following hypothesis.
H5 (Delivery Time): Consumers who perceive shorter delivery time in conducting purchases
online than offline are more likely to switch to online channels.
The five-factor model of channel switching is shown in Figure 1.
14
*** Figure 1 about here ***
4. Data and Methodology Four product categories, books, flight tickets, wine, and stereo systems, were chosen and
available in both online and offline channels. Wine and stereo systems are experience goods
whereas books and flight ticket are search goods. It is clear that the quality and taste of wines
and the sound quality of stereo systems are difficult to evaluate and compare online. On the other
hand, the quality of books is relatively easier to evaluate over the Internet, since once a book title
has been chosen, the book itself is basically identical across retailers. The quality of flight tickets
is also easy to evaluate prior to purchase, e.g., number of connections, length of connection,
flexible departure/return time, and same day/staying overnight. Note that these products were
also selected on the basis of price level. Books and wine are low-priced products while flight
tickets and stereo systems are relatively high-priced ones.5
4.1 Pilot Study
A pilot study was conducted via a paper survey. For the pilot study, the questionnaire was tested
and modified before using for online data collection. Possible question misunderstanding was
also considered and the questionnaire was revised accordingly. A convenient student sample was
used for this pilot study. One hundred and seven business-major undergraduates and M.B.A.
students from a major university in the northeastern U.S.A. participated in the study on a
4 In July 2000, seven online retailers-including CDnow, KBkids.com, and Toysrus.com agreed to pay civil penalties totaling $1.5 million for variously failing to provide customers notice of delayed deliveries [41]. 5 We realize that the respondents may not be familiar with wine, resulting in a low level of involvement. The level of involvement a consumer has with a product is based on the relevance of that product to the consumer’s inherent needs, value and interests [94]. Involvement is known to be an important element of Internet purchasing behavior
15
voluntary base. All the 107 respondents completed the survey. Approximately 94% of the 107
respondents reported experience with online purchase in the last 12 months, and therefore are
actual online buyers.
The use of business school students as surrogates of actual online buyers might raise the
issue of external validity [14, 38, 40]. However, it has been suggested that college students are
influential and representative gatekeepers for filtering/spreading opinions and information about
Internet content to broader society [24]. Besides, in the absence of a theory or evidence showing
why college students would react in a peculiar way, the use of students should not represent a
threat to applicability [64]. Thus, for the purpose of pilot testing, it is believed that students are
appropriate for refining and validating the survey instrument.
4.2 Main Study
The data used in the main study were collected via a Web survey. The Web survey is a visual
stimulus and under the respondent’s complete control with regard to whether and/or how each
question is read and comprehended. Therefore, it is expected that responses to Web
questionnaires will closely resemble those observed via mail questionnaires [26]. In addition, the
inherent characteristics of Web surveys are similar to those of online shopping. Web surveys are
therefore particularly suitable for conducting research in the online shopping realm. Further,
Web surveys also have significant advantages such as speed, the ability to collect large amount of
data without interviewers, the elimination of postage and stationary costs, and low processing
costs without separate data entry.
[17]. But as discussed earlier, we intentionally chose to study wine and wished to demonstrate that the effects of certain factor (channel risk perceptions) may be pronounced and outweigh those of other factors.
16
Considerable effort was exerted to carefully design and extensively tests the instrument to
make it as user friendly as possible. The main purpose of respondent-friendly design for the Web
survey is to decrease measurement and non-response errors. Such design may also have
coverage benefits by helping ensure an equal chance for people with various types of browsers,
computer equipment, and operating systems to receive and complete the Web questionnaire. We
followed the principles for the design of Web surveys as suggested in Dillman [26]. Meanwhile,
the Web survey was also designed to eliminate those possible causes of respondent frustration
observed by Dillman and Bowker [27].
In order to reduce the possibility of multiple submissions by an individual, cookie
technology was used to ensure all respondents answered the questionnaire only once. 6 In
addition, we restrained the size of response categories within tables to reduce measurement errors
due to attitude scales with physical distances between points on the scale, which changes as a
result of (1) changing the screen configuration from 800*600 to one of the other used sizes (e.g.,
640*480 or 1024*768) or (2) using a lower level browser. Further, the Web questionnaire was
constructed to emulate visual aspects of the paper questionnaire. Thus, a similarly colored
background and placement of questions on the page were designed to ensure a very similar
stimulus to traditional mail surveys.
The invitations to participate in the study were distributed through an email list. A list of
50,000 email addresses was purchased from a private company. This sample frame has known
characteristics (consists of individuals who are at least 20 years of age and had sent in registration
6 A cookie simply consists of a text-only string that enters into the memory of a browser. This string contains the domain, path, lifetime, and value of a variable that a website sets. If the lifetime of this variable is longer than the time the user spends at that site, then this string (completion state variable) is saved to file in client’s computer for future reference.
17
cards of warranties for a wide variety of computer hardware and software) in comparison with
many of the Web surveys having no specific sample frames and soliciting responses from a
variety of sources, such as advertising, announcement on Web pages, and message boards of
commercial and academic Web sites. It is clearly impossible to calculate the response rate for
these kinds of Web surveys that do not have a pre-selected sample frame. Although it was not
our objective to have the sample limited to those purchasers of computer hardware and software,
the availability of a sample frame with known characteristics led us to adopt this list. Moreover,
the subjects in our sample frame most likely have adequate equipment to conduct online
shopping. Indeed, approximately 98% of the participants in our study had made purchases on the
Internet in the previous 12 months.
The survey was administered over a period of four weeks (4/3 – 5/1/02). It is important to
note that many other Web surveys were conducted over a period of several months to a year, in
contrast to the four weeks for this research. Email invitations were actually delivered to 18,988
addresses (38%) of the original 50,000-email list, due to returned emails as undeliverable and
other unexpected delivery problems. To protect the anonymity of the survey receipts, special
coding was developed to send out email individually; i.e., concealing the names of survey
recipients and place one’s own email in the regular “send to” field. In so doing, the survey
recipients only saw one email address – that of the individual who sent it. The email included an
explanation of the purpose of the survey and confidentiality. The email also included the link to
the survey URL and contained “institution.edu,” the names of professors, and contact phone,
sending a signal that we are members of a higher education institution (and thus increasing our
credibility).
Response rates of Web surveys are likely to be very low [90]. The response rate in this
study is not an exception and hovers around 2% (a total of 337 complete responses were
18
collected). This might lead to the possibility of a non-response bias, that is, if the respondents are
different from the non-respondents on the questions that are being asked. However, non-response
error is not conceptually equivalent to response rate [27].7 Also, calculating response rates can
be challenging, because calculating the total numbers who received the email and “open” it is
questionable (the denominator in the response rate). Although to test actual delivery and opening
of the email to non-respondents, a read receipt acknowledge can be attached. In this way, it is the
“confirmed” email “contact” rate. But not all mail systems are compliant with this read receipt
acknowledge attached and many people may choose to return the receipt without even reading
the email, still causing difficulty in computing the “confirmed” response rate.
Nonetheless, we attribute the occurrence of low response rate to five factors: (1) survey
fatigue, (2) lengthy questionnaire, (3) the use of Java technology, (2) no inclusion of incentives,
and (5) methodological differences. Survey fatigue is common, as public opinion polls have
become more popular with the media and telemarketers using surveys for data mining research.
The email list that we purchased from a private company might be a weak response pool. Many
individuals on the list should have been contacted by a variety of surveys and are sensitive to
“spamming.” Given the possibly large amount of spamming and unwanted solicitations via email,
a low email survey response rate is likely. However, we believe that this might have a relatively
little impact on our study, as our email contained “institution.edu” and other contact information
for increasing our credibility. The survey URL was also clearly located in the university domain.
Another factor that undoubtedly contributes to the low response rate is the lengthiness of the
questionnaire. In addition to four demographic questions, the questionnaire contains 38 questions
7 But we recognize that there still is a possibility that the higher the proportion of sampled respondents who respond to a survey, the lower the likelihood of non-response error.
19
for each product category. Since the survey requires the subjects to assess four product
categories, the total number of questions is 156, which is likely to significantly reduce the
subjects’ interest in responding to the survey. Our analysis showed that the respondents on
average spent 28 minutes answering the survey, with a standard deviation of 4 minutes.
Apparently, completing this questionnaire requires considerable effort and patience, thus leading
to a low response rate. We also found that the use of Java Script technique for the survey might
make it impossible for many subjects to access the survey and to submit the completed Web
questionnaire. This would reduce the response rate as well. Further, nonresponse might occur
because some subjects were prevented from accessing some aspects of the questionnaire due to
the incompatibilities of hardware and software (e.g., various levels and types of Web browsers).
Such incompatibilities might cause some of the response features to be disabled or perhaps
rearranged on the screen. No inclusion of incentives could be a factor of the low response rate as
well. Past research has shown that survey response rates can be increased by 15-20% by
including such incentives as cash and gift certificates [57]. But because of the budget constraints,
we were unable to offer such incentives. Besides, unlike mailed surveys, it is impossible to
include prepaid incentives such as a dollar bill with an email invitation to the Web survey, even
though we recognized that incentives paid upon competition (post-payment) are possible with
Web surveys. Lastly, a methodological difference may also help explain the low response rate of
this study. This study used a separate Web survey to capture respondent data, rather than an
email survey [63]. Using the Web form requires respondents to access another computer-
mediated communication medium and perform an addition task (use a Web browser), in order to
participate in the survey. Additional steps and tasks with the Web survey might further reduce
the survey response rate.
20
Finally, with the deluge of emails from various ListServs and advertising campaigns,
many individuals are sensitive to “spamming” and thus may take offense at receiving multiple
emails from a researcher they do not know. Consequently, we did not send out a reminder email
after the first email to increase the response rate. The questionnaire is provided in Appendix A,
including the measures of the theoretical factors. The questions pertaining to the factors
examined are common to both online and offline shopping. By having both sets of questions
(online and offline), we were able to compare the attitudes and behavior of consumers in
comparable scenarios where the only difference was the purchase channel.
4.3 Reliability of Instrument The reliabilities of the multi-item scales, assessed by Cronbach’s alpha, for measuring channel
risk perceptions and price search intentions are satisfactory (both online and offline) and
provided in Appendix A. For the two-item measure of purchase intentions, the Pearson
correlation between the two items was over 0.76. The alphas for price search intentions (4 items)
were 0.92 and 0.9 for online and offline, respectively. Although the alphas were 0.61 and 0.60
for online and offline channel risk perceptions (5 items), they still indicated acceptable reliability
[21]. The validity of these results is also supported by [80], which found that Cronbach’s alpha
should be in the range of 0.56 – 0.94 for factors measured by 3-7 items. For all the multi-item
measures, the scales were obtained by adding the item scores. The means, standard deviations,
and intercorrelations of the scales are given in Table 1.
*** Table 1 about here ***
4.4 Demographics
21
The respondents’ ages range from 21 to 91 years old, with a mean of 47 (median = 47). Not
surprisingly, examining the respondents by gender revealed a male dominance (72% male, 28%
female); perhaps, this is in part due to that the respondents are purchasers of computer hardware
and software. Analysis of the respondents by educational background revealed a relatively high
level of education. A significant number of the respondents (67%) have a four-year college
degree. This compares favorably with the national average of 24.4% for the same educational
level [89]. Annual household gross income ranges from $0 - $20,000 to over $120,000, with a
median of $70,000 - $90,000. Comparisons of the respondents with the Internet customers of
other third-party data are shown in Table 2. The last column in Table 2 is the U.S. population
demographics according to the latest U.S. Census. Compared to those of the survey data, the
demographic characteristics of our respondents appear to be somewhat older, more educated,
wealthier, and male predominant. Appendix B presents the respondents’ online activity profiles,
indicating that these nonstudent respondents are comprised of not only actual but also frequent
online buyers: 98% of respondents reported experience with online purchasing in the last 12
months, and 73% of respondents made at least 7 purchases online, 36% at least 12 purchase, and
17% at least 24 purchases in the last 12 months. This sample, therefore, cannot be thought to
represent the population of online shoppers. But this frequent online buyers segment is important
[54] as online retailers are striving for determining what features attract frequent online buyers to
return and what they can do to further encourage repeat customers to become loyal customers.
For example, in a survey by Yahoo! Store, a store hosting service, more than 85 percent of stores
received fewer than 10 percent of their orders from frequent buyers [4].
*** Table 2 about here ***
22
5. Results
5.1 Channel Switching Tendency To analyze consumers’ tendency of switching from offline to online channels, a binary switching
variable was coded 1 if online purchase intentions were greater than offline purchase intentions
and zero otherwise. At the aggregate level, the analysis indicated that the overall channel-
switching tendency from offline to online was approximately 52% across the four product
categories. As Table 3 shows, the order of switching tendency (flight tickets, books, stereo
systems, wine) is consistent with the products’ search and experience attributes. These results
suggest that books and flight tickets should be more successful in alluring consumers to switch
from physical to online channels, whereas wine and stereo systems dominated by experience
attributes (taste and sound quality) should fare less well in inducing channel switching. Our
findings support that merchandises purchased on the basis of search attributes are more amenable
to electronic retailing, whereas merchandises purchased on the basis of experience attributes are
more likely to be purchased in physical stores. To examine the effects of the five factors
identified in the study on consumer channel switching, five logistic regression models are tested
and discussed in the following section.
*** Table 3 about here ***
5.2 Factors Influencing Channel Switching
A logistic regression model was constructed to predict consumer channel switching as a function
of the difference scores between online and offline channels of the five factors. The channel-
switching model is represented as a logistic regression model with the following structure:
Table 2: Comparison of Demographics in the Current Study with Other Third-Party Data
Sources
This Study Ernst&Young * InsightExpress * U.S. Population *
Age (Median) 47 40-49 38 35.3 Gender (Male) 72% 69% 49% 49.1% Income (Median)
$70,000- $90,000
$50,000 - $70,000 $49,800 $50,046
Education (College Grad) 67% N.A. N.A. 24.4%
* Ernst & Young Internet Shopping Study 19998 * InsightExpress Study 20019 * Latest U.S. Census 200010
8 The second annual Ernst & Young Internet shopping study: the digital channel continues to gather steam, 1999/11/12, available at http://www.ey.com/publicate/consumer/pdf/Internetshopping.pdf. 9 InsightExpress e-RDD Study, 2001, available at http://www.insightexpress.com/audiences/methodology.asp. 10 U.S. Census 2000, available at http://www.census.gov/prod/2002pubs/c2kprof00-us.pdf.
50
Table 3: Channel Switching Tendency (from offline to online)
Across Categories Books Flight
Tickets Wine Stereo Systems
Switching Tendency
52%
66%
83%
18%
40%
51
Table 4: The Relationship between Channel Switching and Purchase Decision Variables
Search Effort −.037 −.1 −.186 −.046 −.037 Evaluation Effort −.251*** −.299*** −.183* −.172* −.225*** Delivery Time −.142*** −.238*** −.089 −.217** −.140* Nagelkerke 2R 11 .38 .37 .48 .26 0.36 Hit Ratio 74% 79% 88% 83% 71% Dependent variable: switch = 1, non-switch = 0 Based on one-tailed test * p < 0.05 ** p < 0.01 *** p < 0.001
11 Nagalkerke’s R-square is comparable to the R-square in multiple regressions, also ranging from 0 to 1 with higher value indicating greater model fit. In contrast, the Cox and Snell R-square measure is limited and it cannot reach the maximum value of 1.
52
Table 5: Comparison of Consumers Attracted to Online Shopping vs. Traditional Stores