118
134/ Journal of Marketing,April 2006
Trust
Trust has been defined as “a willingness to rely on an
exchange partner in whom one has confidence” (Moorman,
Zaltman, and Deshpandé 1992, p. 315); “a generalized
expectancy held by an individual that the word, promise,
oral or written statement of another individual or group can
be relied upon” (Rotter 1980, p. 1); and “a belief in a per-
son’s competence to perform a specific task under specific
circumstances” (Sitkin and Roth 1993, p. 373). Reflected in
these and other definitions of trust is a cognitive aspect (i.e.,
trusting beliefs) and a behavioral aspect (i.e., trusting inten-
tions) (Kim et al. 2004; Moorman, Zaltman, and Deshpandé
1992).
Trusting beliefs represent a “sentiment, or expectation
about an exchange partner’s trustworthiness” (Moorman,
Deshpandé, and Zaltman 1993, p. 315). Although various
trusting beliefs have been studied in the literature, the
majority can be conceptually clustered into three dimen-
sions: ability, benevolence, and integrity (McKnight,
Choudhury, and Kacmar 2002). “Ability beliefs” reflect
consumers’ confidence that the firm has the skills necessary
to perform the job (Mayer, Davis, and Schoorman 1995),
“benevolence beliefs” reflect confidence that the firm has a
positive orientation toward its consumers beyond an “ego-
centric profit motive” (Mayer, Davis, and Schoorman 1995,
p. 717), and “integrity beliefs” reflect confidence that the
firm adheres to a set of moral principles or professional
standards that guide its interactions with customers. These
trusting beliefs are related, yet distinct. For example, con-
sumers may believe that the firm cares about its customers
and thus intends to deliver a smooth, error-free transaction
(i.e., the firm is benevolent), but they may also believe that
the firm lacks the ability to do so. Likewise, although
integrity and benevolence beliefs are similar, the former
focuses on meeting objective standards of corporate citizen-
ship, and the latter focuses on customer welfare that goes
beyond normal business activity. For example, despite con-
sumers’ beliefs that the firm follows a professional code of
conduct (i.e., has integrity), they may still question the
firm’s genuine concern for its customers (i.e., its
benevolence).
Although ability, benevolence, and integrity beliefs are
acknowledged as conceptually distinct (e.g., Kumar, Scheer,
and Steenkamp 1995), they are often combined into a
global measure of trusting beliefs (e.g., Doney and Cannon
1997). Whereas combining these beliefs into a single varia-
ble is a parsimonious approach to studying trust, it can
make it difficult to identify what action should be taken to
build trust (Smith and Barclay 1997). Because a global
measure likely obscures the reason certain signals are more
effective than others in affecting online purchase intentions,
we treat each trusting belief separately.
Trusting intentions represent “a willingness to make
oneself vulnerable to another in the presence of risk” (Kim
et al. 2004, p. 105). What distinguishes trusting intentions
from other types of behavioral intentions is that they
involve risk (Moorman, Zaltman, and Deshpandé 1992). As
reflected in the opening quotation, purchasing online
involves risk, especially when a person lacks experience
with the online firm. Specifically, the consumer must be
willing to transfer resources (e.g., credit card and other per-
sonal information) to the online firm, the consequences of
which could be damaging. For example, among the real
and/or perceived risks are that the firm may overcharge, fail
to deliver the product, deliver an inferior product, or fail to
protect personal information. To the extent that consumers
are concerned about these and other risks of purchasing
online, online purchase intentions reflect trusting intentions.
The distinction between trusting beliefs and trusting
intentions has been acknowledged by some researchers
(e.g., Moorman, Zaltman, and Deshpandé 1992; Sirdesh-
mukh, Singh, and Sabol 2002) but ignored by others who
have studied only trusting beliefs, implicitly assuming that
these beliefs imply trust (e.g., Doney and Cannon 1997;
Ganeson 1994; Kumar, Scheer, and Steenkamp 1995;
Mayer, Davis, and Schoorman 1995; Morgan and Hunt
1994). For example, Morgan and Hunt (1994) argue that
trusting beliefs are sufficient for measuring trust because
such beliefs imply that trusting intentions will follow. In
contrast, Moorman, Zaltman, and Deshpandé (1992) argue
that trust is limited when trusting beliefs do not accompany
a corresponding trusting intention or when trusting inten-
tions occur without corresponding trusting beliefs (e.g.,
under conditions of coercion or limited alternatives). In
other words, these researchers argue that both trusting
beliefs and trusting intentions must be present for trust to
exist. Likewise, we argue that trusting beliefs are a neces-
sary but not sufficient condition for trust to exist, because
increasing trusting beliefs will not always have a corre-
sponding positive effect on trusting intentions.
Conceptual Framework
Drawing from research on trust (Mayer, Davis, and Schoor-
man 1995; Moorman, Deshpandé, and Zaltman 1993), con-
sumer goals (e.g., Hoffman and Novak 1996), and market-
ing signals (e.g., Kirmani and Wright 1989; Prabhu and
Stewart 2001), we develop a conceptual framework for
understanding how different signals influence ability,
benevolence, and integrity beliefs and thus influence online
purchase intentions (see Figure 1). Because our objective is
to understand how to increase consumers’ willingness to
buy online, we focus on those whose goal is most consistent
with buying online, namely, consumers who search for
product information (or searchers; Hoffman and Novak
1996; Moe 2003; Schlosser 2003). Indeed, searchers think
about and are persuaded more by product information
(Schlosser 2003) and have higher visitor-to-buyer conver-
sion rates than those who do not search (Moe 2003). We
begin by examining the relationship between searchers’
trusting beliefs and intentions.
Searchers’Trusting Beliefs and Online Purchase
Intentions
Searching reflects purposive, task-specific behaviors, such
as the planned acquisition of information during prepur-
chase deliberation (Hoffman and Novak 1996; Janiszewski
1998). Similar to those who read a text to find an answer to
a question (Rosenblatt 1978), searchers are likely motivated
to find the right answer efficiently. Such a fact-gathering,
89
Converting Web Site Visitors into Buyers / 135
Ability
Online purchase intentions
Signals
Trusting Beliefs
Moderators
Trusting Intentions
Benevolence
Integrity
Searchers only
Browsers only
Strong privacy/security statement
(Study 2)
Web site investment
(Studies 1–4)
Perceived risk
(Study 4)
Goal:
search or browse
(Study 3)
FIGURE 1
Conceptual Framework of the Effect of Online Signals on Trusting Beliefs and Intentions
knowledge-seeking stance is typically outcome oriented,
concentrated, impersonal, and objective (Rosenblatt 1978).
Given this performance orientation, we expect that when
considering whether to purchase online, searchers will
focus on the trusting belief that is most relevant to perfor-
mance: ability (Mayer, Davis, and Schoorman 1995). If this
is the case, searchers’ ability beliefs should largely influ-
ence their online purchase intentions. In contrast, their
beliefs about the firm’s trustworthiness on non-
performance-related dimensions (i.e., benevolence and
integrity) should have relatively little effect on their online
purchase intentions.
H
1
: Searchers’ online purchase intentions depend on their
beliefs about the firm’s ability rather than their beliefs
about the firm’s benevolence or integrity.
Signaling Ability Through Web Site Investment
If trust in a firm’s abilities is critical to increasing online
purchase intentions, a fundamental question is, How can
firms use online cues to communicate that their abilities can
be trusted? To address this question, we draw from research
on marketing signals. Signals are the actions or announce-
ments that convey a firm’s abilities and intentions (Porter
1980). Marketers often use observable signals (e.g., price,
warranties, advertising expenditures) to communicate the
level of some unobservable quality (e.g., product quality;
Kirmani and Rao 2000). Signaling may be especially
important in an online purchasing context because of the
inherent asymmetry of relevant information between buyers
and sellers. Specifically, information about the quality of a
given online transaction is generally unobservable by con-
sumers before purchase.
1
This assumes conditions of normality. Because a poorly
designed Web site (e.g., one with hyperlinks that do not work)
Investing in Web site design may be one observable sig-
nal that firms can use to communicate their abilities and
boost searchers’ online purchase intentions. Indeed, con-
sumers can distinguish between expensive and inexpensive
marketing tactics, such as ad production elements (Kirmani
and Wright 1989). Moreover, they make inferences about
companies (e.g., the company’s ability to make quality
products and its credibility) based on these perceived mar-
keting expenditures (for a review, see Kirmani and Rao
2000). The attribution literature provides insight into these
findings (Kirmani and Wright 1989). Specifically, when
interpreting others’ performance, people infer that investing
time and energy promotes success (Weiner 1986). Likewise,
consumers will likely infer that a firm that has invested in
Web site design can successfully handle online transactions.
As in prior research, we consider investment in broad
terms (Kirmani and Wright 1989); that is, investment
reflects expenditures of time, money, and effort to Web site
design. Importantly, it refers to investments in the front-end
(design) elements of a Web site (i.e., its observable charac-
teristics) and not back-end technologies, such as order ful-
fillment software, security encryption, and firewall capabil-
ities, which are typically unobservable before purchase. Yet
because ability is a stable, internal characteristic (Weiner
1972), people will likely generalize their trust in a firm’s
ability in one area (design) to other related areas (e.g., order
fulfillment). Thus, instead of being purely cosmetic, Web
site design likely communicates important performance
information. However, it likely communicates less about the
firm’s goodwill, ethics, values, or intentions to mislead than
it does about the firm’s ability.1 Consequently, consumers
124
136/ Journal of Marketing,April 2006
likely violates consumers’ expectations about norms of conduct
for e-commerce firms, it will likely raise concerns about the firm’s
benevolence and integrity.
will likely use perceived Web site investment to infer a
firm’s ability more than its benevolence or integrity. If this
is the case and ability beliefs largely influence searchers’
online purchase intentions (H
1
), it follows that
H
2
: Searchers’ online purchase intentions will be higher at a
high-investment Web site than at a low-investment Web
site.
H
3
: Beliefs about the firm’s ability will mediate the relation-
ship between site investment and searchers’ online pur-
chase intentions.
The first two studies test these hypotheses, and the last
two studies examine two potential moderators (goal and
perceived risk). Furthermore, to test the generalizability of
these effects, we varied across studies the samples recruited
(students versus nonstudents), companies (a fictitious ver-
sus well-known firm), and products (home furnishings and
accessories versus cameras).
Study 1
Method
Sample and design. The sample consisted of 111
respondents who participated in exchange for $10 and were
recruited through an electronic and a printed newsletter dis-
tributed to university employees. The sample was 68%
female, with a mean age of 37.5 and a mean income of
$35,000 to $49,999. Respondents had a median education
of four years of college and used the Internet an average of
four to six times per week. We randomly assigned respon-
dents to a high- or low-investment site.
Web site investment manipulation. We manipulated Web
site investment through the presence of sophisticated Web
site technology and visual design elements. Specifically, the
high-investment site had a white background, sophisticated
fonts (images for the navigation bar; Garamond font), and
an enhanced zoom feature created with Design Within
Reach. This enabled users to zoom in on any part of an
image and to choose among three preselected zooms that
executed automatically with a single click. In contrast, the
low-investment site featured the default background color
and font (gray; Times New Roman) and a limited zoom fea-
ture, which, when clicked, simply provided a larger view of
the focal product. The content and layout of both sites were
identical.
To test the effectiveness of this manipulation, 43 under-
graduates viewed screen captures of the homepage (font
and background color were used to manipulate investment)
and a zoom page (technology was used to manipulate
investment). The order was counterbalanced, and partici-
pants viewed only those pages specific to the high- or low-
investment site. After viewing each page, participants
reported how much time, effort, and money they believed
the firm invested in each page on a scale from 1 (“very lit-
tle”) to 7 (“a great deal”). We averaged the responses to
these items (αs > .94) and analyzed them with a 2 (invest-
ment: high versus low) × 2 (page: zoom versus home) × 2
(order: viewed home or zoom page first) analysis of vari-
ance (ANOVA). In support of the investment manipulation,
participants perceived the high-investment site as requiring
greater investment than the low-investment site (Ms = 4.09
versus 3.18; F(1, 39) = 9.02, p < .01). Furthermore, the
effectiveness of the investment manipulation was unaffected
by the type or order of page viewed (Fs(1, 39) < 3.35, not
significant [n.s.]). Thus, it appears that font, background
color, and use of technology communicate investment. Par-
ticipants also reported how informative, entertaining, and
well organized they found the pages to be on the same
seven-point scale. As we expected, investment did not affect
these variables (Fs(1, 39) < 1.97, n.s.).
Procedure. Participants sat at a computer and received a
paper booklet. The first page instructed participants that
they would be visiting a site for a new firm called Urban-
Furniture (UF) and to limit their visit to the homepage and
living-room sections of the site. All participants were told to
imagine that they wanted to purchase contemporary furni-
ture and were considering UF. As in prior research
(Schlosser 2003), to instill a searching goal, participants
were asked to write down two questions they had for UF
about its products before visiting the site.
Participants then visited either the high- or the low-
investment site, which was preloaded on their computer.
Afterward, they completed the survey, which contained
three items that measured their online purchase intentions
(α = .91) and a modification of Mayer and Davis’s (1999)
scale of trustworthiness (for the measures used in this and
the other studies, see the Appendix). This scale measured
beliefs about UF’s ability (α = .90), benevolence (α = .88),
and integrity (α = .71). Then, to test the effectiveness of the
investment manipulation in the main experiment, partici-
pants completed the three-item Web site investment scale
(see the Appendix; α = .95).
At the end of the survey, participants reported their edu-
cation and income levels as well as how often they used the
Internet to purchase goods in the last six months on a scale
ranging from 1 (“not at all”) to 7 (“quite often”). We
included these measures to control for individual differ-
ences in Internet experience both directly (i.e., through self-
reported use of the Web) and indirectly (i.e., using demo-
graphic variables associated with Web use). To control for
variance due to mechanical error, we asked participants to
report the extent to which they encountered problems at the
UF site (e.g., error messages, server delays, crashing) on a
scale from 1 (“not at all”) to 7 (“quite often”).
Results
Manipulation checks. We analyzed the investment
manipulation with a one-way analysis of covariance
(ANCOVA), controlling for reported problems with the site,
prior Web purchase history, education, and income. In sup-
port of the manipulation, perceived Web site investment was
higher among those in the high-investment condition than
among those in the low-investment condition (Ms = 3.46
versus 1.97; F(1, 108) = 32.45, p < .0001).
Trusting beliefs and online purchase intentions. We
used hierarchical regression to test H
1
. We first modeled
157
Converting Web Site Visitors into Buyers / 137
online purchase intentions as a function of ability beliefs.
As we predicted, ability significantly influenced these
intentions (β = .27, t(108) = 2.88, p < .005; R2 = .07,
F(1, 108) = 8.31, p < .01; see Table 1). Furthermore, the
addition of benevolence and integrity beliefs to the model
did not contribute significantly to explaining searchers’
online purchase intentions (ΔR2 = .007, F(2, 106) < 1). We
replicated this pattern of results with a stepwise regression
analysis, which identifies the best subset of belief variables
that predict online purchase intentions. Thus, regardless of
the regression procedures we used, the results are consistent
with H
1
. Furthermore, given the inherent correlation
between trusting beliefs, we tested for multicollinearity by
examining the maximum variance inflation factor (VIF).
Multicollinearity is problematic for interpreting regression
analyses when the maximum VIF is greater than 10 (Neter,
Wasserman, and Kutner 1990). We found that multi-
collinearity was not an issue in the preceding analysis (the
maximum VIF = 3.10).
Web site investment, trusting beliefs, and online pur-
chase intentions. We analyzed ability beliefs with a one-
way ANCOVA. As we expected, ability beliefs were higher
for those who visited the high-investment site than for those
who visited the low-investment site (Ms = 2.41 versus 1.80;
F(1, 109) = 13.27, p < .005). In contrast, Web site invest-
ment did not affect benevolence or integrity beliefs
(Fs(1, 109) <
2.00, n.s.). Thus, Web site investment is more
effective in communicating the trustworthiness of a firm’s
ability than its benevolence and integrity.
In support of H
2
, online purchase intentions were higher
among those who visited the high-investment site than
among those who visited the low-investment site (Ms =
–.13 versus –.78; F(1, 109) = 4.06, p < .05). To test whether
ability beliefs mediated this effect (H
3
), we added ability
beliefs as a covariate to the ANCOVA. Consistent with the
requirements for mediation (Baron and Kenny 1986), abil-
ity was significant (F(1, 108) = 4.31, p < .05), and the
investment effect became nonsignificant (F(1, 108) = 1.48,
n.s.). We found further support for H
3
using the criteria that
Sobel (1982) endorsed for testing mediation (Goodman I
test statistic = 2.22; p < .05). Because investment did not
affect benevolence and integrity beliefs, these beliefs cannot
be considered mediators.
Conclusions
Study 1 provides support for H
1
–H
3
. Specifically,
searchers’ online purchase intentions were influenced by
their trust in the firm’s ability rather than their trust in its
benevolence and integrity. As a result, ability signals (i.e.,
Web site investment) influenced their online purchase inten-
tions. Furthermore, their trust in the firm’s ability mediated
this effect.
It is possible that searchers’ online purchase intentions
were influenced by their ability beliefs rather than by their
benevolence and integrity beliefs because there were no sig-
nals regarding the firm’s benevolence and integrity. In the
presence of such signals, ability beliefs may have less influ-
ence. Indeed, the impact of a given signal is weakened
when other, more relevant signals are present (Kirmani and
Wright 1989). If benevolence and integrity signals are more
relevant, the effect of investment on searchers’ online pur-
chase intentions should be weaker when such signals are
present. However, if ability is more relevant, investment
should affect searchers’ online purchase intentions regard-
less of whether benevolence and integrity signals are pre-
sent. We directly test this in Study 2 by manipulating the
presence and strength of a firm’s privacy and security state-
ment. Because benevolence represents the firm’s orientation
toward customers and integrity represents whether the firm
will do what it promises (Mayer, Davis, and Schoorman
1995), one method of signaling a firm’s benevolence and
integrity may be through formal statements of its intentions
to consumers, such as through privacy and security state-
ments. If so, such statements should affect consumers’
beliefs about a firm’s benevolence and integrity rather than
its ability. Yet if ability is a stronger driver of searchers’
TABLE 1
Hierarchical Regression Analysis of Trusting Beliefs on Online Purchase Intentions
Model 1
Model 2
Study
Goal
Variable
ββ
R2
F for R2
ββ
ΔΔR2
F for ΔΔR2
1
Searchers
Ability
.27*
.07
08.31*
.19*
.01
.38*
Benevolence
.10*
Integrity
.02*
2
Searchers
Ability
.39*
.15
15.57*
.39*
.00
.08*
Benevolence
.06*
Integrity
–.06*
3
Searchers
Ability
.42*
.18
15.70*
.43*
.03
1.50*
Benevolence
.15*
Integrity
–.20*
Browsers
Ability
.15*
.02
01.62
.10*
.10
4.28*
Benevolence
.33*
Integrity
–.01*
*
p
<
.05.
127
138/ Journal of Marketing,April 2006
2Our analysis was limited to either strong or weak privacy and
security statements. Thus, for ease of exposition, we use the term
“privacy/security” to describe levels of privacy and security.
online purchase intentions (H
1
), investment should affect
searchers’ online purchase intentions regardless of whether
a privacy and security statement is present. That is, H
2
and
H
3
should be supported even when such a statement is
provided.
Study 2
Method
Sample and design. A total of 79 undergraduate students
participated in exchange for extra course credit. We ran-
domly assigned participants to one of six conditions in a 2
(investment) × 3 (privacy/security statement: strong or weak
versus absent) design.2 We included the absent condition to
examine whether the presence of a strong statement is better
than having no statement. We included the weak condition
to examine whether merely having a privacy/security state-
ment might signal benevolence and integrity or whether the
contents of the statement influence such beliefs.
Privacy/security statement manipulation. We con-
structed the strong and weak privacy/security statements on
the basis of a content analysis of the privacy/security state-
ments gathered from more than 25 sites. For the strong
statement, explicit information was available about how UF
collects and uses customer information. There was also a
promise of confidentiality, a contact number, an opt-in fea-
ture, encryption information, and a 100% guarantee against
information theft. In contrast, the weak statement informed
the consumer that personal information would be collected
and made available to other vendors that “are offering prod-
ucts we feel are of interest to you.” There was no opportu-
nity to opt in or out of such correspondence. Consumers
were also informed that UF “tries to safely transmit your
account information.” No account protection or guarantee
was offered.
To pretest this manipulation, 37 undergraduates read
either the strong or the weak statement and rated its strength
on six items. Among the items were “I believe Urban-
Furniture is concerned about my privacy” and “I believe
that Urban-Furniture is concerned about the security of my
financial information.” Participants responded on a scale
ranging from 1 (“disagree strongly”) to 7 (“agree
strongly”), and we averaged the responses (α = .94). In sup-
port of the privacy/security manipulation, participants
agreed more that UF would preserve their privacy and secu-
rity when they read the strong privacy/security statement
than when they read the weak statement (Ms = 5.56 versus
2.61; F(1, 36) = 100.53, p < .01).
Procedure. The procedure and survey were the same as
in Study 1, with a few exceptions. To increase involvement,
participants were told to imagine that they accepted a job in
Manhattan after graduation and were searching for living-
room furniture for their apartment. After viewing the site,
participants in the strong and weak privacy/security condi-
tions read these statements before completing the survey. To
test whether reading such statements might artificially
increase participants’ risk perceptions of shopping online,
we added seven items to the end of the survey that mea-
sured such concerns (see the Appendix). We averaged
responses to provide a perceived risk score (α = .89).
Because the sample is homogeneous in terms of education
and income, we deleted these demographic questions. In
addition, because we measured online purchase experience
in Study 1 with a single item that captures only recent
online purchase experience, we added an item that mea-
sures general online purchase experience (i.e., how often
participants shop online) on a scale from 1 (“not at all”) to 7
(“quite often”). We averaged these items to capture online
purchase experience (r = .78).
Results
Manipulation checks. In support of the investment
manipulation, a 2 (investment) × 3 (privacy/security state-
ment) ANOVA yielded a significant investment effect
(F(1, 73) = 6.84, p = .01): Participants perceived the high-
investment site as requiring a greater investment than the
low-investment site (Ms = 3.36 versus 2.48). No other
effects were significant (Fs(1, 73) < 1).
To examine whether reading a privacy/security state-
ment might increase participants’ perceived risks of shop-
ping online, we compared risk perceptions using a 2 × 3
ANOVA. None of the effects were significant (the invest-
ment effect: F(1, 73) = 1.22, n.s.; the direct and interactive
effects of the privacy/security statement: Fs(2, 73) < 2.11,
n.s.). On average, participants perceived buying online as
risky (M = 5.71, which is significantly higher than the mid-
point of 4; higher numbers reflect greater perceived risk,
t(78) = 14.16, p < .01).
Trusting beliefs and online purchase intentions. As in
Study 1, we tested H
1
using hierarchical regression. As we
predicted, ability beliefs influenced searchers’ online pur-
chase intentions (β = .39, t(77) = 3.73, p < .005; R
2
= .15,
F(1, 77) = 15.57, p < .01, see Table 1). Furthermore, the
addition of benevolence and integrity beliefs to the model
did not significantly contribute to explaining online pur-
chase intentions (ΔR2 = .002, F(2, 75) < 1). A stepwise
regression analysis yielded the same results. Multicollinear-
ity was not a problem (the maximum VIF = 2.58).
Web site investment and trusting beliefs. Here and else-
where, we analyzed the data with a 2 (investment) × 3
(privacy/security statement) ANCOVA, controlling for
online purchase experience and problems with the site.
Ability beliefs were higher for those who visited the high-
investment site than for those who visited the low-
investment site (Ms = 2.68 versus 2.13; F(1, 69) = 6.54, p <
.05), thus replicating the results of Study 1. No other effects
on ability beliefs were significant (Fs(1, 69) <
2.34, n.s.).
Furthermore, as we expected, Web site investment did not
affect benevolence or integrity beliefs (Fs(1, 68) < 2.29,
n.s.).
Whereas investment appears to communicate a firm’s
ability but not its benevolence and integrity, privacy/
security statements appear to communicate benevolence
and integrity but not ability: The privacy/security effect was
significant for benevolence beliefs (F(2, 69) = 3.30, p < .05)
124
Converting Web Site Visitors into Buyers / 139
and integrity beliefs (F(2, 69) = 4.03, p < .05) but not for
ability beliefs (F(2, 69) = 2.34, n.s.). Benevolence and
integrity beliefs were significantly higher among those who
read a strong statement than among those who read a weak
statement (benevolence: Ms = 3.04 versus 2.65; F(2, 68) =
2.89, p < .05 [one-tailed test]; integrity: Ms = 3.19 versus
2.66; F(2, 68) = 7.12, p < .05). These beliefs were also
higher for those who received a strong statement than for
those who received no statement (benevolence: Ms = 3.04
versus 2.44; F(2, 68) = 6.55, p < .05; integrity: Ms = 3.19
versus 2.76; F(2, 68) = 5.02, p < .05). However, the mere
presence of a privacy/security statement does not appear to
signal a firm’s benevolence and integrity; the difference
between a weak statement and no statement was not signifi-
cant at p < .05. It appears that both the presence and the
strength of the statement signal the firm’s benevolence and
integrity.
Web site investment and online purchase intentions. We
analyzed online purchase intentions with a 2 × 3 ANCOVA.
In support of H
2
, online purchase intentions were higher
among those who visited the high-investment site than
among those who visited the low-investment site (Ms =
–.16 versus –.79; F(1, 69) = 3.06, p < .05 [one-tailed test]).
A privacy/security statement did not moderate this effect
(F(2, 69) < 1). Thus, regardless of whether benevolence and
integrity signals (i.e., privacy/security statements) were pre-
sent, ability signals (i.e., Web site investment) significantly
influenced searchers’ online purchase intentions.
The only other significant effect was a privacy/security
effect (F(1, 69) = 4.95, p = .01). In support of the argument
that benevolence and integrity signals (e.g., a strong
privacy/security statement) should have little effect on
searchers’ online purchase intentions, online purchase
intentions did not differ between those who received the
strong privacy/security statement and those who received no
statement (Ms = –.36 versus .09; F(1, 46) = 1.10, n.s.).
However, online purchase intentions were lower among
those who received a weak privacy/security statement than
among those who received a strong statement (Ms = –1.15
versus –.36; F(1, 41) = 3.82, p < .05 [one-tailed test]) or no
statement (Ms = –1.15 versus .09; F(1, 49) = 8.58, p < .01).
This finding is consistent with existing research that nega-
tive cues regarding a firm’s character tend to be unexpected,
violating established norms about “business as usual”
(Garfinkel 1963). In such situations, people appear unwill-
ing to buy online from the firm.
To test whether ability beliefs mediate the investment
effect on online purchase intentions (H
3
), we added ability
beliefs as a covariate to the 2 × 3 ANCOVA. Consistent
with the requirements for mediation, ability was a signifi-
cant covariate (F(1, 68) = 8.47, p < .01), and the investment
effect became nonsignificant (F(1, 68) < 1). We replicated
this finding with the Sobel (1982) test (Goodman I test sta-
tistic = 1.93; p = .05). As in Study 1, benevolence and
integrity beliefs cannot be considered mediators, because
investment did not significantly affect these beliefs. How-
ever, it is possible that benevolence and integrity beliefs
mediate the privacy/security effect on online purchase
intentions. To test this, we added these beliefs as covariates
to the 2 × 3 ANCOVA. Inconsistent with the requirements
for mediation but in support of the prediction that searchers’
online purchase intentions are influenced by their ability
rather than by their benevolence and integrity beliefs (H
1
),
neither of these beliefs were significant (Fs(1, 67) < 2.02,
n.s.), and the privacy/security effect on online purchase
intentions remained significant (F(2, 67) = 5.75, p < .01).
Conclusions
Replicating the results of Study 1, we found that searchers’
ability beliefs, rather than their benevolence and integrity
beliefs, influenced their online purchase intentions. As a
result, ability signals (Web site investment) influenced their
online purchase intentions more than did signals of the
firm’s benevolence and integrity (the presence of a strong
privacy/security statement).
Because the objective of this research is to predict
online purchase intentions, thus far our focus has been on
people whose goals are most consistent with prepurchase
deliberation, namely, searchers. For those with this goal,
ability beliefs are a stronger driver of online purchase inten-
tions than are benevolence and integrity beliefs. However,
there may be goals that highlight the importance of a differ-
ent component of trust than ability beliefs. Specifically, for
those whose goal for visiting the site is more personal and
less outcome oriented (i.e., browsers), a different pattern of
results may emerge. Browsing is a moment-by-moment
activity rather than a search process for a specific piece of
information (Janiszewski 1998). As such, it is exploratory
(Moe 2003) and reflects recreational behavior (Hoffman
and Novak 1996). Similar to those who read a text for enter-
tainment (Rosenblatt 1978), browsers are likely focused on
what they are “living through” during their site visits. Thus,
whereas searchers tend to be more objective and outcome
oriented and thus are likely to disengage from having a per-
sonal experience with the site, browsers’ experiences are
likely more personal.
We argue that distinguishing between these goals has
important implications for the relative impact of each trust-
ing belief on online purchase intentions. Whereas searchers
focus on performance and thus base their online purchase
intentions on their ability beliefs, browsers likely have a
more personal experience with the site and thus will be
influenced by the most personal aspect of trust: benevo-
lence beliefs. Benevolence beliefs reflect consumers’beliefs
that the firm cares about their welfare and well-being (e.g.,
“I trust that the firm is concerned about my wants and
needs, even if doing so results in profit reductions”),
whereas integrity beliefs reflect beliefs about the firm’s
moral standards, regardless of how it feels about the indi-
vidual (e.g., “I trust that the firm is guided by sound busi-
ness principles and standards”). Likewise, ability beliefs
reflect beliefs about the firm’s expertise, regardless of how
it feels about the individual (e.g., “I trust that the firm has
the necessary skills to be successful”). Thus, because
browsers’ experiences are highly personal, their online pur-
chase intentions should depend on the most personal
dimension of trust (benevolence) rather than on the less per-
sonal dimensions (i.e., ability and integrity).
125
140/ Journal of Marketing,April 2006
H
4
: Browsers’ online purchase intentions depend on their
beliefs about the firm’s benevolence rather than their
beliefs about the firm’s ability or integrity.
We also propose that the distinction between searching
and browsing has important implications for the influence
of Web site investment on online purchase intentions.
Whereas prior research has demonstrated the importance of
separating signals from interpretation, because not everyone
interprets a signal in the same manner (Prabhu and Stewart
2001), it may be equally important to separate signal inter-
pretation from response because even similarly interpreted
signals may lead to meaningfully different responses. In
particular, we argue that though both searchers and
browsers will likely interpret investment as signaling the
firm’s abilities, the impact of this belief on their online pur-
chase intentions will vary. Specifically, if searchers’ online
purchase intentions are affected by their ability beliefs, abil-
ity signals (or Web site investment) should influence their
online purchase intentions. However, if browsers’ online
purchase intentions are unaffected by their ability beliefs,
ability signals (even if interpreted as such) should have rel-
atively little influence on their online purchase intentions.
Consequently, we hypothesize the following:
H
5
: Web site investment influences searchers’ but not
browsers’online purchase intentions.
Because the firm studied thus far was an unknown Internet-
only firm, an additional objective of Study 3 was to repli-
cate the results for searchers with a well-established firm
that sells a different set of products (electronics rather than
home furnishings and accessories).
Study 3
Method
Sample and design. A total of 152 undergraduate stu-
dents participated in exchange for extra course credit. We
randomly assigned participants to one of four conditions in
a 2 (investment) × 2 (goal: searching versus browsing)
experimental design.
Web site investment manipulation. In contrast to Studies
1 and 2, we manipulated investment using only technology.
Specifically, for the high-investment site, we used Macro-
media’s Shockwave technology to allow participants to
experience an online demonstration and to roll over the
product image to gather additional information about its
features. For the low-investment site, we conveyed the same
information through text and static graphics rather than
through the use of this technology.
Procedure. The procedure was similar to that used in
Study 2, with a few exceptions. Each participant was seated
at a computer terminal, which contained the instructions,
the site, and the survey. Participants were told that they
would be visiting a portion of Kodak’s site devoted to a spe-
cific model of digital camera. Those assigned to browse
were instructed to “have fun, looking at whatever you con-
sider interesting and/or entertaining.” Those assigned to
search were instructed that before doing so, they should
type two questions they have for Kodak about this digital
camera. To ensure that both groups would be attending to
information relevant to them, we did not specify what to
look for. These instructions are identical to those used in
prior research to instill a searching versus browsing goal
(Schlosser 2003). Participants then visited the site. To con-
trol for the amount of time browsers versus searchers spent
at the site and any possible impact of this on the dependent
variables, all participants viewed the site for five minutes,
which is comparable to the time imposed in prior research
(Schlosser 2003). We did not investigate privacy/security
statements in this study.
After visiting the site, participants completed the online
survey, which was identical to that used in Study 2, except
that the risk questions were replaced by the goal manipula-
tion check. Participants were asked the extent to which their
time at the site was spent looking for specific information (a
searching activity) or looking to be entertained (a browsing
activity) on a scale from 0 (“not at all”) to 5 (“a lot”). In
addition to the Web site investment items, participants rated
how informative they found the site to be on a scale from 1
(“very little”) to 7 (“a great deal”). We also made a slight
change to the trusting beliefs measure: We asked partici-
pants to focus on the trustworthiness of Kodak’s Internet
marketing department. We made this change to direct par-
ticipants’ attention toward the e-commerce side of the firm
rather than the firm in general or the brand. For example,
participants may trust the offline firm’s ability, benevolence,
and integrity, but they may not believe that its Internet mar-
keting managers share these same qualities.
Results
Manipulation checks. We analyzed the manipulation
check items with a 2 (investment) × 2 (goal) ANOVA. In
support of the goal manipulation, searchers reported spend-
ing more time searching (Ms = 2.52 versus 1.86;
F(1, 150) = 14.08, p < .01) and less time browsing (Ms =
1.31 versus 1.97; F(1, 150) = 13.80, p < .01) than did
browsers. In support of the investment manipulation, the
high-investment site was perceived as a greater investment
than the low-investment site (Ms = 4.51 versus 2.58;
F(1, 148) = 88.76, p < .01). This effect emerged for both
browsers (Ms = 4.15 versus 2.34; F(1, 74) = 20.19, p < .01)
and searchers (Ms = 4.86 versus 2.10; F(1, 74) = 85.14, p <
.01). No other effects were significant. The sites did not dif-
fer in perceived informativeness (F(1, 148) < 1).
Trusting beliefs and online purchase intentions. To test
whether ability beliefs explain searchers’ online purchase
intentions (H
1
) and whether benevolence beliefs explain
browsers’ online purchase intentions (H
4
), we conducted
separate hierarchical regression analyses for searchers and
browsers; we modeled ability beliefs as a function of online
purchase intentions (Model 1) before adding benevolence
and integrity beliefs (Model 2). As we predicted, ability
beliefs were significantly related to searchers’ online pur-
chase intentions (β = .42, t(74) = 3.96, p < .01; R2 = .18,
F(1, 74) = 15.70, p < .01; see Table 2). The addition of
benevolence and integrity beliefs to the model did not sig-
nificantly contribute to searchers’ online purchase inten-
tions (ΔR
2
= .03, F(2, 72) = 1.50, n.s.). For browsers, ability
150
Converting Web Site Visitors into Buyers / 141
TABLE 2
Study 4: Hierarchical Regression Analysis of Trusting Beliefs on Online Purchase Intentions When Risk Is
High Versus Low
A: Results from a Hierarchical Regression Analysis
Model 1
Model 2
Social Risk
Variable
ββ
R2
F for R2
ββ
ΔΔR2
F for ΔΔR2
High
Product attitudes
.60**
.53
7.91**
.58**
.02
1.80
Autotelic NFT
–.09**
–.14**
Instrumental NFT
–.03**
–.05**
Online buying
.16**
.06**
Site problems
.10**
.17**
Ability
.29**
.24**
Integrity
.18**
Low
Product attitudes
.44**
.27
2.40**
.41**
.02
1.12
Autotelic NFT
–.12**
–.10**
Instrumental NFT
.02**
.00**
Online buying
.05**
.14**
Site problems
.09**
.01**
Ability
.08**
.17**
Integrity
–.17**
B: Results from a Stepwise Regression Analysis
High
Product attitudes
.64**
.41
33.92**
.56**
.07
5.94**
Ability
.27**
Low
Product attitudes
.48**
.23
13.47**
*
p
< .10.
**
p
<
.05.
Notes: NFT = Need for touch.
beliefs alone did not significantly explain their online pur-
chase intentions (R2 = .02, F(1, 74) = 1.62, n.s.). However,
the addition of benevolence and integrity beliefs to the
model significantly contributed to browsers’ online pur-
chase intentions (ΔR2 = .10, F(2, 72) = 4.28, p < .05). Con-
sistent with H
4
, browsers’ benevolence beliefs were signifi-
cantly related to their online purchase intentions (β = .33,
t(74) = 2.70, p < .01), whereas ability and integrity beliefs
were not (t(74) < 1). Multicollinearity was not a problem in
the analyses (maximum VIFs < 1.50).
Web site investment, trusting beliefs, and online pur-
chase intentions. We analyzed each belief with a 2 × 2
ANOVA. As we expected, ability beliefs were higher
among those who visited the high-investment site than
among those who visited the low-investment site (Ms =
3.51 versus 2.99; F(1, 148) = 21.74, p < .01). Investment
did not affect benevolence and integrity beliefs
(Fs(1, 148) < 1). Furthermore, people’s goals for visiting
the site neither directly affected nor interacted with invest-
ment to influence their ability, benevolence, or integrity
beliefs (Fs(1, 148) < 2.27, n.s.). Thus, investment appears to
signal a specific trusting belief (i.e., ability rather than
benevolence or integrity), replicating the results of Studies
1 and 2 with a different set of stimuli. Moreover, this find-
ing persisted despite differences in people’s goals.
In support of H
5
, searchers had higher online purchase
intentions after visiting the high-investment site than after
visiting the low-investment site (Ms = –.03 versus –.76;
F(1, 74) = 4.92, p < .05), whereas browsers’ online purchase
intentions were unaffected by investment (Ms = –.48 versus
–.08 at the high- and low-investment sites; F(1, 74) = 1.26,
n.s.). This investment × goal interaction was significant
(F(1, 148) = 5.56, p < .05).
In addition to demonstrating that the site effect on
online purchase intentions varies across goals, we tested
whether the mediating effects of ability vary across goals
using the procedure that Baron and Kenny (1986) outline.
Specifically, we regressed online purchase intentions on
goal, site, goal × ability, and site × ability, and we found a
significant goal × ability effect (t(146) = 1.82, p < .05 [one-
tailed test]). For searchers, mediation was supported: Abil-
ity was significant (F(1, 73) = 10.45, p < 01), and the
investment effect became nonsignificant (F(1, 73) < 1). A
Sobel (1982) test supports this finding (Goodman I test sta-
tistic = 2.74, p < .01). However, for browsers, mediation
was not supported: Ability had little effect on online pur-
chase intentions (F(1, 73) = 2.64, n.s.). Furthermore,
because investment did not affect benevolence and integrity
beliefs, they cannot be considered mediators.
Conclusions
In Study 3, we examined a boundary condition for the effect
of Web site investment on online purchase intentions: con-
sumers’ goals for visiting sites. We found that the effects of
Web site investment and ability on online purchase inten-
tions are specific to searchers and do not generalize to
browsers. For browsers, the most personal component of
trust (i.e., benevolence rather than ability) influences their
123
142/ Journal of Marketing,April 2006
online purchase intentions. Consequently, although
browsers recognized that Web site investment signals abil-
ity, it had relatively little influence on their online purchase
intentions.
A remaining question is whether the findings are really
a matter of trust. Recall that trusting intentions involve risk
(Moorman, Zaltman, and Deshpandé 1992), which causes
people to consult their trusting beliefs to determine whether
to perform the trusting behavior. If the findings are driven
by trust, Web site investment should influence consumers’
online purchase intentions under conditions of risk.
Although buying online is typically risky, certain situational
factors (e.g., buying a gift for a significant versus an
insignificant other) can make buying online more or less
risky. When there is relatively little risk, buying online
should involve relatively little trust. Consequently, people
are less likely to consult their trusting beliefs when deciding
how to act. Indeed, the degree of trust necessary to influ-
ence behavior has an approximate linear relationship to the
degree of risk involved (Corritore, Kracher, and Wieden-
beck 2003; Mayer, Davis, and Schoorman 1995). Thus,
when the purchasing scenario involves relatively little risk,
ability beliefs (and signals designed to influence such
beliefs) should have relatively little effect on searchers’
online purchase intentions.
H
6
: Ability beliefs affect searchers’ online purchase intentions
only when buying involves risk.
H
7
: Web site investment affects searchers’ online purchase
intentions when buying involves risk.
Another objective for Study 4 was to test the importance
of trust in determining online purchase intentions versus
online purchase experience and a desire to examine prod-
ucts physically before purchase. For people who are moti-
vated to touch products, barriers to touch can decrease con-
fidence in product evaluation, though conveying haptic
information through text and graphics can help (Peck and
Childers 2003b). If trust is critical in determining online
purchase intentions, ability beliefs should be related to
online purchase intentions even when we account for these
other factors.
Study 4
Method
Sample and design. A total of 98 undergraduate students
participated in exchange for extra course credit. We ran-
domly assigned them to one of four conditions in a 2
(investment: high- versus low-investment site) × 2 (risk:
high versus low) factorial design. They visited the UF site
used in Studies 1 and 2.
Risk manipulation. We manipulated risk by varying
social risk. Specifically, the high-risk scenario was as fol-
lows: “Imagine that you graduated and obtained your dream
job at a firm. You are invited to your boss’ house-warming
party, which takes place in a week.” The low-risk scenario
was as follows: “Imagine that you are invited to a former
roommate’s house-warming party, which takes place in a
month.” Participants in both scenarios were told that this
person likes modern furniture and accessories, such as those
3
Moorman, Deshpandé, and Zaltman (1993) do not measure
benevolence beliefs; thus, we do not address them here.
offered at UF’s site, and that they intend to buy a home
accessory (e.g., a vase) as a house-warming gift for this
person.
To test the effectiveness of these scenarios, 41 under-
graduate students read either scenario and then rated how
nervous and concerned they would be about making this
purchase and how risky they considered this purchase on a
scale from 1 to 7 (higher numbers reflected greater risk).
We averaged these items to form a perceived risk score (α =
.76). As we expected, perceived risk was higher among
those who read the high-risk scenario than among those
who read the low-risk scenario (Ms = 4.22 versus 3.12;
F(1, 37) = 7.12, p = .01). Participants then looked at screen
captures of the high- or low-investment site and answered
the Web site investment items (α = .94; see the Appendix).
As we expected, perceived investment was higher for the
high-investment site than for the low-investment site (Ms =
5.12 versus 3.81; F(1, 37) = 14.78, p < .01). Moreover, risk
did not directly affect or moderate these investment percep-
tions (Fs(1, 37) < 1.39, n.s.).
Procedure. The procedure and survey were the same as
that used in Study 2, with a few exceptions. At the begin-
ning of the experiment, participants read either the high- or
the low-risk scenario before receiving the search instruc-
tions. Participants did not receive a privacy/security
statement.
Unlike Study 2, online purchase intentions were specific
to home accessories (the gift they were to buy). Participants
also reported their attitudes toward UF’s home accessories
to control for any individual differences in their liking of
UF’s offerings (see the Appendix). In addition, to test the
robustness of our effects related to ability versus integrity
beliefs, we replaced Mayer and Davis’s (1995) scale with
an adaptation of Moorman, Deshpandé, and Zaltman’s
(1993) ability and integrity items.3 Three items measured
ability beliefs (α = .93), and two items measured integrity
beliefs (r = .31; see the Appendix). Finally, to test the
impact of trust on online purchase intentions relative to con-
sumers’ desire to touch products before purchase, we
administered Peck and Childers’s (2003a) need for touch
(NFT) scale, which measures the need for autotelic and
instrumental touch (see the Appendix).
Results
Trusting beliefs and online purchase intentions. To test
H
6
, we conducted a hierarchical regression analysis sepa-
rately for searchers in the high- and low-risk conditions; we
modeled ability beliefs and variables believed to influence
online purchase intentions (attitudes toward home acces-
sories, both types of NFT, online purchase experience, and
site problems) as a function of online purchase intentions
(Model 1) before adding integrity beliefs (Model 2). As we
predicted, for searchers in the high-risk condition, ability
beliefs were significantly related to online purchase inten-
tions (β = .29, t(48) = 2.39, p < .05; R2 = .52, F(6, 43) =
7.91, p < .01; see Table 2). The only other significant varia-
ble was attitudes toward home accessories (β = .60, t(48) =
Documents you may be interested
Documents you may be interested
