Facebook Marketing Article

1
3
How online social ties and product-related risks inuence purchase
4
intentions: A Facebook experiment
5
6
7 Jyun-Cheng Wang
a,
Q1 , Ching-Hui Chang
b
8
a
Institute of Service Science, National Tsing-Hua University, 101 Kuang-Fu Road, Sec. 2, Hsinchu, Taiwan, ROC
9
b
Institute of Technology Management, National Tsing-Hua University, 101 Kuang-Fu Road, Sec. 2, Hsinchu, Taiwan, ROC
10
11
1 3
a r t i c l e i n f o
14 Article history:
15 Received 11 January 2012
16 Received in revised form 13 March 2013
17 Accepted 14 March 2013
18 Available online xxxx
19 Keywords:
20 Social networking sites
21 Information diagnosticity
22 Tie strength
23 Purchase intention
24 Product-related risks
25
2 6
a b s t r a c t
27 Drawing on information processing theory and the stimulusorganismresponse model, we developed
28 research hypotheses about consumers decision-making processes. Specically, we examined the effects
29 of online tie strength, perceived diagnosticity, and product-related risks on consumers purchase inten-
30 tions. We conducted a eld experiment on Facebook to test these hypotheses. We found that the product
31 information and recommendations provided by friends with whom consumers have strong ties are per-
32 ceived as having a high level of diagnosticity. The latter increases the probability that the consumers will
33 purchase the product in question. Product-related risks moderate the effect of tie strength on perceived
34 diagnosticity. For high-risk products, the information and recommendations provided by strong-tie con-
35 tacts have a greater effect on purchase intentions than the information and recommendations provided
36 by weak-tie contacts. However, we did not nd this effect for low-risk products. We discuss the implica-
37 tions of our ndings for both theory and practice.
38 2013 Published by Elsevier B.V.
39
40
41 1. Introduction
42 Consumers can obtain product information from online
43 sources, which are important to their purchasing decisions. Sev-
44 eral studies have discussed the impact of online consumer reviews
45 on the purchasing process (Chevalier and Mayzlin 2006; Dellaro-
46 cas et al. 2007; Duan et al. 2008a,b; Godes and Mayzlin 2004;
47 Koh et al. 2010; Lee et al. 2008; Liu 2006; Park and Kim 2008; Park
48 et al. 2007). As internet social networking sites (SNSs) have be-
49 come more popular, they have forged connections among internet
50 users and become important sources of information for users.
51 SNSs have been attracting more attention from both researchers
52 and marketers. In particular, researchers who have recently begun
53 studying this phenomenon regard it as a potential source of inter-
54 esting and valuable information. For example, Trusov et al. (2009)
55 studied the effect of word-of-mouth (WOM) marketing on the
56 growth of a social networking site and compared WOM marketing
57 with traditional marketing techniques. They found that WOM
58 referrals have longer carryover effects than traditional marketing
59 techniques. Kelly et al. (2010) examined the factors that inuence
60 advertising avoidance in SNSs. The researchers showed that users
61 are more likely to avoid advertising in their online social network-
62 ing environments if the following conditions are true: (1) the users
63 expect that the experience will be negative, (2) the advertising is
64 not relevant to the users, or (3) the users are skeptical about the
65 advertising message. Marketers have also begun to recognize the
66 potential of such sites as effective marketing tools. SNSs allow a
67 company to easily identify its target customers and instantly
68 distribute information to a broad range of people.
69 Marketers and researchers have realized the importance of
70 social ties with respect to customers decision-making processes
71 (Arndt 1967). The different types of social ties have different
72 impacts on purchasing decisions. For example, strong ties (friends)
73 exert more inuence over a receivers decision making than weak
74 ties (acquaintances) (Brown and Reingen 1987). However, few
75 scholars have studied SNSs because marketing oriented around
76 social networking is a relatively new phenomenon. In SNSs, users
77 interact with one another, share information, gather ideas and
78 opinions, and inuence one anothers perceptions (Centola 2010).
79 The information generated by these interactions is subsequently
80 dispersed and updated. This information then impacts consumers
81 decision-making processes. SNSs are changing existing marketing
82 practices on a fundamental level because people are relying more
83 on the members of their own networks, such as friends and peers,
84 to guide their decision making instead of relying only on authority
85 gures, experts, the mainstream media, and mass advertising
86 (Brymer 2009). This development supports Watts and Doddss
87 (2007) argument that major cascades of inuence are not driven
88 by specic opinion leaders. Therefore, marketers must consider a
89 new marketing logic based on the social networking phenomenon.
90 Scholars should examine how the online social interactions among
1567-4223/$ - see front matter 2013 Published by Elsevier B.V.
http://dx.doi.org/10.1016/j.elerap.2013.03.003
Corresponding author. Tel.: +886 3 571 5131x42247; fax: +886 3 561 6345.
E-mail address: jcwang@mx.nthu.edu.tw (J.-C. Wang).
Electronic Commerce Research and Applications xxx (2013) xxxxxx
Contents lists available at SciVerse ScienceDirect
Electronic Commerce Research and Applications
j our nal homepage: www. el sevi er . com/ l ocat e/ ecr a
ELERAP 524 No. of Pages 10, Model 5G
5 April 2013
Please cite this article in press as: Wang, J.-C., and Chang, C. H. How online social ties and product-related risks inuence purchase intentions: A Facebook
experiment. Electron. Comm. Res. Appl. (2013), http://dx.doi.org/10.1016/j.elerap.2013.03.003
91 SNS users affect purchase decision making and whether theories
92 about the inuence of social ties in the ofine context are applica-
93 ble to SNSs.
94 We develop research hypotheses about the effect of online tie
95 strength in SNSs on consumers purchase intentions. According to
96 the information processing model and the stimulusorganismre-
97 sponse (SOR) model, after consumers receive information about
98 a product, they rst process the information before purchasing the
99 product (Bettman 1979, Hoyer and MacInnis 2004, Mehrabian and
100 Russell 1975). We argue that the effect of online tie strength on
101 consumers purchase intentions is mediated by their information-
102 processing behavior. Perceived diagnosticity is adopted in this
103 study as a mediator. Perceived diagnosticity represents consumers
104 perceptions of their information sources capacity to assist them in
105 evaluating the quality of products (Jiang and Benbasat 2005, Kempf
106 and Smith 1998, Mudambi and Schuff 2010). Information sources
107 are a criterion for evaluating the product information. Therefore,
108 in our model, we contend that if consumers receive product infor-
109 mation, they rst evaluate the information source to help under-
110 stand the products quality and then decide whether to buy it. In
111 addition, product-related risks are important factors that inuence
112 how consumers evaluate products and the information sources
113 (Dowling 1999). We believe that online tie strength may interact
114 with product-related risks. Therefore, the objective of this study
115 is to understand how online tie strength, perceived diagnosticity,
116 and product-related risks affect consumers intentions to purchase
117 a product. We conducted a eld experiment on Facebook, one of
118 the most popular SNSs, to investigate the following three ques-
119 tions. Are the recommendations and information provided by dif-
120 ferent social ties perceived as having different levels of
121 diagnosticity? Do product-related risks moderate the relationship
122 between online tie strength and perceived diagnosticity? Does per-
123 ceived diagnosticity have a positive effect on purchase intentions?
124 In addition, the SNSs databases can accurately record the interac-
125 tions among SNS users and thereby create a valuable source of data
126 for social network analysis. In conducting our analysis, we attempt
127 to take advantage of the online interaction data collected from
128 websites as the basis for tie strength. Our work differs from previ-
129 ous studies, which used self-reported data gathered by surveys to
130 measure the strength of social ties (Brown and Reingen 1987,
131 Duhan et al. 1997, Frenzen and Nakamoto 1993, Wen et al.
132 2009), in that our survey alleviates concerns about the accuracy
133 of the informants retrospective reports (Bernard et al. 1984,
134 Marsden 1990).
135 This paper provides evidence regarding the effect of online tie
136 strength on purchase intentions. Our methodology used data on
137 online interactions to measure the strength of social ties. By using
138 these online interaction data, we took an alternative perspective in
139 dening tie strength, which allows us to rst examine the relation-
140 ships between online tie strength and the behavior variables. Our
141 work also demonstrates that the data collected from websites
142 can be useful to marketing companies. We developed theoretical
143 hypotheses regarding the underlying mechanisms of consumers
144 decision-making processes within the context of SNSs. Using
145 empirical evidence, we argue that online tie strength, perceived
146 diagnosticity, and product-related risks play signicant roles in
147 the decision-making process. This research is also important from
148 a managerial perspective because it shows that product-related
149 risks can be used as a practical criterion for users attempting to
150 determine from which contacts they should solicit product
151 recommendations.
152 2. Theoretical background
153 A social tie is dened as the set of social interactions between
154 two or more individuals. Previous studies have found that social
155 ties are important to individuals decision-making processes
156 (Arndt 1967). Social ties can be divided into different types. Grano-
157 vetter (1973) introduced the concept of tie strength, which refers
158 to the closeness of the relationships among people. Tie strength
159 ranges from weak to strong. If a relationship is dened as a weak
160 tie, then the individuals are only acquaintances and do not know
161 each other well. If the tie is dened as a strong tie, then the individ-
162 uals are friends who know each other well.
163 Several studies have examined the effects of tie strength on
164 consumer behavior. Brown and Reingen (1987) used retrospective
165 data collected exclusively from successful referrals to analyze
166 WOM-referral behavior. They found that weak and strong social
167 ties played different roles. Consumers generally perceived strong
168 ties to be more inuential than weak ties in their decision-making
169 processes. Frenzen and Nakamoto (1993) studied the impacts of
170 consumer decisions to transmit or withhold word-of-mouth infor-
171 mation on the ow of information in a market. They demonstrated
172 that consumers are likely to share all types of information with
173 strong-tie contacts.
174 Duhan et al. (1997) studied the factors that inuence the likeli-
175 hood that a consumer will use recommendations from strong-tie
176 sources and weak-tie sources. The researchers found that consum-
177 ers rely on both types of sources during the decision-making pro-
178 cess. That is, consumers rely on strong-tie sources if the perceived
179 difculty of the task is high. Conversely, consumers tend to rely on
180 weak-tie sources if the product is more technical. Bruyn and Lilien
181 (2008) observed the reactions of 1100 individuals to receiving an
182 unsolicited e-mail invitation from one of their acquaintances to
183 participate in a survey. The researchers found that, although the
184 characteristics of the social ties inuenced the recipients behav-
185 iors, these ties had different effects at different stages of the deci-
186 sion-making process. For example, tie strength had a positive effect
187 on awareness only during the decision-making process and trig-
188 gered the recipients interest afterward.
189 Wen et al. (2009) conducted an experiment using SNS pages to
190 investigate the effects of tie strength, endorser expertise, and prod-
191 uct type on advertising effectiveness. They found that strong-tie
192 endorsers are more effective than weak-tie endorsers at recom-
193 mending products with primarily hedonic value, regardless of the
194 strong-tie contacts expertise in the endorsed products. However,
195 for products with primarily utilitarian value, the high-expertise
196 endorsers had a stronger positive effect on the consumers pur-
197 chase intentions than the other endorsers, regardless of the tie
198 strength.
199 There is ample evidence that tie strength inuences consumer
200 behavior. Strong ties and weak ties have different effects on con-
201 sumers decision-making processes in different situations. How-
202 ever, because SNS-oriented marketing is a relatively new
203 phenomenon, few of these studies have investigated SNSs.
204 Although scholars such as Wen et al. (2009) have recently shown
205 greater interest in studying SNSs, their studies have not examined
206 the relationship between SNSs and consumers decision-making
207 processes in detail. For example, Wen et al. (2009) did not examine
208 how the consumers decision-making processes involve their cog-
209 nitive systems in processing product information and evaluating
210 its sources. Additionally, the study did not consider the effect of
211 product-related risks, which are important factors that inuence
212 consumers decision making (Dowling 1999).
213 Prior research has seldom used empirical data sources to assess
214 online social relationships. The data sources used to measure the
215 strength of social ties were essentially self-reported surveys. In Ta-
216 ble 1, we summarize previous studies with regard to the sources of
217 data on social ties. We believe that empirical data on online social
218 relationships are particularly important to understanding the pro-
219 cesses underlying e-WOM. We measured tie strength with online
220 data, including friend lists and rich interaction histories. One
2 J.-C. Wang, C.-H. Chang / Electronic Commerce Research and Applications xxx (2013) xxxxxx
5 April 2013
221 characteristic of our study is the use of online interaction data as
222 the basis of measuring tie strength. Our methods of examining so-
223 cial ties on SNSs can help address some of the limitations of previ-
224 ous research by utilizing readily available data to track online
225 interactions. In this way, our work also alleviates concerns about
226 the accuracy of the informants retrospective reports (Bernard et
227 al. 1984, Marsden 1990).
228 3. Research model and hypotheses
229 In the information processing model, consumers decision-mak-
230 ing processes involve cognitive behavior (Bettman 1979, Hoyer and
231 MacInnis 2004). Consumers are no longer passive learners who
232 produce stimulus response reactions. Instead, the model focuses
233 on how inputs from the decision environment are processed
234 through consumers cognitive systems and how they lead to nal
235 response. These processes are similar to the SOR model. The S
236 OR model posits that environmental and informational cues act
237 as stimuli that affect an individuals cognitive and affective reac-
238 tions, which, in turn, affect behavioral intentions (Mehrabian and
239 Russell 1975). Stimuli (S) may appear in different formats (Jacoby
240 2002). Past works have said that website features are important
241 stimuli for the online purchase process (Eroglu et al. 2003, Jiang
242 et al. 2010, Koufaris 2002, Parboteeah et al. 2009). The organism
243 (O) refers to an individuals cognitive systems, including the cogni-
244 tive network and schema. The response (R) represents the psycho-
245 logical reactions such as attitudinal and behavioral reactions.
246 Based on the concepts discussed above, we proposed the theo-
247 retical framework, which we called the informationcognition
248 intention model. Information is shown as an input for consumers.
249 Product information recommended by friends is a constituent of
250 this input. Before such information can affect behavioral inten-
251 tions, it must have some mental effect. Cognition is represented
252 as one of the major intermediate recommendation effects. The
253 mediating processes can be considered to be lters of the initial
254 information input. The behavioral intentions are consequently
255 generated. Following the informationcognitionintention routes,
256 we hypothesize that after consumers receive product information,
257 they rst use this information to understand and evaluate the qual-
258 ity and characteristics of the products. Afterward, depending on
259 their evaluations, the consumers develop different levels of pur-
260 chase intentions.
261 More specically, we operationalize the information as prod-
262 uct information recommended by different social ties and product-
263 related risks, the cognition as perceived diagnosticity, and the
264 intention as the purchase intentions of consumers (Fig. 1).
265 3.1. Tie strength and perceived diagnosticity
266 Measuring perceived diagnosticity is a particularly important
267 objective of the present study because helping consumers evaluate
268 products is an important goal that affects the purchasing process
269 for marketers. Perceived diagnosticity is dened as the extent to
270 which consumers consider particular aspects of their shopping
271 experiences to be helpful in evaluating products (Kempf and Smith
272 1998). Jiang and Benbasat (2005) extended this concept by show-
273 ing that perceived diagnosticity can describe the perceived ability
274 of a Web interface to convey product information that helps cus-
275 tomers evaluate the quality and performance of products. Mudam-
276 bi and Schuff (2010) applied the concept to the context of online
277 product reviews and found that depending on the type of product,
278 the characteristics of product reviews, review rating, and review
279 length have different effects on their perceived diagnosticity. How-
280 ever, Mudambi and Schuff (2010) did not study other characteris-
281 tics of the reviews, such as product information sources. Source
282 credibility theory suggests that people are more likely to be per-
283 suaded if the source is considered to be credible (Hovland et al.
284 1953, Ohanian 1990). According to this theory, information sources
285 seem to be a criterion for evaluating the truthfulness of the infor-
286 mation. We argue that different information sources could have
287 different effects on perceived diagnosticity.
288 Rogers (1995) argues that strong-tie sources are perceived as
289 more credible and trustworthy than weak-tie sources. In other
290 words, the information derived from strong-tie sources carries less
291 risk. Social exchange theory has shown that strong-tie sources are
292 likely to transmit information of higher economic value than weak-
293 tie sources (Frenzen and Nakamoto 1993, Sahlins 1972). Informa-
294 tion is more likely to be considered useful if it comes from close
295 sources, whereas the information from acquaintances is more
296 likely to be seen as less valuable or more suspicious. Consumers
297 can more easily evaluate products if the information about these
298 products is provided by strong-tie sources because consumers
299 trust these sources and the value of the information provided by
300 these sources. In other words, consumers believe that a strong-
301 tie source can help them evaluate the quality and performances
Table 1
Previous studies and sources of social ties.
Topics Source of data on social ties References
Word-of-mouth referral behavior Ofine environment Brown and Reingen (1987)
Piano classes
Self-reported surveys
Impact of tie strength on individual transmission information decisions Ofine environment Frenzen and Nakamoto (1993)
Discount scenario
The factors that inuence the likelihood of consumers using recommendation sources Ofine environment Duhan et al. (1997)
Medical services
Inferences from self-reported surveys
Multistage model of word-of-mouth inuence Online environment Bruyn and Lilien (2008)
Survey invitation message by e-mail
E-mails
Advertising effectiveness on social network sites Online environment Wen et al. (2009)
Product endorsement message on Facebook
Purchase
intentions
Tie strength
(strong vs. weak)
Perceived
diagnosticity
Product-related risk
(high vs. low)
H1
H2
H3
Information Cognition Intention
Fig. 1. Theoretical framework.
J.-C. Wang, C.-H. Chang / Electronic Commerce Research and Applications xxx (2013) xxxxxx 3
5 April 2013
302 of products. Based on these arguments, we hypothesize that per-
303 ceived diagnosticity is higher for the recommendations and infor-
304 mation provided by strong-tie sources.
305 Hypothesis 1. The recommendations and information provided by
306 strong-tie sources will have a higher level of perceived
307 diagnosticity.
308 3.2. Product-related risks
309 The more that consumers have to pay for a product, the higher
310 the perceived risk of purchasing the product will be (Bettman
311 1973, Dowling 1999). Therefore, choosing a helpful source is an
312 important concern for customers. Consumers collect and consider
313 more information about the sources trustworthiness if the product
314 risks are relatively high. Previous studies have noted that the dis-
315 tinction between high-risk products and low-risk products causes
316 consumers to utilize different psychological processes while evalu-
317 ating a product (Dowling and Staelin 1994, Payne et al. 1993). Be-
318 cause the evaluation process for high-risk products tends to be
319 more careful, consumers focus on their sources trustworthiness
320 in these situations. In contrast, consumers are less likely to care-
321 fully evaluate low-risk products. In these cases, consumers pay less
322 attention to the trustworthiness of their sources. This nding sug-
323 gests that the impact of tie strength on perceived diagnosticity is
324 likely to differ across product risk levels (i.e., source and product
325 risks interact with one another). If a products risks are high, then
326 the information provided by strong-tie sources has higher per-
327 ceived diagnosticity than that provided by weak-tie sources. If a
328 products risks are low, then the information provided by strong-
329 tie and weak-tie sources do not have signicantly different
330 diagnosticity.
331 Hypothesis 2. Product-related risks moderate the relationship
332 between online tie strength and perceived diagnosticity.
333 334
335 3.3. Perceived diagnosticity and purchase intentions
336 From the businesses point of view, the ultimate goal of market-
337 ing is persuasion (Braun-Latour and Zaltman 2006). Both market-
338 ers and academic researchers believe that purchase intentions
339 provide valuable insights into the effectiveness of marketing.
340 Therefore, we chose purchase intentions as the dependent variable,
341 which we used to measure the effectiveness of recommendations
342 derived through SNSs. Previous studies have shown that perceived
343 diagnosticity can alleviate information asymmetry through the sig-
344 nals and incentives, prevent customers to purchase low-quality
345 products (Pavlou et al. 2007), and strengthen customers con-
346 dence in their purchase decisions (Kempf and Smith 1998). If cus-
347 tomers feel that the product information is diagnostic, they are
348 certain about estimating product quality, and more condent
349 about their purchase decisions (Kempf and Smith 1998). Higher
350 perceived diagnosticity, which enables consumers to understand
351 the product more thoroughly, improves customers cognitive eval-
352 uations of the product. We suggest that if the product information
353 is positive and diagnostic, customers will be convinced that the
354 quality of the product is high, and then their purchase intentions
355 will likely be more positive. Therefore, for positive product infor-
356 mation, these improved evaluations lead them to develop higher
357 levels of purchase intentions. If the product information is negative
358 and diagnostic, customers will be convinced that the quality of the
359 product is low, and then their purchase intentions will likely be re-
360 duced. The purpose of this study is to create new marketing oppor-
361 tunities based on the social networking phenomenon. We only
362 consider the effects of positive recommendations.
363 Hypothesis 3. For positive recommendations, higher perceived
364 diagnosticity will lead to higher purchase intentions.
365 4. Research method
366 This study investigates the effects of online tie strength, per-
367 ceived diagnosticity, and product-related risks on consumers pur-
368 chase intentions. We conducted a eld experiment on Facebook to
369 test our hypotheses. Founded in 2004, Facebook has become the
370 most popular SNS. Facebook held an initial public offering (IPO)
371 on May 17, 2012. There are now more than a billion monthly active
372 users worldwide and more than half of whom use Facebook on a
373 mobile device. Facebook features an application platform that al-
374 lows developers to implement applications and integrate them
375 into the site. The popularity of these Facebook applications allowed
376 us to pursue our research in a realistic setting. We used a Facebook
377 application as the platform for this study. This Facebook applica-
378 tion offered several advantages. We could quickly and accurately
379 access information about the interactions among users on
380 Facebook. In addition, we were able to recruit a relatively large
381 sample population in a short period of time.
382 We used a 2 2 between-subjects design to investigate the
383 effects of strong vs. weak ties and high vs. low risks. Table 2 shows
384 that there were four possible combinations: strong tie and high
385 risks, strong tie and low risks, weak tie and high risks, and weak
386 tie and low risks.
387 4.1. Online tie strength
388 One critical aspect of our study was the procedure we used to
389 quantify the strength of the social ties between the individuals
390 who participated in our experiment and their friends. Previous
391 studies have measured tie strength by using many different met-
392 rics. Sometimes the measures have included recency of contact
393 (Lin et al. 1978). Other scholars have used frequency of interactions
394 as a surrogate for tie strength (Gilbert et al. 2008, Granovetter
395 1973). In the present research, we dened tie strength as the fre-
396 quency of interactions among contacts on SNSs. We used a mea-
397 sure that we called interaction frequency, which we dened as
398 the number of wall posts exchanged between two people. Our
399 Facebook application allowed us to access these data, which we
400 used to estimate the strength of the social ties between each pair
401 of friends. We dened strong-tie friends as the three people with
402 whom a participant had interacted most frequently during the past
403 three months and weak-tie friends as the three people with whom
404 the participant had interacted least during the same period.
405 Because of system limitations, we were only able to retrieve
406 Facebooks interaction data from the three months prior to the
407 experiment.
408 4.2. Products described in the experiment
409 We chose snacks as the product analyzed in our experiment be-
410 cause the use of snacks reduces the need to measure the expertise
411 level of the information source. In addition, information about
412 snacks is often found on Facebook, and most people are potential
Table 2
The four experimental treatments.
Online tie strength Product-related risks
High risks Low risks
Strong tie Treatment 1 Treatment 2
Weak tie Treatment 3 Treatment 4
5 April 2013
413 consumers of snacks. Our study examined snack products at differ-
414 ent price levels (i.e., low-priced and high-priced products). Several
415 studies have shown that product price inuences consumer behav-
416 ior. For example, customers tend to search for more information if
417 the price of a product is relatively high (Beatty and Smith 1987). Be-
418 cause product price is a stimulus for customer thinking, high-priced
419 products lead customers to think harder about their purchase
420 intentions (Wathieu and Bertini 2007). Product price is related to
421 perceived risk, to the customers perceptions of the probability of
422 an uncertain outcome and the potentially adverse consequences
423 of purchasing a product (Dowling and Staelin 1994). Higher value
424 products are perceived as riskier than products that have fewer
425 such characteristics. We operationalize low-priced products as less
426 risky products and high-priced products as risky products.
427 We asked some people who did not participate in the main
428 experiment to evaluate the snack products. Based on the results,
429 we selected three existing snack products for each price group.
430 The low-priced snacks ranged from US $2 to $4, and the high-
431 priced snacks ranged from US $30 to $50. Within each group, we
432 chose products that were similar in popularity and avor and that
433 used similar recommendation words to eliminate possible con-
434 founding effects. Each recommendation post contained a message
435 from the friend, a photo of the product, the price, and the products
436 description. The recommendation messages were all positive.
437 4.3. Experimental design
438 We asked the subjects in each group to imagine a scenario in
439 which a friend invited them to a party and asked them to bring
440 some snacks. While they were pondering which snacks to buy, they
441 noticed that their Facebook wall posts contained relevant informa-
442 tion (Fig. 3). We used two types of information to construct the
443 wall posts. First, we replicated the 10 most recent posts on the sub-
444 jects actual Facebook wall on the day before he or she participated
445 in the experiment. Second, we added three posts containing prod-
446 uct recommendations (e.g., the snack is delicious.). The contents of
447 the three recommendations were quite similar for all of the subject
448 groups. The differences lay in the type of contact who had suppos-
449 edly posted the recommendation and in the price of the product.
450 For example, for the strong-tie and high-risk group, strong-tie
451 friends recommended products with high prices (see Fig. 2).
452 4.4. Questionnaire items
453 We measured two key constructs by using questionnaire items
454 with responses on a seven-point Likert scale ranging from
455 strongly disagree to strongly agree. To measure the perceived
456 diagnosticity, we adopted the questionnaire items developed by
457 Kempf and Smith (1998) and Jiang and Benbasat (2005, 2007)
458 and modied these items to t the context of our study. In addi-
459 tion, the questionnaire items from Coyle and Thorson (2001) were
460 adopted to measure the participants purchase intentions. These
461 questions focused mainly on the subjects willingness to purchase,
462 the likelihood of purchasing, and their interest in purchasing the
463 product. The questionnaire items are listed in full in Appendix A.
464 4.5. Pilot test
465 We conducted a pilot test with 120 participants to check for
466 confounding effects and other problems that could occur during
467 the actual experiment. The respondents were asked to view exper-
468 imental Facebook pages and to complete a nal questionnaire. We
469 obtained feedback and suggestions from the subjects after they
470 had completed the pilot test. In general, the respondents in all four
471 groups thought the experimental procedure had been successfully
472 performed. Their major suggestions and recommended adjust-
473 ments involved the phrasing of the questions and the design of
474 the Facebook application.
475 4.6. Participants and procedures
476 We advertised the Facebook application by using Facebooks
477 applicationdistribution mechanism, which introduced random
478 exogenous variation through the participants. We awarded prizes
479 to the respondents as an incentive for completing the experiment.
480 Each Facebook user who chose to participate in the experiment
481 was randomly assigned to one of the four experimental groups.
482 After the Facebook users signed up to use our Facebook applica-
483 tion, we provided them with instructions that explained the proce-
484 dures. Next, we asked the subjects to provide their demographic
485 information and explained the hypothetical task. Then, the applica-
486 tion showed them a customized wall post. Finally, we asked the
487 subjects to complete a questionnaire.
488 5. Data analysis and results
489 We conducted a eld experiment by inviting randomly selected
490 users on Facebook to participate in a Facebook application created
491 for this experiment. We recruited the 420 subjects by using
492 Facebook. Among the subjects, 216 (51.4%) were female, and 204
493 (48.6%) were male. Most of the participants (78.5%) were between
494 21 and 30 years old. Most had been using Facebook for more than a
495 year (74.5%), and a majority (68.1%) checked their Facebook wall
496 messages at least once a day.
497 5.1. Manipulation checks
498 To assess our operational denitions of online tie strength and
499 product-related risks, we asked the following two questions in
500 the questionnaire: Do you think you interact often on Facebook
501 with the friends who provided recommendations? and Do you
502 think the price you saw for the snack product is high? The sub-
503 jects answers were similar to our operational denitions. Table 3
504 shows the levels of agreement. Our manipulation of online tie
505 strength and product-related risks appears to have been successful.
506 5.2. Validity of the measurement instrument
507 We conducted a factor analysis and varimax rotation for
508 the eight items in the two constructs. We performed a
509 KasierMeyerOlkin (KMO) test to determine whether the data
510 were suitable for factor analysis. A KMO value higher than 0.5 is
511 generally considered sufcient (Hinton et al. 2004). The KMO value
512 for the combined construct was 0.88. The factor loading scores for
513 all of the questionnaire items were above 0.6, which indicates that
514 the constructs were valid (Hair et al. 1998).
515 We adopted a Cronbachs a coefcient between 0.70 and 0.90 as
516 our criterion for internal consistency (Nunnally et al. 1967). The
517 Cronbachs a values for both of the constructs were more than
518 0.8 (i.e., Purchase intentions: 0.870 and Perceived diagnosticity:
519 0.914), which indicates that each item was related to its respective
520 construct.
521 5.3. Hypothesis testing
522 We performed a univariate analysis by using ANOVA to deter-
523 mine the effects of online tie strength and product-related risks
524 on the perceived diagnosticity. The results are shown in Table 4.
525 We found that online tie strength had a signicant effect
526 (F = 6.856, p < 0.01). We then conducted a t-test to compare the
527 levels of perceived diagnosticity of information from strong- and
528 weak-tie sources. We found that strong ties result in higher levels
5 April 2013
529 of perceived diagnosticity (M
strong
= 5.22 vs. M
weak
= 4.96, p < 0.05).
530 This nding conrms hypothesis H1.
531 In addition, the interaction between online tie strength and
532 product-related risks had a signicant effect (F = 4.677, p < 0.05).
533 Table 5 shows the mean and standard deviation for perceived diag-
534 nosticity, and Fig. 3 is a graphical representation of the results
535 regarding online tie strength and product-related risks. When the
536 products risk was high, the level of perceived diagnosticity was
537 higher when the product was recommended by a strong-tie source
538 than when the product was recommended by a weak-tie source
539 (M
strong
= 5.31 vs. M
weak
= 4.84, p < 0.01). In contrast, when the
540 products risk was low, the difference between the strong-tie and
541 weak-tie recommendations was not signicant (M
strong
= 5.13 vs.
542 M
weak
= 5.09, p = 0.746). This nding supports hypothesis H2.
543 We analyzed the effects of online tie strength and perceived
544 diagnosticity on purchase intentions by using a regression analysis
545 to test H3. Table 6 summarizes the results, which show that per-
546 ceived diagnosticity was positively correlated with purchase inten-
547 tions (b = 0.596, p < 0.01), fully mediated the relationship between
548 online tie strength and purchase intentions, and explained 36.3% of
549 the variance in purchase intentions. This nding supports hypoth-
550 esis H3.
551 6. Discussion and conclusions
552 SNSs can provide a marketing opportunity. Using the informa-
553 tion processing theory of consumer behavior, we conducted an
554 experiment on Facebook to investigate the effects of online tie
555 strength, perceived diagnosticity, and product-related risks on con-
556 sumers purchase intentions. Our hypotheses were supported by
557 three experimental results. First, online tie strength is an important
558 factor in consumer decision making. Consumers tend to perceive
The post on the subjects actual Facebook wall
A photo The products description and the price
1
A message from the friend on Facebook
1
A photo The products description and the price
1
A message from the friend on Facebook
1
Fig. 2. An example of Facebook wall posts customized by this study. (In the strong-tie and high-risk treatment, the subjects saw that products with high prices were
recommended by their strong-tie friends. In the strong-tie and low-risk treatment, the subjects saw that products with low prices were recommended by their strong-tie
friends. In the weak-tie and high-risk treatment, the subjects saw that products with high prices were recommended by their weak-tie friends. In the weak-tie and low-risk
treatment, the subjects saw that products with low prices were recommended by their weak-tie friends.)
5 April 2013
559 the diagnosticity of product information and recommendations to
560 be higher if this information is provided by strong-tie sources. Con-
561 sumers believe that a strong-tie source can help them understand
562 and evaluate the quality and performances of products. This nding
563 is consistent with those of previous studies. For example, Brown
564 and Reingen (1987) and Rogers (1995) found that strong ties were
565 more inuential than weak ties.
566 In contrast with previous studies, we used the frequency of
567 interactions between pairs of people on Facebook to measure the
568 strength of their social ties, whereas previous studies have mostly
569 used self-reports collected through surveys (Brown and Reingen
570 1987, Duhan et al. 1997, Frenzen and Nakamoto 1993, Wen et al.
571 2009). We did so because, although a subject may, for example,
572 answer Father in response to a question about strong ties on a
573 self-administered survey, he or she may not have a strong online
574 tie with his or her father. We used the frequency of interactions
575 among friends as an objective measure of tie strength because
576 doing so allowed us to obtain objective online data. Examining on-
577 line social relationships through SNSs can help address some of the
578 limitations in previous studies by utilizing readily available data to
579 track online interactions. This study provides empirical evidence
580 suggesting that online social relationship data are important and
581 provide an effective means of analyzing consumer behavior. Addi-
582 tionally, the evidence indicates that theories about the impact of
583 social relationships on traditional WOM communication may be
584 applicable to SNSs.
585 Second, we found that product-related risks moderate the rela-
586 tionship between online tie strength and perceived diagnosticity. A
587 moderation effect is often sought if a causal relationship between
588 an independent variable and a dependent variable is weak or not
589 found empirically (Baron and Kenny 1986, Bennett 2000, Chaplin
590 1991, Frazier et al. 2004). Results in Table 4 show that the main ef-
591 fect of product-related risks was not signicant (F = 0.136,
592 p = 0.713), and the interaction term was signicant (F = 4.677,
593 p < 0.05). Such results reinforce our argument regarding modera-
594 tion effect of product-related risks. Therefore, product-related risks
595 should be considered for studies examining the effects of social
596 relationships on consumer behavior. Product-related risks affect
597 how consumers search for and evaluate information. Overall,
598 although a strong tie is considered as more effective than a weak
599 tie, product-related risks matter. If a products risk is high, then
600 the information and recommendations provided by strong-tie
601 sources have a more signicant effect than the information and
602 recommendations provided by weak-tie sources. However, we
603 did not nd this effect when the consumers considered low-risk
604 products. As we expected, when the consumers evaluated products
605 in different risk ranges, they underwent different psychological
606 processes (Dowling and Staelin 1994, Payne et al. 1993). Extant re-
607 searches suggested that as risk increases, consumers search for
608 more and more information such as word of mouth, advertising,
609 and interpersonal sources to reduce the degree of uncertainty
610 (Dowling and Staelin 1994, Garner 1986, Gemiinden 1985). Thus,
611 in a high-risk situation consumers will be more prone to carefully
612 analyze all the information and, as a result, sources trustworthi-
613 ness may help overcome these fears and form a more condent
614 opinion about products (Zhang and Li 2006). In turn, when risk is
615 low, consumers will be less inuenced by sources since they do
616 not need these reinforcements to overcome worry about the possi-
617 bility of non-desired outcomes (Lian and Lin 2008). This result
618 sheds new light on previous research on tie strength (Wen et al.
619 2009) because previous studies did not consider the effect of prod-
620 uct risks. The present study shows that product risks moderate the
621 effect of online tie strength on perceived diagnosticity.
622 Third, the mediators of perceived diagnosticity appeared to be
623 powerful elements in explaining purchase intentions based on rec-
624 ommendation sources. Recommendation sources can inuence
625 perceived diagnosticity by determining whether a consumer will
626 purchase a product. If consumers perceive the diagnosticity of
627 the positive information that they have about a product to be high,
628 then their purchase intentions are high. Previous studies have
629 shown that perceived diagnosticity inuences consumer attitudes,
630 which then lead to purchase intentions (Jiang and Benbasat 2007).
631 In this study, we did not consider the attitudinal variables and fo-
632 cused on the direct impact of perceived diagnosticity on purchase
633 intentions. Indeed, we found that perceived diagnosticity directly
634 inuences purchase intentions. This result supports Fangs (2012)
635 ndings, which found that perceived diagnosticity directly inu-
636 ences customers transaction intention with a specic seller from
637 a website.
Fig. 3. The interaction between online tie strength and product-related risks.
Table 3
Results of the manipulation checks.
Variable Types Percent agreement
Online tie strength Strong 87.3%
Weak 76.0%
Product-related risks High 78.9%
Low 73.9%
Table 4
ANOVA test results for perceived diagnosticity.
Source F Sig.
Online tie strength 6.856 .009
**
Product-related risks .136 .713
Online tie strength Product-related risks 4.677 .031
*
*
p < 0.05.
**
p < 0.01.
Table 5
Descriptive statistics for perceived diagnosticity.
Mean (N, SD) Online tie strength
Strong Weak
Product-related risks
High 5.31 4.84
(107, 0.86) (102, 1.17)
Low 5.13 5.09
(105, 1.05) (106, 0.98)
5 April 2013
638 6.1. Contributions
639 6.1.1. Theoretical Implications
640 In this study, we developed a model for investigating the mech-
641 anisms underlying consumer decision making in the context of
642 Facebook. We provide empirical evidence showing that online tie
643 strength, perceived diagnosticity, and product risks play signicant
644 roles in the decision to accept recommendations. Past studies have
645 examined the effects of tie strength on consumer decision making.
646 We build on Wen et al.s (2009) study on the effects of tie strength,
647 endorser expertise, and product type on purchase intentions. The
648 results of their study showed that depending on the type of prod-
649 uct being considered, tie strength and endorser expertise have
650 varying effects on purchase intentions. However, their study did
651 not investigate the psychological dimensions of consumers
652 information processing behaviors. In contrast, we examined the
653 mediating effects of perceived diagnosticity on the consumer
654 decision-making process. The resulting model is more descriptive
655 than previous models because we consider how a consumers deci-
656 sion-making process involves his or her cognitive system while he
657 or she processes information about the products and sources of
658 product information (Bettman 1979, Hoyer and MacInnis 2004).
659 In addition, the results also contribute to the perceived diagnos-
660 ticity literature. We used the concept of perceived diagnosticity,
661 which represents the extent to which consumers consider particu-
662 lar shopping experiences as helpful for evaluating products (Kempf
663 and Smith 1998). This concept was extended from the concept of
664 Jiang and Benbasat (2005). They noted that perceived diagnosticity
665 reects the perceived ability of a Web interface to convey product
666 information to customers that helps them understand and evaluate
667 the quality and performance of products. Because Mudambi and
668 Schuff (2010) found that the characteristics of the review have dif-
669 ferent effects on the information diagnosticity of that review, they
670 link the concept of information diagnosticity to online reviews.
671 Previous studies have only applied the concept of perceived diag-
672 nosticity to the information derived from sellers Web interfaces
673 and online reviews (Jiang and Benbasat 2005, Mudambi and Schuff
674 2010). Our ndings extend the literature (Jiang and Benbasat 2005,
675 Kempf and Smith 1998, Mudambi and Schuff 2010) on information
676 diagnosticity by applying the concepts to recommendations and
677 information provided by SNS contacts that differ from the informa-
678 tion examined by existing studies.
679 6.1.2. Managerial implications
680 Our ndings will increase the companies understanding of the
681 role that social ties play in consumers decision-making processes.
682 This study found that social connections on SNSs can inuence
683 consumer purchasing decisions. Social ties help increase the per-
684 ceived diagnosticity of information. The results of this study can
685 be used to develop guidelines for creating more effective market-
686 ing strategies.
687 For marketing companies, our results imply that a message
688 from a strong-tie friend has a powerful inuence on consumer
689 decision making. Companies can choose appropriate users to rec-
690 ommend a specic product to their friends. In doing so, it is hoped
691 that these users enhance their friends purchase intentions. For
692 example, companies can give their new product to the target con-
693 sumers strong-tie friends and require them to recommend the
694 product to their friends. In this way, the company can help con-
695 sumers to evaluate the product and create interest among poten-
696 tial users. In many cases, offering free products to the right group
697 of people may be less expensive than traditional advertising. Our
698 results also suggest that marketers should consider different strat-
699 egies, depending on whether a products risks are high or low. For
700 high-risk products, companies should invite individuals who have
701 strong ties with the targeted consumers to recommend the compa-
702 nies products to these consumers. For low-risk products, social
703 relationships have less inuence on consumer purchasing deci-
704 sions. The recommendations from a weak-tie friend are just as
705 inuential as from a strong-tie friend. Furthermore, we found that
706 for low-risk products, perceived diagnosticity of the information
707 from both strong-tie and weak-tie friends is high (mean > 5).
708 According to such results, it would seem that SNSs are the excel-
709 lent platforms to promote low-risk products because the majority
710 of SNSs friends are weak-tie friends rather than strong-tie friends.
711 Marketers can use SNS platforms to promote low-risk products.
712 For SNSs, our results show that social information is valuable
713 data. Social information is important and provides an effective
714 way to analyze consumer behavior. Using such information, com-
715 panies can make highly targeted, relevant, and social marketing
716 strategies. We suggest that SNSs can gather data on social relation-
717 ships, such as friend lists and interaction histories, to create social
718 information-based products. Besides, SNSs can also position them-
719 selves as a platform for social commerce, allowing it to facilitate
720 social buying on a level never before seen. This could generate
721 money for the company while making the buying process easier
722 for consumers and sellers. Facebook opened its IPO in May, 2012.
723 However, the earnings of Facebook are not as expected so the stock
724 price quickly fell. Analysts worried that too many users are access-
725 ing Facebook on mobile devices now, resulting in lower growth
726 rates for ad revenues as well as reduced operating prots. There
727 are a lot of questions about the efcacy and future of Facebooks
728 ad-dominant revenue model. Facebook should monetize through
729 another business model apart from advertising. According to our
730 results, Facebook has plenty of revenue options beyond advertis-
731 ing. Facebook has access to more information about consumers
732 demographics, interests, and social networks than any other com-
733 panies. It could create the social commerce platform and sell data
734 analytics products. It is a challenge and an opportunity as well for
735 Facebook in identifying effective ways to generate more revenue
736 streams. In addition, as people are increasingly daunted by the
737 sheer amount of data available on the Facebook, the term
738 Facebook fatigue is being mentioned. SNSs should continue to
739 address the issue of information overload that some people some-
740 times experience. According to our ndings, recommendation
741 information issued from a strong-tie friend carries more weight
Table 6
Regression and mediation analysis for purchase intentions.
IV on DV IV on mediators IV on DV with proposed mediators
Purchase intention = a + b Online tie strength Perceived diagnosticity = a + b Online tie strength Purchase intention = a + b
1
Online tie strength + b
2
Percevied diagnosticity
Online tie strength Perceived diagnosticity
Beta Beta Beta Beta
.125
*
.126
*
.050 .596
**
Note: IV = Independent Variable; DV = Dependent Variable.
*
p < 0.05.
**
p < 0.01.
5 April 2013
742 than from a weak-tie friend. We suggest that SNSs should create a
743 mechanism to help consumers lter the usefulness of information
744 and to identify interesting messages from their social circles. Be-
745 cause thousands of messages appear on users review pages, most
746 of the messages may be ignored. By using social tie data and lter
747 mechanisms, consumers can nd the interesting messages quickly,
748 and companies can easily nd their target customers.
749 6.1.3. Methodological implications
750 Prior research has more often used self-reported surveys in-
751 stead of real interaction data to measure the strength of social ties
752 (Brown and Reingen 1987, Duhan et al. 1997, Frenzen and Nakam-
753 oto 1993, Wen et al. 2009). This study provides a new approach to
754 measuring online tie strength by collecting users interaction data
755 from SNSs. SNSs allow companies to easily identify target custom-
756 ers and gather large quantities of information about these consum-
757 ers (e.g., sex, education level, and age). This method provides new
758 research opportunities (Lewis et al. 2008), allows us to bypass the
759 problem of accuracy that emerges while using retrospective infor-
760 mation provided by informants (Bernard et al. 1984, Marsden
761 1990), and demonstrates that data collected through websites
762 can be useful to marketing companies. In addition, because we
763 pursued our study in a realistic Facebook environment, we were
764 better able to gauge the consumers actual reactions and opinions.
765 6.2. Limitations and future research
766 This study was somewhat limited by its experimental design.
767 The mental processes involved in assessing information diagnos-
768 ticity are part of an overall experience. Consumer decision making
769 may be affected by various factors, such as prior experience, prior
770 attitudes, product preferences, new product information and eval-
771 uations of the sources of this information. Future studies should
772 consider the effects of individual characteristics, such as product
773 preferences, product knowledge, and personal involvement. This
774 study demonstrates that perceived diagnosticity has a direct effect
775 on purchase intentions. Future research is encouraged to investi-
776 gate other possible mediators such as attitudes toward products
777 that may enhance the association between perceived diagnosticity
778 and purchase intention.
779 Additionally, this study did not examine the effects of negative
780 recommendations and information. Studying the effects of nega-
781 tive recommendations and information would likely provide addi-
782 tional insights. For example, previous studies have shown that
783 positive and negative information exerts distinct effects on con-
784 sumers behaviors (Charlett et al. 1995, Dellarocas 2003). More
785 studies are needed to explore the effect of different recommenda-
786 tions on decision-making processes within the context of SNSs. In
787 addition, we only examine the nancial risks for product-related
788 risks. We did not examine the effects on other types of risks, such
789 as psychological risks, social risks, performance risks, and physical
790 risks (Hirunyawipada and Paswan 2006, Jacoby et al. 1974). Future
791 studies could choose different product-related risks to conrm that
792 the results hold.
793 Finally, we did not distinguish whether the impacts of recom-
794 mendation on purchase intentions are derived from inuence
795 (friends induce to purchase) or homophily (friends have similar
796 backgrounds and tastes). This study focused on the effect of the en-
797 tirety of the information, including friends inuence, recommen-
798 dations, and homophily. Some studies have developed a dynamic
799 matched sample estimation framework to identify peer effects by
800 further distinguishing inuence and homophily effects in dynamic
801 networks (Aral et al. 2009). There are important implications for
802 distinguishing inuence and homophily. Companies need to devel-
803 op different marketing strategies in responding to either situation.
804 For example, if consumers purchase a product due to inuence,
805 companies might adopt friend to friend methods. If consumers
806 purchase a product due to homophily, companies would just seg-
807 ment the market based on observable preferences. Future studies
808 could emphasize in distinguishing the effect of inuence and hom-
809 ophily on purchase intentions.
810 Appendix A. Questionnaire items in constructs 811
813
814 Construct 815 Sources 816 Measurement items
817 Perceived
818 diagnosticity
819 Kempf and
820 Smith (1998)
821 and Jiang and
822 Benbasat
823 (2005, 2007)
824 The recommendations
825 helped me evaluate the
826 products.
828 helped me judge the
829 quality of the products.
831 helped me assess the
832 performances of the
833 products.
834 Purchase
835 intentions
836 Coyle and
837 Thorson
838 (2001)
839 It is likely that I will buy
840 the recommended
841 products.
842 I will purchase the
843 recommended products
844 the next time I need
845 snacks.
846 If a friend called me to
847 get my advice about
848 which snacks to buy, I
849 would advise them to
850 buy the recommended
851 products.
852 I will denitely buy the
853 recommended
854 products.
855 I have strong intentions
856 to purchase the
857 recommended
858 products.
859
860
861
862 References
863 Aral, S., Muchnika, L., and Sundararajana, A. Distinguishing inuence-based
864
contagion from homophily-driven diffusion in dynamic networks. Proceedings
865 of the National Academy of Sciences, 106, 2009, 2154421549.
866
Arndt, J. Word of Mouth Advertising: A Review of the Literature. Advertising Research
867
Foundation, New York, 1967.
868 Baron, R. M., and Kenny, D. A. The moderatormediator variable distinction in social
869
psychological research: conceptual, strategic, and statistical considerations.
870 Journal of Personality and Social Psychology, 51, 1986, 11731182.
871
Beatty, S. E., and Smith, S. M. External search effort: an investigation across several
872
product categories. Journal of Consumer Research, 14, 1987, 8395.
873 Bennett, J. A. Mediator and moderator variables in nursing research: conceptual
874
and statistical differences. Research in Nursing & Health, 23, 2000, 415
875 420.
876
Bernard, H. R., Killworth, P., Kronenfeld, D., and Sailer, L. The problem of informant
877 accuracy: the validity of retrospective data. Annual Review of Anthropology, 13,
878
1984, 495517.
879
Bettman, J. R. Perceived risk and its components: a model and empirical test. Journal
880 of Marketing Research, 10, 1973, 184190.
881
Bettman, J. R. An Iinformation Processing Theory of Consumer Choice. Addison-Wesley,
882 Reading, MA, 1979.
883
Braun-Latour, K. A., and Zaltman, G. Memory change: an intimate measure of
884
persuasion. Journal of Advertising Research, 46, 2006, 5772.
885 Brown, J. J., and Reingen, P. H. Social ties and word-of-mouth referral behavior.
886
Journal of Consumer Research, 14, 1987, 350362.
887 Bruyn, A. D., and Lilien, G. L. A multi-stage model of word-of-mouth inuence
888
through viral marketing. International Journal of Research in Marketing, 25, 2008,
889 151163.
890 Brymer, C. The Nature of Marketing: Marketing to the Swarm as Well as the Herd.
891
Palgrave Macmillan, Hampshire, Great Britain, 2009.
5 April 2013
892 Centola, D. The spread of behavior in an online social network experiment. Science,
893
329, 2010, 11941197.
894 Chaplin, W. F. The next generation in moderation research in personality
895
psychology. Journal of Personality, 59, 1991, 143178.
896 Charlett, D., Garland, R., and Marr, N. How damaging is negative word of mouth?
897
Marketing Bulletin, 6, 1995, 4250.
898
Chevalier, J. A., and Mayzlin, D. The effect of word of mouth on sales: online book
899 reviews. Journal of Marketing Research, 43, 2006, 345354.
900
Coyle, J. R., and Thorson, E. The effects of progressive levels of interactivity
901 and vividness in web marketing sites. Journal of Advertising, 30, 2001,
902
6577.
903
Dellarocas, C. The digitization of word of mouth: promise and challenges of online
904 feedback mechanisms. Management Science, 49, 2003, 14071424.
905
Dellarocas, C., Zhang, X., and Awad, N. F. Exploring the value of online product
906 reviews in forecasting sales: the case of motion pictures. Journal of Interactive
907
Marketing, 21, 2007, 2345.
908 Dowling, G. R. Perceived risk. In P. E. Earl and S. Kemp (eds.), The Elgar Companion to
909
Consumer Research and Economic Psychology, Edward Elgar, Cheltenham, UK,
910
1999, 419424.
911 Dowling, G. R., and Staelin, R. A model of perceived risk and intended risk-handling
912
activity. Journal of Consumer Research, 21, 1994, 119134.
913 Duan, W., Gub, B., and Whinston, A. B. Do online reviews matter? An empirical
914
investigation of panel data. Decision Support Systems, 45, 2008, 10071016.
915
Duan, W., Gub, B., and Whinston, A. B. The dynamics of online word-of-mouth and
916 product salesan empirical investigation of the movie industry. Journal of
917
Retailing, 84, 2008, 233242.
918 Duhan, D. F., Johnson, S. D., Wilcox, J. B., and Harrell, G. D. Inuences on consumer
919
use of word-of-mouth recommendation sources. Journal of the Academy of
920
Marketing Science, 25, 1997, 283295.
921 Eroglu, S. A., Machleit, K. A., and Davis, L. M. Empirical testing of a model of online
922
store atmospherics and shopper responses. Psychology and Marketing, 20, 2003,
923 139150.
924
Fang, Y.-H. Does online interactivity matter? Exploring the role of interactivity
925 strategies in consumer decision making. Computers in Human Behavior, 28,
926
2012, 17901804.
927
Frazier, P. A., Tix, A. P., and Baron, K. E. Testing moderator and mediator
928 effects in counselling psychology. Journal of Counseling Psychology, 51, 2004,
929
115134.
930 Frenzen, J., and Nakamoto, K. Structure, cooperation, and the ow of market
931
information. Journal of Consumer Research, 20, 1993, 360375.
932
Garner, S. J. Perceived risk and information sources in services purchasing. The Mid-
933 Atlantic Journal of Business, 24, 1986, 4959.
934
Gemiinden, H. G. Perceived risk and information search: a systematic meta-analysis
935 of the empirical evidence. International Journal of Research in Marketing, 2, 1985,
936
79100.
937 Gilbert, E., Karahalios, K., and Sandvig, C. The network in the garden: an empirical
938
analysis of social media in rural life. In 26th Annual SIGCHI Conference on Human
939
Factors in Computing Systems, Florence, Italy, ACM Press, New York, 2008, 1603
940 1612.
941
Godes, D., and Mayzlin, D. Using online conversations to study word-of-mouth
942 communication. Marketing Science, 23, 2004, 545560.
943
Granovetter, M. S. The strength of weak ties. The American Journal of Sociology, 78,
944
1973, 13601380.
945 Hair, J., Anderson, R., Tatham, R., and Black, W. Multivariate Data Analysis. Prentice-
946
Hall, Upper Saddle River, NJ, 1998.
947 Hinton, P., Brownlow, C., and McMurray, I. SPSS Explained. Routledge, New York,
948
2004.
949 Hirunyawipada, T., and Paswan, A. K. Consumer innovativeness and perceived risk:
950
implications for high technology product adoption. Journal of Consumer
951
Marketing, 23, 2006, 182198.
952 Hovland, C. I., Janis, I. L., and Kelley, H. H. Communication and Persuasion. Yale
953
University Press, New Haven, CT, 1953.
954 Hoyer, W., and MacInnis, D. Consumer Behavior. Houghton Mifin, Boston, MA, 2004.
955
Jacoby, J. Stimulusorganismresponse reconsidered: an evolutionary step in
956
modeling (consumer) behavior. Journal of Consumer Psychology, 12, 2002, 51
957 57.
958
Jacoby, J., Kaplan, L. B., and Szybillo, G. J. Components of perceived risk in product
959 purchase. Journal of Applied Psychology, 59, 1974, 287291.
960
Jiang, Z., and Benbasat, I. Virtual product experience: effects of visual and functional
961 control of products on perceived diagnosticity and ow in electronic shopping.
962
Journal of Management Information Systems, 21, 2005, 111147.
963 Jiang, Z., and Benbasat, I. Investigating the inuence of the functional mechanisms
964
of online product presentations. Information Systems Research, 18, 2007, 454
965 470.
966
Jiang, Z., Chan, J., Tan, B. C. Y., and Chua, W. S. Effects of interactivity on website
967 involvement and purchase intention. Journal of the Association for Information
968
Systems, 11, 2010, 3459.
969
Kelly, L., Kerr, G., and Grennan, J. Avoidance of advertising in social networking
970 sites: the teenage perspective. Journal of Interactive Advertising, 10, 2010, 1627.
971
Kempf, D. S., and Smith, R. E. Consumer processing of product trial and the inuence
972 of prior advertising: a structural modeling approach. Journal of Marketing
973
Research, 35, 1998, 325337.
974
Koh, N. S., Hu, N., and Clemons, E. K. Do online reviews reect a products true
975 perceived quality? An investigation of online movie reviews across cultures.
976
Electronic Commerce Research and Applications, 9, 2010, 374385.
977 Koufaris, M. Applying the technology acceptance model and ow theory to online
978
consumer behavior. Information Systems Research, 13, 2002, 205223.
979 Lee, J., Park, D.-H., and Han, I. The effect of negative online consumer reviews on
980
product attitude: an information processing view. Electronic Commerce Research
981
and Applications, 7, 2008, 341352.
982 Lewis, K., Kaufman, J., Gonzalez, M., Wimmer, A., and Christakis, N. Tastes, ties, and
983
time: a new social network dataset using Facebook.com. Social Networks, 30,
984 2008, 330342.
985
Lian, J. W., and Lin, T. M. Effects of consumer characteristics on their acceptance of
986
online shopping: comparisons among different product type. Computers in
987 Human Behavior, 24, 2008, 4865.
988
Lin, N., Dayton, P. W., and Greenwald, P. Analyzing the instrumental use of relations
989 in the context of social structure. Sociological Methods and Research, 7, 1978,
990
149166.
991
Liu, Y. Word of mouth for movies: its dynamics and impact on box ofce revenue.
992 Journal of Marketing Research, 70, 2006, 7489.
993
Marsden, P. V. Network data and measurement. Annual Review of Sociology, 16,
994 1990, 435463.
995
Mehrabian, A., and Russell, J. A. Environmental effects on afliation among
996 strangers. Humanitas, 11, 1975, 219230.
997
Mudambi, S. M., and Schuff, D. What makes a helpful online review? A study of
998
customer reviews on Amazon.com. MIS Quarterly, 34, 2010, 185200.
999 Nunnally, J., Bernstein, I., and Berge, J. Psychometric Theory. McGraw-Hill, New York,
1000
1967.
1001 Ohanian, R. Construction and validation of a scale to measure celebrity endorsers
1002
perceived expertise, trustworthiness and attractiveness. Journal of Advertising,
1003
19, 1990, 3952.
1004 Parboteeah, D. V., Valacich, J. S., and Wells, J. C. The inuence of website
1005
characteristics on a consumers urge to buy impulsively. Information Systems
1006 Research, 20, 2009, 6078.
1007
Park, D.-H., and Kim, S. The effects of consumer knowledge on message processing
1008 of electronic word-of-mouth via online consumer reviews. Electronic Commerce
1009
Research and Applications, 7, 2008, 399410.
1010
Park, D.-H., Lee, J., and Han, I. The effect of on-line consumer reviews on consumer
1011 purchasing intention: the moderating role of involvement. International Journal
1012
of Electronic Commerce, 11, 2007, 125148.
1013 Pavlou, P. A., Liang, H., and Xue, Y. Understanding and mitigating uncertainty in
1014
online exchange relationships: a principal-agent perspective. MIS Quarterly, 31,
1015
2007, 105136.
1016 Payne, J. W., Bettman, J. R., and Johnson, E. J. The Adaptive Decision Maker. Cambridge
1017
University Press, Cambridge, 1993.
1018 Rogers, E. Diffusion of Innovations. Free Press, New York, 1995.
1019
Sahlins, M. Stone Age Economics. Aldine Atherton, Chicago, 1972.
1020 Trusov, M., Bucklin, R. E., and Pauwels, K. Effects of word-of-mouth versus
1021
traditional marketing: ndings from an internet social networking site.
1022
Journal of Marketing, 73, 2009, 90102.
1023 Wathieu, L., and Bertini, M. Price as a stimulus to think: the case for willful
1024
overpricing. Marketing Science, 26, 2007, 118129.
1025 Watts, D., and Doddss, P. S. Inuentials, networks, and public opinion formation.
1026
Journal of Consumer Research, 24, 2007, 441458.
1027
Wen, C., Tan, B. C. Y., and Chang, K. T.-T. Advertising effectiveness on social network
1028 sites: An investigation of tie strength, endorser expertise and product type on
1029
consumer purchase intention. In 30th International Conference on Information
1030 Systems, Phoenix, Arizona, Association for Information Systems, Atlanta, 2009.
1031
Paper 151.
1032 Zhang, H., and Li, H. Factors affecting payment choices in online auctions: a study of
1033
eBay traders. Decision Support Systems, 42, 2006, 10781088.
1034
5 April 2013

Facebook Marketing Article

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Facebook Marketing Article

Transféré par

Droits d'auteur :

Formats disponibles

1

Vous aimerez peut-être aussi