The Dark Side of Social Media within the FMCG Market
by
Frans Gerben van den Berg
Rijksuniversiteit Groningen
Faculty of Economics and Business Administration
Msc Marketing Management and Marketing Research
Business Administration
31 December 2013
Admiraal de Ruijterweg 65-3 1057 JXAmsterdam 06-30306761 fgvandenberg@gmail.com Studentnummer: 1604511 SupervisorsManagement Summary
Nowadays, social media has become very important in the media landscape. Literature has proven that the digital innovations made it possible for consumers to share their opinion and communicate with each other. People can like or comment on each other’s messages and can be seen as a form of word-of-mouth (WoM) recommendation.
However, consumers not only share their positive product experiences through social media. They also share their negative experiences or criticism on a certain product. Negative publicity often hurts and can affect the sales of companies. Moreover, according to the literature negative information has more impact than positive information, because it is giving more weight by consumers. The research question of this study is: What is the effect of food safety criticism by social media on consumer purchase behavior of food products?
Furthermore, there are several factors that influence the way consumers respond to criticism and eventually influence the consumer purchase behavior. This study will focus on two moderating effects: Involvement and amount of likes. Moreover, this research focuses on the Fast Mover Consumer Goods industry (FMCG), the effect of criticism is narrowed by food safety issues and by social media as information source. In order to determine what the effect of the several hypotheses were an online traditional conjoint survey was conducted under 255 consumers. Next to the preferences of the different attributes the data of the survey also gave insight in demographic, social influence and involvement aspects. The results showed that criticism has indeed a negative effect on consumer purchase behavior and that an increasing number of likes leads to a stronger negative effect of criticism on consumer purchase behavior. However, the extent of involvement results in a stronger negative influence of criticism on consumer purchase behavior under high involvement than under low involvement this was not expected.
Table of Contents
Management Summary ... 2
1 Introduction ... 4
1.1 – Background of problem ... 4
1.2 – Research question ... 6
1.3 – Theoretical relevance ... 6
1.4 – Managerial relevance ... 7
1.5 – Structure of thesis ... 8
2 Theoretical Framework ... 9
2.1 -‐ Effect of criticism on consumer purchase behavior ... 9
2.2 -‐ Moderating effect of involvement ... 12
2.3 – Moderating effect of the number of Likes ... 15
2.4 – Conceptual Model ... 18 3 Research Design ... 19 3.1 – Research Method ... 19 3.2 – Data collection ... 19 4 Results ... 24 4.1 – Descriptives ... 24
4.2 – Factor Analysis: Involvement ... 27
4.3 – Conjoint Analysis ... 29
5 Conclusion and recommendations ... 37
5.1 – Conclusion ... 37
5.2 – Management Recommendations ... 39
5.3 – Limitations and future research recommendations ... 40
References ... 41
Appendix I – SPSS output ... 46
1 Introduction
1.1 – Background of problem
There has been an enormous increase in marketing expenditures in social media (Kozinents et al. 2010), it has migrated into ‘mainstream’ and marketers have taken notice of this. The percentage of companies using social media for marketing has reached 88% in 2012, up from 42% in 2008 (Williamson, 2010). In one such story, the New York Times reported that a popular blog endorsement had helped one company grow its sales from $100,000 to $4 million (Jaret, 2006). Social networking sites have become extremely popular: Facebook, for example, claims to have attracted over 800 million active members (as of fall 2011) since starting in 2004 (www.facebook.com).
Recent years have witnessed the rise of new media channels (e.g. Facebook, YouTube, Google and Twitter) which enable customers to take a more active role as market players and reach (and be reached by) almost everyone anywhere at any time (Henning-Thurau et al. 2010). Due to the rise of these new media channels consumers are able to share their opinion not only with their network but also with the entire globe. However, not only the ways consumers communicate have been changing dramatically, the same is true for the ways consumers obtain and exchange information about products (Henning-Thurau et al. 2010). Examples are product reviews, blogs or messages on social media. This new media gives consumers extensive opportunities to provide and gather information on products.
The digital innovations of the last decade made it effortless for consumers to communicate with each other and create a dialog (Deighton and Kornfeld, 2009). Consumers not only share their positive product experiences through social media with the world, they also share their negative experiences or criticism on certain products.
Although an old and popular wisdom suggests that “any publicity is good publicity”, Berger et al. (2010) found that negative information hurts product evaluation and purchase by consumers. Furthermore, several authors including Ahluwalia et al. (2000) discovered that negative information is given greater weight than positive information by consumers. In the literature it is a well-known argument and is better known as the ‘negativity effect’; the greater weighing of negative as compared with equally extreme positive information in the formation of judgments (Bunker 1996; Herr, Kardes, and Kim 1991). These findings underline that the effect of criticism will have more impact on consumers than positive news.
However, the way consumers respond to criticism through social media, when taking notice of it, will probably differ per consumer. This is due to several factors that influence the way consumers handle criticism and eventually affect the purchase behavior. The factors of influence that will be discussed in this study are involvement and the amount of likes of a message.
Involvement determines to what extent a consumer processes information (Gupta & Harris, 2010). Research revealed that variations in people’s involvement with an issue can affect how they process and respond to the issue (Greenwald and Leavitt 1984; Kardes 1988). Petty, Cacioppo and Schumann (1983) divided involvement in two degrees that differentiate consumers in the way they process and respond to argument quality; (1) High – and (2) low involvement. When the motivation to process information is high and information is carefully considered consumers are high involved. On the other hand, under low involvement the motivation to process information is low and consumers are likely to seek some heuristic cue or other ways to process information. Involvement moderates the process of selective influence of a message and in particular consumer behavior (Gamliel and Herstein, 2013).
1.2 – Research question
This study will focus on the influence of product criticism on consumer purchase behavior within the industry called ‘Fast Moving Consumer Good’ (FMCG) market. Within this business the focus will be on food products that can be purchased in retail stores (supermarkets and other wholesales). The deteriorating effects on sales of a criticizing article of the non-governmental organization Foodwatch gave several insights for this research.
As the literature showed, social media is currently one of the most important media tools for marketers and in this research the focus will be on criticism through social media. Furthermore, criticism on food products concerns a lot of different issues. In this study the focus will be on food safety of the FMCG product, Unox Runder knaks. The product has been chosen because it suits the food safety concerns and the product appeals to most types of consumers. Additionally, the recent horsemeat incident in the Netherlands made it relevant to conduct this study (NOS, 2013), as, in this case, the criticism will probably affect the producer of the product. Examples of producers within the FMCG market are Unilever, Procter & Gamble and Nestlé. The research question that will be the main focus of this master thesis is the following: What is the effect of food safety criticism by social media on consumer purchase behavior of food products?
Attention will be devoted to the following sub questions:
- What is the effect of food safety criticism on consumer behavior?
- How does the number of likes moderate the effect of product criticism on consumer purchase behavior?
- How does involvement moderate the effect of product criticism on consumer purchase behavior?
1.3 – Theoretical relevance
This research’s main relation will give insight and knowledge about the effect of criticism through social media on consumer purchase behavior. Furthermore, the two moderating effects that influence the main relation are quite interesting. These are: (1) the effect of involvement and (2) the effect of the amount of likes.
As shown in the introduction, social media is a rather interesting topic, since its tremendous increase in use and expenditures by several companies. The reason to involve social media in this study was that there has been preliminary research conducted on the success of marketing activities on social media, however little is known about factors that influence post popularity (Ryan and Zabin 2010; Shankar and Batra 2009). This study will research what the effect of the popularity of a message (amount of likes) is on the consumer purchase behavior. Furthermore, the effect of involvement and the extent of accepting criticism will be evaluated. The influence of the moderator involvement is a well-known effect in the marketing literature thanks to Petty et al. (1983). In this study involvement will be focused on food safety. The moderating effects make this research interesting and especially social media makes it relevant, since there is little research that describes the effect of this type of media on influencing consumers.
1.4 – Managerial relevance
The possible danger of criticism lies in the impact of a product-harm crisis, which is among a firm’s worst nightmares. A product-harm crisis can be defined as well-publicized events wherein products are found to be defective or even dangerous (Dawar and Pillutla, 2000). Because of the increasing complexity of products and closer research by manufacturers and policy makers as well as higher demands by consumers, product-harm crises are expected to occur ever more frequently (Dawar and Pillutla 2000), while heightened media attention will also make them more visible to the general public (Ahluwalia et al. 2000).
and (iv) a decreased cross impact of its marketing-mix instruments on the sales of competing, unaffected brands”.
However, the problem of social media is that the impact is rather small without the notice of traditional media, since traditional media still has a tremendous influence in the current media landscape. Traditional media can be seen as one of the driving forces behind enlarging social media campaigns. Stephen & Galak (2012) underline this argument by expecting that traditional media drive social media. For example, grassroots political campaigns and social movements usually begin by building WoM and social media, and then only when they have achieved a relatively high level of prominence they begin to attract the attention of the traditional media; examples are “Kony 2012” campaign and “Occupy Wall Street” movement in 2011. In general, journalists, editors and producers in traditional media often turn to social media for story ideas (Arrington, 2009).
1.5 – Structure of thesis
2 Theoretical Framework
2.1 -‐ Effect of criticism on consumer purchase behavior
The main effect that will be researched in this study is the influence of criticism on a FMCG food product on consumer purchase behavior as mentioned in the introduction. Criticism is focused on food safety in this research. First, the different categories in food safety issues will be described. Thereafter the different effects of criticism from the literature will be presented. In conclusion, the hypothesis will be presented and summarized.
The focus of criticism will be on food safety and there are several food safety issues. Food safety continues to be a concern of consumers and a focal point of the food industry (Brewer & Rojas, 2007). There are different issues in the food industry that can be identified. Brewer and Rojas (2007) evaluated consumer attitudes regarding food safety and came up with four food safety categories:
- “Chemical issues” category
Consumers in this category are most concerned with pesticide residues, hormones in meat, poultry and preservatives.
- “Health issues” category
Consumers in this category are most concerned with fat and cholesterol content, but also with calorie and carbohydrate content.
- “Microbiological issues” category
Consumers in this category are most concerned about restaurant sanitation and meat being thoroughly cooked, followed by microbiological contamination and mad cow disease in beef.
- “Regulatory issues” category
Consumers were very concerned about the inspection of imported foods and restaurant sanitation.
Beforehand, there are several important factors in evaluating negative information that influence the purchase behavior. “The credibility of the information channel, the severity of the problem, the source of the criticizing message, the probability that the problem occurs, the veracity of the negative claim and the culpability of the responsible company” are all important for consumers in evaluating negative information (Romeo, Weinberger and Antes, 1994). Most consumers consider publicity as a relatively credible source of information and value it as more influential than other marketing communications (Bond and Kirshenbaum, 1998). Considering criticism or negative information it is argued that negative information attracts more attention by consumers than positive information (Fiske, 1980). This negative effect of criticism, as mentioned in the introduction, is a well-known and important finding in the literature known as: ‘the negativity effect’. This theory argues that consumers place more weight on negative than positive information in forming overall evaluations of a target (Fiske 1980; Skowronski and Carlston 1989). Importantly, this effect has been found in both person perception as product evaluation context (Herr, Kardes and Kim, 1991).
Several different authors found that negative information, publicity or criticism on a product has a mostly negative influence on the purchase behavior of consumers. Negative reviews, messages or rumors hurt product evaluations, brand evaluation and reduce purchase likelihood of a consumer (Tybout et al. 1981, Wyatt & Badger 1984, Huang & Chen 2006). This can affect business by deteriorating a firm’s net present value and sales (Goldenberg et al. 2007, Reinstein and Snyder 2005). An example of how negative information can heavily influence sales of not only a product but also an industry is Coca Cola. In 1999 the Coca Cola Corporation withdrew 30 million cans and bottles in the Benelux and France following a scare in Belgium (Guardian, 1999). The cause was that nearly 100 people suffered attacks of nausea and stomach cramps after drinking Coke. The reason was a bad quality batch of Coca Cola bottles. Unfortunately, the whole industry experienced inconvenience of this negative information.
review of a book is more powerful in influencing the sales negatively than a positive review is in influencing the sales positively (Chevalier & Mayzlin, 2006).
On the other hand, negative publicity can have a positive effect on consumer purchase behavior, although it will depend on the existing product awareness (Berger et al., 2010). Although negative publicity hurt products that already have broad awareness it can help unknown products for both purchase likelihood and actual sales.
In general consumers have finite attention and therefore will not be aware of every (criticizing) message about a product, movie or cd (Berger et al., 2010) and therefore consumers evaluate information by several factors. In addition, consumers may have known about a product, but information varies its accessibility (Higgins and King, 1981) and is less likely to affect consumer purchase behavior if the product is not top of mind.
These insights demonstrate that criticism on a product can have a devastating influence on the business performance and will have an enormous impact on the consumer purchase decision. Considering the influence of food safety criticism on a product which influence the purchase behavior of consumer, the following hypotheses is proposed:
2.2 -‐ Moderating effect of involvement
In this paragraph involvement will be explained. Second, involvement will be placed into the context of criticism on food safety. In addition, the different attitude scales of food safety will be described and adjacent to this the connection with the Elaboration Likelihood Model (ELM) will be discussed.
As described before, it is expected that involvement moderates the effect of criticism on consumer purchase behavior by influencing to what extent consumers process criticism on a product. The amount of involvement of a consumer determines to what extent the information is processed and considered valuable for the purchase decision of a consumer. According to the ELM the extent of involvement determines the route of interpreting information. Consumers can either be high- or low involved with a particular issue.
The ELM of attitude change has different routes of inducing persuasion and may work best depending on whether the elaboration likelihood of the communications situation is high or low (Petty and Ciaoppo, 1981). There are two different routes that can be distinctive in to what extent consumers interpret information, (1) the central route and (2) the peripheral route (Petty et al., 1983).
Consumers under high involvement conditions follow the central route and use their knowledge of an issue or product to evaluate the relevant arguments that are presented to them and their attitudes are formed by this information processing activity (Petty et al., 1983). Consumers that follow the central route carefully consider the information that is central to the true merits of their attitude. Approaches following this route emphasize factors such as:
- Mental justification of attitude inconsistent behavior.
- Understand, learn and reserve issue- or product-relevant information. - Nature of a person’s typical response to external communications.
- The way a consumer combines and integrates issue- or product-relevant beliefs into an overall reaction.
On the other hand, consumers under low involvement use the peripheral route and their attitudes are based on a more superficial analysis of information. Moreover, they are not focusing on the key message arguments that are used to influence attitudes and appear to be influenced by simple acceptance and rejection cues in the persuasion context and less influenced by argument quality (Petty et al., 1983). Consumers attitude changes according to the peripheral route do not occur due to personal considerations of the pros and cons of an issue, but because the issue is associated with positive or negative cues. Examples of peripheral cues are attractive sources, pictures or the use of humor.
A consumer may reject an argument simply because the position presented is valued as too extreme. These cues and inferences can shape the attitudes or allow a low involved consumer to decide what influence it will have on their behavior without the need for engaging in any extensive thought about issue- or product-relevant arguments (Petty et al., 1983). Instead of diligently considering the relevant arguments of a particular issue a consumer following the peripheral route may accept an argument simply because it was presented during a pleasant lunch or because the source is an expert. Consumers that follow the peripheral route are low involved by personal relevance, perform less informational searches and are mostly persuaded by peripheral cues. Attitude changes under the peripheral route or low involvement are relatively temporary and unpredictable of behavior (Petty et al., 1983). Since these attitudes are temporary or unpredictable, these consumers can be easily regained because these consumers make their decisions based on peripheral cues and not on well-considered decision based on relevant information.
In short, consumers with low involvement in an issue lack knowledge to make an informed decision by argument quality.
Involvement in this study is focused on food safety. Consumers who are concerned about food safety issues will either reject food or become willing to pay more to avoid the specific food safety concerns (Shin et al., 1992). Considering the two different attitudes toward food safety and place these attitudes in the ELM in the framework, two routes can be distinguished:
(1) Price-sensitive consumers are low involved, because their main driver is the price cue regarding food safety. The assumption can be made that these consumers are not thoroughly informed about the food safety issues.
(2) Safety-sensitive consumers are highly involved with food safety and are probably thoroughly informed about the food issues, because they studied the issue.
However, social media recommendations seem to limit search and consideration efforts of consumers with lower motivation to process information (low involvement) (Gupta & Harris, 2010). Consumers with low motivation are likely to seek some cue or other way to minimize their cognitive effort to process information. A recommendation on social media (WoM) can be a simple decision making cue influencing the beliefs about the product (e.g. “the criticized product is horrible”) (Gupta & Harris, 2010). The authors also argue that when the motivation to process is high, recommendations on social media (WoM) serve as an additional argument. However, the persuasion of a social media message is a joint function of the recipients’ involvement in the issue and the communicator ’s credibility (Hass, 1981). In conclusion, involvement moderates the effect of criticism on consumer’s purchase behavior by influencing the extent of information that consumers will process information and therefore the way they react on criticism. High involved consumers are very well informed about a particular issue and make decisions based on the argument quality of a comment (Petty et al., 1983). On the other hand, low-involved consumers are not well informed and make decisions based on peripheral cues, for instance a recommendation on social media. Considering these different routes the following hypotheses is proposed:
2.3 – Moderating effect of the number of Likes
In this paragraph the moderating effect of likes on the effect of criticism on consumer purchase behavior will be discussed. First the social networking sites will be explained to clarify the expanding of messages. Second, the enlarging effect of social media, which influences the importance of the message, will be construed. Finally, the influence of the amount of likes on consumer opinions will be discussed.
Social networking sites can be best described as networks of friends for social or professional interactions (Trusov, Bucklin and Pauwels, 2009). As noted in the introduction social media has gained importance in the current media landscape and the emergence of this type of media made it possible for consumers to communicate with other consumers who share their interest and involvement of a certain topic (Mangold & Faulds, 2009, Kozinents, 1999). Since the evolvement and development of social media, the influence has a considerable impact on the consumer purchase behavior. Consumers make their opinion public by liking or commenting on a post and therefore, liking and commenting can be seen as a similar effect as word-of-mouth (WoM) communication (de Vries et al. 2012). Moreover, recommendations (WoM) on social media influence product consideration and product choice (Gupta & Harris, 2010). Negative WoM communication, presented directly before or after respondents have seen a message, reduces purchase intentions (Smith and Vogt, 1995). Other research has shown that WoM communication of social networking site users significantly influences new social media users (Trusov, Bucklin, and Pauwels 2009). Considering these insights the number of likes generates WoM communication and influences the popularity of the message. In other words, the amount of likes moderates the effect of criticism.
many-to-many communication (Goldfarb and Tucker, 2011; Hoffman and Novak, 1996). The power of social media lies within the interactivity, because consumers can share their experiences about products with or without permission of the firm in question (Mangold and Faulds, 2009). Beirut (2009) concludes that some consumers see social media as a way of making their message heard and impacting certain problems, such as environmental problems. Consumers with criticism on a product can also share their experiences and let their message be heard.
frequently and exchange more information than consumers with a weak tie relationship (Brown & Reingen, 1987). Strong tie consumers will have more influence on consumers. Source credibility mainly depends on the source expertise and source bias of an information source (Buda & Zhang, 2000; Birnbaum & Stegner, 1979).
In conclusion, the amount of likes enhances the negative information by increasing the scope of the audience. The reason is that social media is a medium with interactivity. Furthermore, the amount of likes has a positive effect on influencing the purchase behavior because likes can be interpreted as WoM communication and research have indicated that communicating a message by WoM is superior to traditional marketing methods. Therefore the following hypothesis is proposed:
2.4 – Conceptual Model
The conceptual model indicates the relations that will be researched in this study. The hypotheses are mentioned in the conceptual model with the expected effect of the relation between the several variables.
Figure 1 – Conceptual Model
H1 – Criticism on a food product through social media will have a negative impact on the purchase behavior of consumers
3 Research Design
3.1 – Research Method
The aim of this study is to understand the influence of criticism on consumer’s purchase behavior and how the moderators (1) involvement and (2) amount of likes influence the main effect. The method to collect data is a traditional conjoint analysis. First, a factor analysis, using common factor analysis, will be performed in order to assess one involvement variable towards food safety. Second, a conjoint analysis was designed to understand how respondents develop preference by calculating how much each attribute influences the purchase behavior (Hair et al. 2010). To calculate the utility scores of the attribute a regression analysis will be performed. Third, the willingness to pay will be calculated. Finally, segments/classes will be made with Latent Class to create a profile of the different type of consumers.
With these methods the following objectives will be analyzed in order to detect: • Involvement variable will be created
• Utility scores of the attributes will be identified • Willingness to pay will be calculated.
• Classes will be described. 3.2 – Data collection
A traditional conjoint analysis was performed using a factorial design approach, which means that the respondent had to evaluate all possible profiles. A respondent had to evaluate profiles constructed with selected levels from each attribute (see levels in table 3.1). This format is known as a full-profile approach.
Attribute Attribute level
Price € 0,85
€ 1,25
€ 1,70
Social Media 10 likes
10.000 likes
Criticism None
Moderate criticism
Heavy criticism
The respondents rated their preference of every profile on a 10-point scale (10 = most likely) when they were asked how likely it would be that they purchase the Unox Runder Knaks. The reason to choose this product was that it is quite neutral or put in other words not segment specific. The survey was conducted under 295 Dutch consumers, which were selected with the non-probability sampling method. Specifically, a judgment sample was used and students, colleagues, family and friends received an email or Facebook message with the survey-link. Normally a sample of 200 is sufficient (Hair et al., 2010).
The survey contained 18 choice sets, because there are two attributes with three levels and one attribute with two levels, which lead to 18 choice sets (3x3x2). The survey contained several demographic -, social influence - and involvement questions. Each choice set consisted of three attributes that changed consistently until all the eighteen profiles where shown (see table 3.1).
Three price levels were constructed in order to calculate the willingness to pay and to discover the influence of price-offs in relation to the purchase behavior. The three different price levels are constructed based on several price offs; €1.70 is the actual price, €1.25 (approximately 25% discount) and €0.85 represents fifty percent discount. The two levels of social media are quite excessive. The reason is that a message is either very popular (almost like a viral campaign) or not that popular. Since there is nothing in between these levels, the levels of 10 and 10.000 likes were constructed.
Moderate criticism
E250 is a preservative that is mainly used in meat and meat products and also in Unox Runderknaks. E250 is a difficult case: It prevents health risks and also produces health risks. It prevents botulism, which is a severe form of food poisoning. On the other hand, it does contain nitrite, which in combination with other substances can form nitrosamines in your body. Nitrosamines are possible carcinogens. This effect has been observed in animal studies, but it is difficult to prove on humans. Unox processed E250 in his expensive Runderknaks cans despite the possible consequences of this E number. Full article
Heavy criticism
The Unox Runderknaks with '100% pure quality meat’ consists partly of turkey offal. Not just turkey, but the last bit of meat from mechanically scraped of their carcasses. Unox processes this cheap 'mechanically separated meat' in his expensive Runderknaks. Full article
Involvement – and social influence scales
The amount of involvement was measured with several constructs. Respondents had to indicate, on a 7-point Likert scale ranging from “totally disagree” to “totally agree”, to what extent they pay attention to quality reassurance attributes while purchasing products. In order to detect the amount of involvement in food safety a scale was developed based on insights from the food literature.
High-involved food safety consumers’ pay more attention to food during their purchases and are therefore better able to discriminate between food products (Bell and Marshall, 2003). Furthermore, high involved food safety consumers demand intangible quality attributes, such as quality assurance labels (Mcearchern & Schroder, 2002). However, low involved food safety consumers are focused on tangible attributes, like price, whereas high-involved consumers additionally seek quality reassurance (Verbeke & Vackier, 2003). The quality reassurance attributes that are used in this study are:
- ‘Beter leven’ label – This label shows if the animal lived in an acceptable environment before it was butchered.
- Ingredient list
Respondents scoring high on these elements are high involved with food safety, when the respondents score low they will automatically be low involved with food safety.
The susceptibility to influence was also measured on a 7-point Likert (interval) scale ranging from “totally insensitive” (1) to “totally sensitive” (7). The respondent had different scales that they had to rate with the above-mentioned ranges.
In order to detect the influence of statements on a person’s behavior a survey question was developed based on social influence. The respondents where asked to what extent they were sensitive to well-defined statements of others on social media. According to the social influence literature there are two types of influences have been identified: informational and normative (Deutsch & Gerard, 1955).
Informational influence is the desire to make informed decisions and therefore search for reliable sources (Bearden & Etzel, 1982). Or an informational social influence can be defined as “an influence to accept information obtained from another as evidence about reality” (Deutrsch & Gerard, 1955).
Normative influence is if it will be accepted by others and thereby an influence to conform with the expectations of others. (Deutsch & Gerard, 1955).
Reflecting the literature on the following influence alternatives that are formulated:
- Statements from experts Informational influence
- Statements from friends Normative influence
- Statements from employees/companies Informational influence
- Statements from NGO’s Informational influence
The questionnaire also includes several relevant socio- demographic characteristics like, age, gender, education, income and type of residence.
Therefore, there are questions in the survey to indicate employees in the FMCG industry and the responsibility of the groceries.
Lastly, the willingness to pay (WTP) denotes the maximum amount of money a consumer is willing to spend for a product exposed to criticism. To calculate the willingness to pay the output from the regression analysis must indicate the utility of the actual price. Therefore, the actual price of the evaluated profile will be added as a variable in the dataset. In addition, the utility of criticism is needed to calculate the WTP for a product that is criticized. In order to calculate the willingness to pay for a consumer the following equation will be used:
𝑊𝑇𝑃 = 𝑢𝑡𝑖𝑙𝑖𝑡𝑦/𝑙𝑒𝑣𝑒𝑙
4 Results
4.1 – Descriptives
In order to perform the analyses correctly the dataset had to be representative and therefore the first step was to clean the dataset. In total, 295 respondents completed the survey. Only a few contained missing values (5) and therefore were deleted from the dataset. Next to the missing values 14 respondents gave comments that they never bought Unox Runder Knaks or similar products and therefore were not able to finish the questionnaire without bias. Approximately 20 respondents rated the several profile cards with the same rate each time and were typified as outliers. Outliers are observations with a unique combination of characteristics and are identified as distinctly different from the other observations (Hair et al. 2010). Therefore it is assumed that these respondents did not thoroughly read the introduction of the analysis or they evaluated the questionnaire as time consuming. That was a common remark with the outliers. The survey was briefly explained in the introduction and respondents were instructed to only pay attention to the attributes that where shown to them. In order to detect these outliers, the answers were ranked using excel. The data of these respondents were deleted from the sample and eventually the sample was reduced to 255 respondents due to the outliers.
The next step was to explore the demographics of the dataset in order to analyze the respondents and better understand them. Of the remaining 255 respondents there were 157 (61,6%) men and 98 (38,4%) women. Furthermore, the average age of the sample is approximately 36 years old. Considering the age per gender, women are slightly younger than the average age, namely 32 years. Where men are on average 38 years.
Figure 4.1 – Age of respondents
The most frequent age of a respondent is 25 years old. The absolute majority of the sample (76,1%) lives in cities and the remaining respondents lives either in villages (21,2%) or on the countryside (2,7%). The question was asked in order to detect a relation between the response of consumers on criticism and their type of residence. 71% of the respondents are responsible for the groceries in their household and 12,9% of the respondents is working in the FMCG industry.
Regarding the level of education, 135 respondents (52,9%) finished a University education and 61 respondents (23,9%) finished a HBO education. Considering the income of the respondents the majority have an income of less than €15.000 a year (39,6%). The overview of the level of education and the income scales are presented in table 4.2. Frequency Percentage Education Mavo 2 0,8 Havo 9 3,5 VWO 41 16,1 MBO 6 2,4 HBO 61 23,9 WO 136 52,9 Income < €15.000 101 39,6 €15.000-€30.000 25 9,8 €30.000-€45.000 37 14,5 €45.000-€60.000 28 11,0 €60.000-€90.000 35 13,7 > €90.000 29 11,4 Residence City 194 76,1 Village 54 21,2 Countryside 7 2,7 FMCG yes 33 12,9 no 222 87,1 Groceries yes 181 71 no 74 29
Table 4.2 – Demographics: Education & Income
the effect of likes the daily time spent on Facebook was conducted. As is shown in figure 4.3 most respondents (26,3%) use Facebook less than 15 minutes a day. However, the graph shows that the use of this type of social media is rather equally divided between less than 15 minutes, 15-30 minutes and 30-60 minutes a day.
Figure 4.3 – Social media usage
The last characteristic is the social influence of several sources. The graph shows that people are very sensitive for statements from experts (informational influence) and moderately insensitive for statements from employees or NGOs.
Figure 4.4 – Degree of influence of arguments from different sources
0 5 10 15 20 25 30 No Facebook account <15 min
per day 15-‐30 min per day 30-‐60 min per day 60-‐120 min per day >120 min per day
4.2 – Factor Analysis: Involvement
In order to define the underlying structure among the several variables that indicate the differences in involvement a factor analysis is performed. Factor analysis is a technique to detect interrelationships or correlations among several variables by distinctively defining sets of variables that are interrelated and can be interpreted as one construct known as factors (Hair et al., 2010).
Factor analysis is a method that can be used for data reduction or data summarization (Hair et al., 2010). In this study it is performed to summarize the data, which means that it makes the identification of the underlying dimensions. Thus, estimates of the factors and the contributions of each variable to the factors (termed loadings). The most appropriate analysis for data summarization is a common factor analysis. In order to test if the several questions measured the correct subject a reliability test was conducted. The Cronbach’s Alpha had a value of 0.708, which indicates that there is sufficient consistency between the several questions about involvement and therefore the quality reassurance attribute questions are valued as reliable.
First, the Kaiser-Meyer-Olkin measure of sampling adequacy consider whether the partial correlations among variables are small, the value (0.577) in table 4.5 indicates that it is mediocre related. With the Kaiser-Meyer-Olkin measure a value of 1 indicates a totally related relationship, which is bad. The Bartlett’s test of sphericity determines the appropriateness of a factor analysis, considering that it is significant it can be performed.
KMO & Bartlett’s Test of Sphericity
KMO Measure of Sampling Adequacy 0.577
Bartlett’s Test of Sphericity Approx. Chi-square 4098,492
df 3
Sig. .000
Table 4. 5 – Appropriateness of the Factor Analysis
# of Factors Initial Eigenvalues Eigen value Variance % Cumulative % 1 1.937 64.567 64.567 2 0.798 26.612 91.179 3 0.265 8.821 100.00
Table 4.6 – Outcome of Common Factor Analysis
Normally, five criteria are used to determine the number of factors. However, the criteria that the factors have to explain at least 5% of variance is not applicable in this analysis because each factor explains at least five percent (Hair et al., 2010). Therefore, the focus in this case will be on (1) the eigenvalue and (2) the cumulative variance. The Eigenvalue must exceed the value of 1 in order to be significant and relevant. The cumulative percentage should at least be 60%. Considering these criteria and the literature as presented in the research method, the quality reassurance labels can be summarized to one factor: Involvement. As the results show (table 4.6) one factor should be chosen.
Factor 1
Involvement ‘Vinkje’ label 0.897 Involvement ‘Beter leven’ label 0.888
Ingredient list 0.586
Table 4.7 –Loadings Involvement
Loadings that exceed 0.3 are significant and that exceed 0.5 are important. Reflecting this criteria in the loading table 4.7 factor 1 covers ‘Het Vinkje’-label, ‘Beter Leven’-label and ingredient list and can be Leven’-labeled as involvement in food safety
4.3 – Conjoint Analysis
To analyze the effect of criticism on consumer purchase behavior a traditional conjoint analysis was performed. The respondents had to evaluate 18 profiles and rate the likelihood that they would purchase the product. To calculate the partworths of the different attribute levels an ordinary least squares regression was performed. The dependent variable is the rating of the consumer on the likelihood that a consumer purchase the product.. In order to perform this analysis dummy variables were used. The base case that was used to for the dummy variables is shown in table 4.8.
Base case
Price € 0.85 Criticism None Likes 10
Table 4.8 – Base case in regression analysis
First, a regression was performed with the bias variables FMCG and groceries to detect if one of these groups could be valued as outliers. The results of the respondents who worked in the FMCG industry were quite different compared with not FMCG employees. The results are shown in table 4.9.
FMCG: Yes FMCG: No Groceries: Yes Groceries: No
Attributes B sig B sig B sig B sig
Constant 7.433 .000 6.188 .000 6.483 .000 6.023 .000 Price €1,20 -.611 .003 -.504 .000 -.576 .000 -.374 .007 Price €1,70 -1.263 .000 -1.036 .000 -1.133 .000 -.901 .000 Criticism Moderate -2.722 .000 -2.017 .000 -2.183 .000 -1.923 .000 Criticism Heavy -2.500 .000 -2.496 .000 -2.526 .000 -2.426 .000 Likes 10.0000 -.091 .588 .224 .001 .174 .016 .204 .073 R² =0.302 F=0.000 N=33 R² =0.245 F=0.000 N=222 R² =0.258 F=0.000 N=181 R² =0.224 F=0.000 N=74 Dependent Variable – Ratings of respondents
Table 4.9 – Coefficient table Aggregate Conjoint Analysis: Bias variables
consistent with the criticism outcome, because these respondents evaluate criticism as a negative influence on purchase behavior despite the amount of likes.
Furthermore, the respondents, which are responsible for the groceries, find price more important in order to purchase Unox Runder Knaks than the respondents, which are not responsible for the groceries.
Considering the results none of these groups will be excluded from the analysis, because the FMCG employees underline heterogeneity within the sample.
Second, a regression was performed on aggregate level with the following results shown in table 4.10:
Attribute Beta P-value
Constant 6.349 0.000 Price €1,20 -0.518 0.000 Price €1,70 -1.065 0.000 Criticism Moderate -2.108 0.000 Criticism Heavy -2.497 0.000 Likes 10.000 0.183 0.003 R² =0.248 F=0.000
Dependent Variable – Ratings of respondents
Table 4.10 – Coefficient table Aggregate Conjoint Analysis
Considering the predictive accuracy of the model, the R², is rather low. Since there are 255 respondents with a lot of variance in their likelihood to purchase the product it is therefore likely that the R² is not high.
However, the significant level of each attribute is more important. The table also shows that all attributes are highly significant. It proves that criticism has an extremely powerful negative effect on the purchase behavior of the consumer. Although, there is a difference between the moderate and the heavy criticism, the difference between the negative effects is smaller than expected considering the pretest. Though, there is a difference in negative effect on the consumer purchase behavior.
criticism has the highest importance (67%) for consumers in deciding to purchase the product (see table 4.11).
Relative importance per attribute
Price 28%
Criticism 67%
Likes 5%
Table 4.11 – Relative importance of the attributes
In order to calculate the willingness to pay for the Unox Runderknaks while consumers accept the heavy criticism the variable actual price had to be created. A conjoint analysis with the actual price, instead of the price dummy variables, was performed. The R² outcome was exactly the same as is shown in table 4.12.
In order to determine what the influence of the moderators: (1) Involvement and (2) amount of likes is on the effect of criticism on the purchase behavior several variables were created. The moderating variables were both multiplied with moderate criticism and heavy criticism, because of their influence on the effect of criticism on the consumer purchase behavior. This resulted in the variables presented in table 4.12. In addition, the interaction between price and income was added to the interaction model.
Involvement negatively influences the purchase behavior of consumers (β= -0.610 and β= -0,484), because both p-values are significant (0.000 and 0.001). Moreover, the Beta’s of the moderating variable indicate that an increasing extent of involvement in food safety has a stronger negative influence on the purchase behavior when a consumer is exposed to criticism. In other words, criticism exposed to high involved food safety consumers will result in declining purchase of the product. In addition, moderate criticism has slightly more impact (β = -.610) than heavy criticism (β = -.484) on involved consumers. Reflecting these outcomes on the formulated hypothesis 2, the hypothesis is rejected. In conclusion, high-involved food safety consumers will be more influenced by criticism than low-involved consumers.
negative influence of criticism on the purchase behavior of consumers. Therefore Hypothesis 3 is supported.
The p-value (.952) of the interaction effect between price and income clearly indicates that the relation is not significant. There is no relation between the income and the price of the product on the purchase behavior of consumers.
Attributes Beta P-value
Main Effect Model
Constant 7.408 0.000 Price Actual -1.253 0.000 Criticism Moderate -2.108 0.000 Criticism Heavy -2.497 0.000 Likes 10.000 0.183 0.003 R² =0.248 F=0.000 Interaction Model Constant 7.226 0.000 Price Actual -1.250 0.000 Criticism Moderate -1,803 0.000 Criticism Heavy -2.255 0.000 Likes 10.000 .548 0.000 #Likes_CriticismM -0.188 0.001 #Likes_CriticismH -0.255 0.000 Involvement_CriticismM -0.610 0.000 Involvement_CriticismH -0.484 0.001 Price_Income -.001 0.952 R² =0.256 F=0.000
Dependent Variable – Ratings of respondent
Table 4.12 – Coefficient table Aggregate Conjoint Analysis Main effects & Interaction effects
In order to detect under what price consumers are willing to buy products with heavy criticism the willingness to pay for Unox Runder Knaks with heavy criticism the following will be calculated. The following equation is used:
𝑊𝑇𝑃 = 𝑢𝑡𝑖𝑙𝑖𝑡𝑦/𝑙𝑒𝑣𝑒𝑙
𝑢𝑡𝑙𝑖𝑡𝑖𝑡𝑦/𝑝𝑟𝑖𝑐𝑒 = 𝑝𝑟𝑖𝑐𝑒 𝑜𝑓 𝑡ℎ𝑎𝑡 𝑙𝑒𝑣𝑒𝑙
𝑊𝑇𝑃 = −2,497
Consumers are willing to pay €1.99 less for Unox Runder Knaks when they are confronted with the heavy criticism. Since the product cost €1.70 in the supermarket, the severe criticism is non compensatory. Considering the WTP with the moderate criticism in mind, consumers are willing to pay €1.68 less. This analysis shows how severe the effect of criticism can be.
Furthermore, analyzing the effect of likes on the willingness to pay, consumers are willing to pay approximately €0.15 more when a product has 10.000 likes. Moreover, this implies that consumers accept a price increase of almost 10 percent. On the other side, when it has no likes people want to pay the same amount of euro’s less for the Unox Runder Knaks. It is important to notice that these arguments only effective when there is no criticism. The amount of likes strengthen the importance to the product whereby the willingness to pay increases.
However, when a product receives criticism and the message is liked 10.000 times the likes will strengthen the importance. This insight also contributes to the support of Hypothesis 3. The more likes criticism on a product has the less likely the product will be purchased.
In order to analyze on a segment level a latent class regression analysis was performed. Normally, the number of classes would be selected based on the information criteria – BIC, AIC & CAIC. Since this is not possible because no minimum can be detected by the information criteria, the R² value is the best to indicate the number of classes. Each increase in the R² between classes signals that the regression model explains more thanks to the added class. Considering the relative increase in the R² by adding a class the optimal number that will be used is three (as indicated by the bold letters in table 4.13).
Class LL BIC(LL) AIC(LL) CAIC(LL) Npar Class.Err. R² 1-class -‐9817.5 19690.3 19654.9 19700.3 10 0 0,26 2-class -‐9108.8 18394.8 18281.5 18426.8 2432 0,0012 0,49 3-class -‐8662.9 17625.2 17433.8 17679.1 54 0,0122 0,56 4-class -‐8430.3 17281.7 17012.5 17357.7 76 0,0196 0,63 5-class -‐8186.3 16915.6 16568.6 17013.6 98 0,0284 0,68 6-class -‐8039.2 16743.3 16318.4 16863.3 120 0,0284 0,70 7-class -‐7938.6 16664 16161.1 16806 142 0,0268 0,71 8-class -‐7816.9 16542.6 15961.8 16706.6 164 0,0259 0,73
Table 4.13 – Latent Class output
To verify if the Latent Class regression is performed correctly, the output of the parameters in Latent Class should be equal to the output from the previous regression model (table 4.12). Table 1 in the Appendix I indicates that the parameters are the same as with the other conjoint analysis, except for the interaction effects.
Class 1 Class 2 Class 3 P-value Intercept 7.6774 8.9329 3.8427 .000 Price Actual -1.4089 -1.4865 -.4792 .000 Criticism Moderate -2.1388 -1.0157 -1.4914 .000 Criticism Heavy -2.8942 -1.2537 -1.6826 .000 Likes 10.000 .7363 0.1843 0.3028 .0053 Likes CriticismM -.7516 -.4535 -.3054 .18 Likes CriticismH -.612 -.2778 -.2723 .30 Involvement CritM -.1486 -.4284 -0.0624 .0064 Involvement CritH -.3511 -.5072 .0061 .000 Gender -.1573 -.453 .0.6103 .085 Age -.0221 -.0035 .0256 .010 Residence -.3355 -.1296 .4651 .170 Education .0395 -.0677 .0282 .740 Groceries .3421 -.4907 .1486 .190 FMCG .502 -.8258 .3238 .014 Facebook .0957 -.0227 -.073 .540 Statement Friends .0724 -.1464 .074 .260 Statement Experts .0359 .2402 -.2761 .026 Statement Firms .0702 -.0076 -.0626 .680 Statement NGO -.0244 -.1426 .167 .180 R² .4076 R² .3032 R².3564
Dependent Variable – Ratings of respondent
Table 4.14 – Latent Class output: Parameters of the classes
Class 1 – Highly criticism focused
Represents consumers, which are vey likely to purchase a product (high intercept) and price is important to them. They are extremely sensitive for criticism and is an important factor to not purchase a product. In addition, the extent of involvement negatively influences the consumer behavior. Consumers in this class are quite involved and many are FMCG employees.
Class 2 – Price sensitive
Given the intercept, this class is extremely likely to purchase the product independent of the other parameters. However, these consumers are mostly focused on the price to determine the purchase. Despite their high involvement and negative influence on the purchase behavior, these consumers are not really influenced by criticism.
Class 3 – Price insensitive
Furthermore, table 2 in appendix I show the descriptives of the several classes. Overall in every class men are best represented and the most respondents finished a WO degree. Interesting is that class 1 has the lowest average age (33,6), most consumers from the countryside (3,2%) compared with the other classes and almost no FMCG employees (89,9%).
Moreover, in class 2 most consumers live in the city (80,1%), most consumers are responsible for their groceries (78%) and interesting is that comparing with the other classes class 2 has the highest percentage without a Facebook account (24,1%).
5 Conclusion and recommendations
After coming up with the results two of the three hypotheses were proved to be significant and therefore supported. The main finding, that criticism has indeed a negative influence on the consumer purchase behavior is proven. However, a critical reflection of the results is needed.
First, the hypothesis will be summarized, discussed and the link between the theoretical framework and the results will be reflected. Second, the academic - and managerial implications will be described. Finally, the recommendations and limitations will be discussed.
5.1 – Conclusion
First, the sub-question: What is the effect of food safety criticism on consumer purchase behavior? will be discussed. The results of hypothesis 1 clearly indicated that criticism has a negative influence on the consumer purchase behavior and is consistent with the theory. Negative reviews, messages or rumors hurt product evaluations, brand evaluation and reduce purchase likelihood (Tybout et al., 1981, Wyatt & Badger 1984, Huang & Chen, 2006). Negative information, especially in food safety related subjects could have a devastating impact on the business as was shown in the example of Coca Cola.
However, the effect of criticism has a reasonable effect on the consumer purchase behavior. The WTP outcome demonstrated that products with severe criticism are non-compensatory. Which means that consumers are not willing to accept a product with severe criticism even if they don’t have to pay for the product. Since, food safety concerns the health of a person and therefore consumers will probably not accept any criticism.
argument (Gupta & Harris, 2010), and therefore would be less affected by criticism through social media. On the other hand, high involved consumers have more knowledge of food safety issues and therefore can be stronger influenced by well-argued criticism statements.
Third, the sub-question: How does the number of likes influence the effect of product criticism on consumer purchase behavior? is discussed. The results of hypothesis 3 demonstrated that the moderating effect of the number of likes negatively influences the consumer purchase behavior and it enhances the criticism. The results are consistent with the theory. Consumers make their opinion publicly by liking or commenting on a post and therefore liking and commenting can be seen as a similar effect as word-of-mouth communication (de Vries et al. 2012). The WoM communication can be evaluated as a recommendation and according to the literature recommendations on social media influence product consideration and product choice (Gupta & Harris, 2010). Furthermore, social contagion also plays a role in influencing the behavior of a consumer. In the decision to ‘adopt’ a post it is assumed that social contagion plays an important role (de Vries et al. 2012). When a friend likes a message it might influence the friends’ fans (Aral & Walker, 2011). An increase in likes, under moderate or heavy criticism, will stronger decrease the likelihood that a consumer purchases the product. Therefore, it can be concluded that an increasing amount of likes enhances the impact of the message.
Finally, to answer the research question: What is the effect of food safety criticism by social media on consumer purchase behavior of food products?
First of all criticism has a negative effect on the consumer purchase behavior due to the fact that criticism is given more weight as Fiske (1980) argued. Although there is a difference in the type of criticism, in this study moderate – and heavy criticism was used, the fact remains that it still has a negative influence on purchase behavior. However, moderate criticism has a slightly less negative effect.
In addition, the amount of involvement in food safety influences the extent of accepting the negative information and their influence the purchase behavior. Given the results consumers high involved in food safety are less likely to purchase the product.
likelihood that a consumer purchases a food product. Social contagion and influence plays an important role (de Vries et al., 2012; Aral & Walker, 2011).
However, to put this into perspective, in today’s information rich environment, some amount of negative information is likely to exist for almost every product or brand. However, as the conclusion shows the amount of likes seems to have an important impact. The effect of the messages is enhanced by the amount of likes. In this information rich environment there is a possibility of information overload and as Berger et al. (200) argued, consumers have finite attention, it also could be that consumers will not notice or remind the criticism while their shopping.
In addition, the results indicated that consumers responsible for groceries pay more attention to price. A logical reason could be that consumers, which are responsible for groceries, have more knowledge of prices and therefore can better decide on the attribute price whether or not to purchase a product. Furthermore, FMCG employees considered the moderate criticism as more influencing. A logical reason could be that employees of Unilever or other companies evaluate the heavy criticism statement as untrue, because it concerns a brand of the company
5.2 – Management Recommendations
An eye-opener concerning a product under severe food criticism is that such a product is non-compensatory, which means that a consumer will not purchase the product even when it is for free (-€0,29). A normal reaction in the FMCG industry is to boost sales through a payoff. However, when a product has heavy criticism there are no possibilities to sale your product.
Interesting would be if in time consumers will forget the criticism and start buying the product again. A FMCG producer could take precautions if this point would be discovered. For instance, by reducing the production.
However, to be able to prevent a negative statement to become viral could prevent possible declines in sales.
On the other side, companies in the FMCG business can also try to stimulate the purchase behavior through marketing campaigns. For instance, by focusing on the good aspects of the food product. Since the topics on social media change rather fast, organizations can react quickly on the negative information. Moreover, companies could also explain how they improved the product due to the negative information. Thereby showing that the care about their consumers and adapt the product to eliminate the parts that affect the food safety criticism.
5.3 – Limitations and future research recommendations
First, one of the limitations in the research is that a factorial design approach was used with the traditional conjoint analysis. A traditional conjoint analysis will be performed using a factorial design approach, which means that the respondent evaluates all possible profiles. However, a disadvantage of this approach is that it can be valued as time consuming to evaluate 18 different choice sets, whereby every time one attribute level changes. Several respondents indicated that the questionnaire was long and therefore some of the respondents were treated as outliers. Therefore, in the future a fractional factorial design would be recommended.
Second, a limitation in the research is that one type of social media is used (Facebook). Considering the various social media and the differences between them it would be recommended to use several types of social media. If several types of social media are used or the most common, then it would be able to interpret the results for all variants of the media type. However, in this study the results are limited for Facebook.
Third, a recommendation for future research is to include the variable brand loyalty into the model. It is interesting to analyze how brand loyal or product loyal consumers deal with severe criticism on their product. And to what extent loyal consumers will continue to buy the product despite the criticism.
References
Ahluwalia, R., Burnkrant, R.E., Unnava, H.R. 2000, Consumer Response to Negative Publicity: The Moderating Role of Commitment, Journal of Marketing Research, 37(2), 203-214
Arrington, M. 2009, NY Times Tech Editor Reads Tech Cninch Every Morning for Story Ideas," TechCrunch.com, (July 13)
[available at http:// techcrunch.com/2009/07/13/nytimes-tech-editor-reads- techcrunch-every-moming-for-story-ideas/].
Bearden, W.O., Etzel, M.J. 1982, Reference group influence on Product and Brand Purchase Decisions, Journal of Consumer Research, 9(2), 183-194
Beirut. 2009, Why do people really tweet? The psychology behind tweeting! Bell, R., Marshall, D.W., The construct of food involvement in behavioral research: scale development and validation, Appetite, 40(3), 235-244
Berger, J., Sorensen, A.T., Rasmussen, S.J. 2010, Positive Effects of Negative Publicity: When Negative Reviews Increase Sales, Marketing Science, 29(5), 815-827 Berlyne, D. E. 1954. A Theory of Human Curiosity, British Journal of Psychology, 45, 180-191
Birnbaum, M. H., Stegner, S. E. 1979, Source Credibility in Social Judgment: Bias, Expertise, and the Judge’s Point of View, Journal of Personality and Social Psychology, 37, 48–74
Bond, J., Kirshenbaum, R. 1998, Under the Radar: Talking to Today's Cynical Consumer. New York: John Wiley & Sons.
Brewer, M.S., Rojas, M. 2008, Consumer Attitudes Toward Issues in Food Safety, Journal of Food Safety, 28, 1-22
Brown, J.J., Reingen, P.H. 1987, Social Ties and Word-of-Mouth Referral Behavior, Journal of Consumer Research, 14(3), 350-362
Brown, J., Broderick, A.J., Lee, N. 2007, Word of Mouth Communication within Online Communities: Conceptualizing the Online Social Network, Journal of Interactive Marketing, 21(3), 1-19
Buda, R., Zhang, Y. 2000, Consumer Product Evaluation: The Interactive Effect of Message Framing, Presentation Order and Source Credibilityi Journal of Product & Brand Management, 9, 229–242.
Bunker, T. 1996, Right on the Money, Brandweek, 37(8), 30-32
Chevalier, J.A., Mayzlin, D. 2006. The Effect of Word of Mouth on Sales: Online Book Reviews, Journal of Marketing Research, 48, 345-354
