Graduate School of Communication Research Master Communication Science
Master’s Thesis
A Blogger and a Health Organisation: Both or Neither?
The Effects of the Presence of a Blogger and a Health Organisation in Healthy Living Blogs on Sugar-Sweetened Beverages Consumption
28 June, 2018
Chamoetal Zeidler 10531947
Chamoetal.zeidler@student.uva.nl
Abstract
Obesity has become a prevalent problem worldwide, partly caused by sugar-sweetened beverages (SSB) consumption. Healthy living blogs are widely popular, but they often communicate wrongful information, which can be harmful for the blog’s followers. Therefore, the blogs’ source aspects must be studied to understand what can be used to communicate correct health information by health organisations. Expectations included a positive effect of the presence of a blogger and a health organisation (i.e., Voedingscentrum and the FDA) on reducing SSB consumption via source credibility, source resistance, behavioural attitude, and behavioural intention, with need for autonomy and external control moderating this effect. An online 2 (blogger: absent/present) x 2 (health organisation: absent/present) experiment was conducted among 320 Dutch and American participants, randomly assigned to one of the four conditions. Results indicated that there was no effect of the blogger or health organisation on source credibility nor source resistance. Source credibility and source resistance did not affect behavioural attitude. A positive effect of behavioural attitude was found on behavioural intention, but no effect of behavioural intention on SSB consumption was found. The effects were not moderated by need for autonomy nor by need for external control. An unmediated effect was found of the blogger on behavioural attitude, indicating that a blogger led to a more negative behavioural attitude. Only for the Dutch participants, a negative effect of behavioural intention on SSB consumption was found. These findings suggest that while the presence of a health organisation has no effect on SSB consumption, the presence of a blogger seems to negatively influence behavioural attitude. Thus, the message source in the form of a blogger must be considered carefully.
Keywords: health communication, healthy living blog, healthy lifestyle, need for autonomy, need for external control, sugar-sweetened beverages consumption
A Blogger and a Health Organisation: Both or Neither?
The Effects of the Presence of a Blogger and a Health Organisation in Healthy Living Blogs on Sugar-Sweetened Beverages Consumption
Worldwide, many people have an unhealthy lifestyle. An unhealthy lifestyle can be a major cause of chronic illnesses, such as cancer, and consequently puts an enormous, preventable strain on healthcare (World Health Organization, 2018). Obesity, a problem caused by an unhealthy lifestyle, has increased tremendously in recent years. It is a primary risk factor for death in the Western world (World Health Organization, 2017a) and can cause several diseases such as diabetes (Boncinelli, Gerini, Pagnotta, & Alfnes, 2016). In 2016, 39 and 13 percent of the world population was considered to be overweight and obese,
respectively (World Health Organization, 2017b). Obesity is often caused by an imbalance between energy intake and expenditure (Te Velde et al., 2012), which is often caused by an unhealthy diet and insufficient physical activity (Boncinelli et al., 2016). In theory, the
obesity problem could be solved by a healthier diet and more exercise (Te Velde et al., 2012). However, changing such behaviours is not as easy as it seems (Sheeran, 2002). Therefore, it is important to study how obesity can be prevented or overcome.
A major contributor to the energy imbalance, and therefore also to the obesity problem, is the consumption of sugar-sweetened beverages (SSB; e.g., Harvard, 2012; Te Velde et al., 2012). SSB include a large amount of sugar which negatively influences the consumer’s health (Harvard, 2012) and it directly influences weight gain and obesity (Malik, Schulze, & Hu, 2006). However, SSB are widely consumed (Harvard, 2012). In the Netherlands
(Nederlandse Vereniging FWS, 2013) and in the United States (Harvard, 2012), SSB are most often consumed by people between the ages of seven to forty. The consumption of SSB by children and adolescents under the age of eighteen is strongly influenced by their parents (Kremers, Van der Horst, & Brug, 2007), while adults over eighteen start taking responsibility
for their own health (Boiarsky, Rouner, & Long, 2013). Since they just start creating patterns for (future) health behaviour, the target group of this study included (young) adults between the ages of eighteen to forty that consume SSB.
Correct health information is important for people’s health-related decision-making (Hu & Sundar, 2010). Hence, correct information must be effectively communicated to promote reducing SSB consumption. The online environment is considered to be the biggest source for obtaining health information (Fox & Jones, 2009) and for information about how to maintain a healthy lifestyle (Boepple & Thompson, 2014; Meitz, Ort, Kalch, Zipfel, & Zurstiege, 2016). This is especially relevant for young adults, who spend most of their time online as compared to other age groups (Anderson, Steen, & Stavropoulos, 2017; Boiarsky et al., 2013). The online environment is overloaded with inaccurate, untrustworthy, and
controversial information, which creates uncertainty and confusion among the readers and, consequently, readers often do not know what health behaviour advice to follow in order to obtain a healthy lifestyle (Lee & Cho, 2017). Therefore, it is important to effectively reach people with the correct information online.
A popular online health information source is blogs (Caplette et al., 2017; Lu, 2013; Meitz et al., 2016). Blogs are online journals, often presented in a reversed chronological order (Colliander & Erlandsson, 2015). Healthy living blogs (HLBs), blogs focused specifically on promoting a healthy lifestyle, reach large audiences with persuasive messages for successful behaviour change (Boepple & Thompson, 2014; Lu, 2013). However, many HLBs often seem to promote an incorrect message which can harm the blog’s readers: many bloggers are not health professionals and do not have health-related training (Boepple & Thompson, 2014). Simultaneously, health organisations trying to promote the correct message are often less successful than HLBs in reaching their target audience through online media (Schneider, 2014). Thus, it is important to understand which of HLBs’ aspects lead to their success in
terms of behaviour change, so health communication professionals can adapt this information to their benefit. The main observed difference between the popular HLBs and health
communication by professionals is the message source: the source of the former is a (popular) blogger and the source of the latter is a health organisation. Thus, a contributor to the success or failure to initiate behaviour change may be the message source in the form of the presence or absence of a blogger and the presence or absence of a health organisation.
Prior marketing research found that a message source influences perceived credibility, brand attitudes, and behavioural intentions (Ballantine & Yeung, 2015). The message source of health information may have the same effects. The persuasiveness of HLBs has been compared extensively to other online sources, such as online magazines, institutional websites, and Facebook (e.g., Boiarsky et al., 2013; Choi, Miao, Almanza, & Nelson, 2013; Hu & Sundar, 2010; Meitz et al., 2016; Neubaum & Krämer, 2015). While HLBs have been compared to other sources, no study to this date has examined different source types of the same HLB and no study has combined the effects of two possible source types on the health message persuasiveness. Health information can be obtained online from websites generated by organisations as well as from lay-people (i.e., HLBs; Choi et al., 2013). Therefore, two source types were chosen for this study: a blogger and a health organisation. This study’s aim is thus to understand whether the explicit presence or absence of a blogger and the explicit presence or absence of a health organisation are necessary for the HLB’s persuasiveness, in terms of behaviour change.
Several underlying mechanisms and moderating variables are proposed that may explain the effect of the source on the message persuasiveness. Understanding the cognitive and affective mechanisms underlying the message persuasiveness provides a means of understanding the effect (Ahluwalia, 2000). Moreover, it is assumed that if the content
2008; 2014). Therefore, the mediators and moderators, which are elaborated on in the theoretical framework, were studied as well. Thus, the following research question is proposed: How do a blogger (versus no blogger) and a health organisation (versus no organisation) affect reducing SSB consumption, and which underlying mechanisms and moderators play a role in this effect?
Theoretical framework Healthy Living Blogs
Blogs are a popular and influential form of online communication (Caplette et al., 2017; Meitz et al., 2016). Some Dutch bloggers receive over 100,000 unique visitors daily (i.e., beautygloss.nl) and some American bloggers even receive close to 1,000,000 unique visitors daily (i.e., thechive.com; Van Reijmersdal et al., 2016). Specifically, HLBs have a great potential to promote healthy lifestyle behaviours, as they could positively influence readers by changing their intention to engage in health behaviours (Boepple & Thompson, 2014;
Caplette et al., 2017). HLBs include user-generated personal content (Miller & Pole, 2010) and share the blogger’s healthy lifestyle, aiming at providing health information and serving as an example for the readers (Boepple & Thompson, 2014). A systematic review of HLBs (Miller & Pole, 2010) found that most bloggers post at least once a week, and about half of the blogs was written by health professionals while the other half was written by lay-people with no health education. Most HLBs focus either on a healthy lifestyle in general (i.e., diet and exercise), a particular disease, or a specific health condition (Miller & Pole, 2010).
Blogs are defined as “frequently updated websites… typically [posted] by a single author” (Neubaum & Krämer, 2015, p. 873). Thus, the blog’s author is presented with their name and often a profile picture (Boepple & Thompson, 2014), in contrast to health organisations’ websites where the organisation is known but the author is not. Readers have different responses to online health information depending on the source (Mazzocchi, Lobb, Traill, &
Cavicchi, 2008). For example, Meitz et al. (2016) found that HLBs were perceived as
significantly more relevant and led to higher source credibility and trustworthiness compared to Facebook (Meitz et al., 2016). Neubaum and Krämer (2015) found that readers of a person-centred HIV blog had a more positive attitude towards condom use than readers of an
institutional website, but the institutional website was perceived as more credible than the blog (Neubaum & Krämer, 2015). Caplette et al. (2017) found that exposure to a HLB for six months significantly increased the fruit and vegetable consumption of readers, compared to no exposure to this blog. While these studies compared different sources, to date, no study has combined the effect of two source types on the message persuasiveness. Health information can be obtained online from websites/blogs generated by health organisations and lay-people (Choi et al., 2013). Often if a blogger writes on a health organisation’s website, the author is not explicitly presented. In contrast, if a blogger posts on their blog on behalf of a health organisation, the blogger is explicitly presented, but the organisation is not. Other online texts may also be presented without a source. Therefore, two elements of the source type were proposed: a blogger and a health organisation. The expectations of the message
persuasiveness with the presence or absence of a blogger and a health organisation on reducing SSB consumption are outlined in this theoretical framework.
A Blogger
A healthy living blogger can serve as an example for the readers (Neubaum & Krämer, 2015). Prior studies (e.g., Green, 2006; Zillmann, 2006) showed that bloggers who serve as an example to their readers can positively change the readers’ attitudes towards the advocated health behaviour. A blogger has stronger effects on readers’ attitudes than an abstract presentation of information without a clear author (Neubaum & Krämer, 2015).
However, the message often does not directly affect the persuasiveness, and underlying mechanisms explaining the effect of the blogger on the persuasiveness are not taken into
account (Burgoon, Alvara, Grandpre, & Voulodakis, 2002). These underlying mechanisms may explain why certain effects are found and can give more insight into the psychological mechanisms of these effects (Van Reijmersdal, et al., 2016). Two important underlying mechanisms to the message persuasiveness are source credibility (Meitz et al., 2016) and source resistance (Burgoon et al., 2002). Source credibility is the extent to which the source of a message is perceived as credible, defined by perceived source expertise and trustworthiness (Hovland & Weiss, 1951). Resistance is a motivational state in which one experiences a wish to counteract an attempt of limiting one’s choices (Brehm, 1966), such as a persuasive attempt experienced from reading a blog-post (Knowles & Lin, 2004). A persuasive attempt causes the sense of threat to one’s freedom, which leads to resistance towards the source (i.e., source resistance; Burgoon et al., 2002), and it includes dismissing the source’s validity (Fransen, Verlegh, Kirmani, & Smit, 2015), expertise, or trustworthiness (Zuwerink Jacks & Cameron, 2003).
Health bloggers that write about their personal stories and experiences are perceived as more authentic and, therefore, more credible (Neubaum & Krämer, 2015). In HLBs, the bloggers often share their personal stories. Therefore, it is expected that a blogger compared to no blogger would increase the source credibility. Subsequently, source credibility is a critical determinant for persuasion (Hovland & Weiss, 1951; Petty & Cacioppo, 1979). Specifically, the credibility of an online source plays an important role while evaluating the information and it affects the attitudes towards health (Meitz et al., 2016). A credible source has more persuasive power than a less credible source, and leads to a more positive attitude and a higher behavioural intention (Choi et al., 2013; Meitz et al., 2016). Thus, the source credibility serves as a mediator of the effects of the message on the message persuasiveness (Meitz et al., 2016).
Theory of Planned Behaviour. An often-used theory to predict behavioural change is the
Theory of Planned Behaviour (TPB; Ajzen, 1991). According to the TPB (Ajzen, 1991) and the Integrative Model (Fishbein & Cappella, 2006), health behaviour is predicted by the intention to perform the behaviour, which is predicted by the attitude (i.e., positive/negative behavioural evaluation), subjective norm (i.e., perceived degree to which important others believe the behaviour should be performed), and perceived behavioural control (i.e., perceived ease or difficulty of performing the behaviour) of the behaviour (Ajzen, 1991; Fishbein & Cappella, 2006). The Integrative Model varies per population and behaviour (Fishbein & Cappella, 2006). Prior studies found that attitude is the most important predictor of intention to reduce SSB consumption compared to subjective norm and perceived behavioural control (Zoellner, Estabrooks, Davy, Chen, & You, 2012). In another study, attitude was the only significant predictor of intention (Jordan, Piotrowski, Bleakley, & Mallya, 2012). Therefore, only the attitude to reduce SSB consumption (behavioural attitude) was taken as a predictor of the intention to reduce SSB consumption (behavioural intention). Past research found that the attitude toward the behaviour postulated in the message positively predicts the behavioural intention (e.g., Gallagher & Updegraff, 2012; Jordan et al., 2012; Van Reijmersdal et al., 2016), which, subsequently, positively predicts performing the behaviour (e.g., Sheeran, 2002; Webb & Sheeran, 2006). Therefore, the following hypothesis is proposed:
H1: Through a higher source credibility, the presence of a blogger (versus no blogger) leads to a more positive behavioural attitude, higher behavioural intention, and lower SSB consumption.
Similarly, source resistance may also serve as a mediator of the message persuasiveness (Brehm, 1966; Brehm & Brehm, 1981; Burgoon et al., 2002). According to the Theory of Psychological Reactance (Brehm, 1966; Brehm & Brehm, 1981), any message aimed at changing the reader’s attitudes can be seen as a threat to one’s freedom. A blog can be
perceived as such a persuasive attempt (Van Reijmersdal et al., 2016). Once one feels that his/her freedom is threatened, it may result in psychological reactance, the motivation to regain one’s freedom threatened by the message. The reactance can result in, among other responses, source derogation (i.e., source resistance, Burgoon et al., 2002). A higher source resistance may lead to a lower persuasion (Fransen et al., 2015; Jenkins & Dragojevic, 2011). When the source is derogated, the reader experiences more negative attitudes towards the advocated topic in the message. However, a blogger makes the readers experience closer feelings toward the blogger (Klimmt, Hefner, & Vorderer, 2009). Therefore, with the presence of a blogger, the readers are expected to experience less resistance towards the source and, consequently, a more positive behavioural attitude. As mentioned earlier, the behavioural attitude positively predicts the behavioural intention which positively predicts performing the behaviour (Ajzen, 1991; Fishbein & Cappella, 2006). This leads to the following hypothesis:
H2: Through a lower source resistance, the presence of a blogger (versus no blogger) leads to a more positive behavioural attitude, higher behavioural intention, and lower SSB
consumption.
A Health Organisation
Next to differing in the presence or absence of a blogger, the HLB’s source can also differ with the presence or absence of a health organisation. According to Boiarsky et al. (2013), exposure to a message from an organisation (versus an unknown source) leads to more positive attitudes and stronger behavioural intentions. This effect is especially strong when the message includes important topics, such as health (Rieh & Hilligoss, 2008). Similar to the effect of the blogger, the message persuasiveness of a health organisation may also be
mediated by two underlying mechanisms: source credibility (Meitz et al., 2016) and source resistance (Burgoon et al., 2002).
A formal source of health information (i.e., a health organisation), has a high credibility (Rieh & Hilligoss, 2008). Health information published by an organisation is perceived as more credible compared to no organisation (Eastin, 2001). Furthermore, Neubaum and Krämer (2015) found that readers perceived the source as more credible when they read the information from a website of an official institution rather than a lay-person’s text. In general, governmental organisations and health authorities are usually a highly trusted information source (Choi et al., 2013; Mazzocchi et al., 2008). Thus, it is expected that health
organisations are perceived as more credible compared to no organisations, because they are perceived as more accurate and reliable (Neubaum & Krämer, 2015). As mentioned earlier, a higher source credibility may lead to more positive attitudes (Choi et al, 2013), and attitude positively predicts the behavioural intention, which predicts performing the behaviour (Ajzen, 1991; Fishbein & Cappella, 2006). Since source credibility may serve as a mediator of the message persuasiveness (Meitz et al., 2016), the following hypothesis is proposed: H3: Through a higher source credibility, the presence of a health organisation (versus no organisation) leads to a more positive behavioural attitude, a higher behavioural intention, and lower SSB consumption.
Source resistance may also serve as a mediator of the effect of the source on the message persuasiveness (Burgoon et al., 2002). According to the Politeness Theory (Brown &
Levinson, 1987), one must feel he/she is respected and treated as an equal by their conversational partner (i.e., the message source) by, for example, being seen by the conversational partner as competent to make one’s own decisions (Jenkins & Dragojevic, 2011). If a blog is perceived as not respecting and treating the reader as equal, and it makes the reader feel incompetent, it may lead to higher source resistance (Jenkins & Dragojevic, 2011). A persuasive message such as a HLB is inherently a threat to one’s freedom and can, therefore, lead to source resistance (Burgoon et al., 2002). However, the Politeness Theory
posits that this occurs less when the source is perceived as having legitimate power and the right to influence the reader (Jenkins & Dragojevic, 2011). The source’s legitimate power depends on the topic. A health organisation is assumed to have legitimate power regarding health information. Therefore, it is expected that a health organisation produces less source resistance than no organisation, which may lead to a more positive behavioural attitude (Jenkins & Dragojevic, 2011). The behavioural attitude positively predicts the behavioural, intention which predicts performing the behaviour (Ajzen, 1991; Fishbein & Cappella, 2006). Therefore, the following hypothesis is proposed:
H4: Through a lower source resistance, the presence of a health organisation (versus no organisation) leads to a more positive behavioural attitude, higher behavioural intention, and lower SSB consumption.
Need for Autonomy and Need for External Control
According to the Self-Determination Theory (SDT; Deci & Ryan, 2000), people have the basic needs for competence (i.e., feeling skilled and effective), relatedness (i.e., feeling connected to the social environment), and autonomy (i.e., feeling the freedom to choose; Ng et al., 2012; Smit & Bol, 2018). While both the need for competence and relatedness have been studied often in a health communication context, research on need for autonomy (NFA) has been neglected in this context (Deci & Ryan, 2000). The SDT posits that an autonomous motivation is necessary for behaviour change. However, people differ in their NFA and need for external control (NFC), and more specifically, in their decision-making process regarding their health behaviour (Deci & Ryan, 2000; Levinson, Kao, Kuby, & Thisted, 2005; Smit, Linn, & Van Weert, 2015). Some people prefer to make their own health behaviour choices (NFA), others prefer advice from an expert (NFC), yet others have high or low preferences for both (Levinson et al., 2005). A latent class analysis of NFA and NFC regarding health-related decisions found that people can be divided into four groups according to their NFA and NFC:
self-reliers (high NFA), confirmation-seekers (high NFA and high NFC), expert-dependents (high NFC), and indifferents (low NFA and low NFC; Smit & Bol, 2018). Thus, people differ in their NFA and NFC and these differences may influence the persuasiveness of health messages (Resnicow et al., 2008; 2014), such as HLBs. Resnicow et al. (2008; 2014) tested this moderation effect in offline settings. It was found that health information was more relevant for participants with a high NFA, but not participants with a low NFA, and autonomy supportive messages increased fruit and vegetable consumption (Resnicow et al., 2008) and colorectal cancer screening rates (Resnicow et al., 2014) among participants with a high NFA, but not participants with a low NFA.
Bloggers (Colliander & Erlandsson, 2015) and health organisations (Choi et al., 2013) are perceived as experts within their field of expertise. People perceive themselves as being autonomous when they experience psychological freedom to choose their health behaviour change (Smit & Bol, 2018). Therefore, people with a higher NFA may perceive the blogger or health organisation as irrelevant for their autonomous choice regarding their health behaviour (Resnicow et al., 2008). This leads to a weakening of the positive effect of the blogger and health organisation on the message persuasiveness for people with high NFA, but not for people with a low NFA, leading to the following hypothesis:
H5a: The effect of the blogger on the message persuasiveness through (1) source credibility and (2) source resistance is moderated by the NFA, such that the effect is weaker for
participants with a higher NFA.
H5b: The effect of the health organisation on the message persuasiveness through (1) source credibility and (2) source resistance is moderated by the NFA, such that the effect is weaker for participants with a higher NFA.
According to the SDT, external regulation is considered as controlling the behavioural decisions (Deci & Ryan, 2000). Bloggers and health organisations are perceived as experts
within their field of expertise (Choi et al., 2013; Colliander & Erlandsson, 2015). Thus, the HLB written by a blogger or a health organisation provide expert advice and is perceived as controlling the behavioural decisions. Therefore, people with a higher NFC perceive the blogger and health organisation as more relevant. This leads to a strengthening of the positive effect of the blogger and health organisation on the message persuasion for people with a high NFC, but not for people with a low NFC, leading to the following hypothesis:
H6a: The effect of the blogger on the message persuasiveness through (1) source credibility and (2) source resistance is moderated by the NFC, such that the effect is stronger for participants with a higher NFC.
H6b: The effect of the health organisation on the message persuasiveness through (1) source credibility and (2) source resistance is moderated by the NFC, such that the effect is stronger for participants with a higher NFC.
See Figure 1 for the illustrated conceptual model discussed in the theoretical framework.
Figure 1. Conceptual model.
American and Dutch Samples
For effects validation, cross-cultural comparison, and generalisation of the results, next to testing the hypotheses on the overall sample, the hypotheses are tested with two sub-samples: the Netherlands and the United States (U.S.). The differences and similarities between the
samples are studied. Differences are, however, unknown, leading to the following sub-question: How do the results differ between the Dutch and American participants?
Method Design and Procedure
An experiment was conducted with two samples recruited via the panels PanelClix (Dutch participants) and Mechanical Turk (MTurk; American participants). Participants were
randomly assigned to one of the four conditions in a 2 (blogger: present/absent) × 2 (health organisation: present/absent) between-subjects design.
The study was approved by the ethical committee of the University of Amsterdam (reference number: 2018-PC-9073). Participants received a link to the questionnaire in Qualtrics at three different measurement times: T0, T1 (T0 + one week), and T2 (T0 + two weeks). One single blog-post has minimal effects on behaviour change, such as reducing SSB consumption (Boiarsky et al., 2013); therefore, multiple exposures are necessary. This led to a two-fold exposure to two blog-posts with an interval of one week (T0, T1), with an additional measurement of behaviour change (T2).
At T0, participants received an explanation of the study, including information on their compensation and an informed consent form, which they had to sign to participate in the study. Dutch participants received compensation after each questionnaire, with a higher compensation for T2 to increase motivation to complete all questionnaires. The compensation for T0, T1, and T2 were €1.75, €1.75, and €1.80 respectively. American participants received their compensation of $2.50 after completing all three questionnaires. Afterwards, the
baseline variables gender, age, educational level, current SSB consumption, BMI (Body Mass Index, calculated by one’s weight and height), whether the participant suffers from diabetes, NFA, and NFC were measured. Then, participants were exposed to a blog-post about the effects of SSB consumption on obesity. A manipulation check was conducted to test whether
participants recognised the blog’s source. At T1, participants were exposed to a blog-post about the effects of consuming SSB on diabetes and the manipulation check was conducted again. Moreover, source resistance, source credibility, behavioural attitude, and behavioural intention were measured. At T2, participants were asked about their SSB consumption and were thanked for their contribution to the study.
Participants
A total of 812 participants started at T0 (starting T1: n = 402, starting T2: n = 339). Only 320 responses (n Dutch = 202, n American = 118)1 were eventually included in the final sample after removing invalid questionnaires. Participants were omitted from the analysis because they did not meet the age requirements of 18-40 (n = 36), meet the requirement of consuming at least one sugar-sweetened beverage a week (n = 32), they completed the
questionnaire in a short (less than ninety seconds), or a long time (3 SD’s above the mean; n = 67), or they did not complete the entire questionnaire (n = 176). Additionally, 36 double id numbers were removed. In total, 529, 344, and 320 participants were considered as
successfully completing T0, T1, and T2, respectively, indicating a drop-out rate of 209 participants (39.5%) from T0 to T2.
The final sample included 143 women (44.7%), and participants ranged in age from 18 to 40 (M = 30.56, SD = 5.28). Most participants were highly educated (n = 167; 52.2%) and did not suffer from diabetes (n = 290, 90.6%). The average BMI was 25.56 (SD = 5.22) and the average daily SSB consumption at T0 was 487.57 ml (SD = 515.69). See Table 1 for the sample characteristics of the complete sample, as well as for the Dutch and American samples specifically.
1
An a-priori power analysis was conducted for the structural equation model with GPower 3.1 (Faul, Erdfelder, Buchner, & Lang, 2009). The model included four latent variables and 23 observed variables. To detect a moderate effect size with 80% power and an alpha of .05, 153 respondents were sufficient. However, rules of thumb for structural equation modelling advise 200 respondents (Kline, 2011). Therefore, the aim was to have 200 participants per sample.
Participants were randomly assigned to one of the four conditions: 75 (23.4%) were exposed to a blog-post with no health organisation or blogger, 81(25.3%) to a blog-post with a blogger, 78 (24.4%) to a blog-post with a health organisation, and 86 (26.9%) to a blog-post with a health organisation and a blogger. See Table A1 for the description of assignment to conditions.
Table 1
Sample characteristics
Stimulus Material and Manipulations
Each participant viewed two blog-posts with or without a blogger and/or a health organisation (see Appendix B for the stimulus material and manipulations). The HLB was fictitious, but its format was based on existing HLBs by mixing information and personal opinions (Choi et al., 2013), and by adding photos (Caplette et al., 2017). The blog-posts’ content was based on information from the Voedingscentrum (i.e., the Dutch nutrition centre; Voedingscentrum, n.d.-a; n.d.-b). Both blog-posts focused on the negative consequences of SSB consumption (T0: obesity, T1: diabetes) and gave tips on how to reduce SSB
consumption. To ensure the authenticity of the blog-posts written by the researcher, the content was assessed by an expert from the Voedingscentrum and the format was assessed by a journalist and by a blogger (Lu, 2013).
The blogger. In the blogger-present condition, the blog-posts were signed by the blogger,
Emily, including her picture and description (Boepple & Thompson, 2014). These elements were missing in the blogger-absent condition.
The health organisation. In the organisation-present condition, two well-known and
comparable health organisations were chosen: Voedingscentrum (Dutch sample) and the U.S. Food and Drug Administration (FDA; American sample). The organisation’s logo was added to the top and bottom of the blog-posts. The logo was missing in the organisation-absent condition.
Manipulation check. The manipulation check tested whether participants correctly
recognised the source, measured by the question “Who is the sender of the blog you have just read?” Participants could indicate whether they saw the blogger, the organisation, both, none, or whether they did not know. The photo of the blogger and logo of the organisations were added to the answers.
Variables and Measures
Source credibility included seven items on a seven-point bipolar scale (Choi et al., 2013). The statements included for example “the source of the blog I just read is:
uninformative/informative” and “inexperienced/experienced.” Due to low factor loadings revealed by the Confirmatory Factor Analysis (CFA; see Results), two items
(“biased/unbiased” and “subjective/objective”) were removed, resulting in a five-item scale. A mean score was calculated (α = .87, M = 5.36, SD = 0.98), with a higher score indicating a higher source credibility.
Source resistance included seven items on a seven-point bipolar scale (Jenkins & Dragojevic, 2011). The statements included, for example: “the source of the blog I just read is: stupid/smart” and “unintelligent/intelligent.” A mean score of the reversed scales was calculated (α = .95, M = 2.64, SD = 1.11), with a higher score indicating a higher source resistance
Behavioural attitude included five items on a seven-point bipolar scale (Jordan et al., 2012). The items included, for example: “I find reducing the consumption of sugar-sweetened beverages bad/good” and “foolish/wise.” Due to low factor loadings revealed by the CFA (see Results), two items (“complicated/simple” and “unpleasant/pleasant”) were removed2,
resulting in a three-item scale. A mean score was calculated (α = .90, M = 6.00, SD = 1.02), with a higher score indicating a more positive behavioural attitude.
Behavioural intention included three items. The first statement, “I am planning to drink less sugar-sweetened beverages in the coming week,” was based on a prior study (De Bruijn & Van den Putte, 2009). The two following statements, “I intend to…” and “I will try to drink less sugar-sweetened beverages in the coming week,” were based on a smoking cessation study (Hummel et al., 2015) and adapted to SSB consumption. A mean score was calculated (α = .96, M = 4.86, SD = 1.58), with a higher score indicating a higher behavioural intention.
Need for autonomy included two statements (Levinson et al., 2005): “I want to make my own decisions when it comes to my health” and “I want to make my own decisions when it comes to reducing my sugar-sweetened beverages consumption” (1 = completely disagree, 7 = completely agree). A mean score was calculated (α = .81, M = 5.49, SD = 1.27), with a higher score indicating a higher NFA.
2 Two factors were found when a Principal Component factor Analysis was conducted. The first factor, with an Eigenvalue of 2.98 explained 59.7% of the variance. It included the items “bad/good”, “foolish/wise”, and “harmful/beneficial”. The second factor, with an Eigenvalue of 1.12, explained 22.31% of the variance. It included the items “complicated/simple” and “unpleasant/pleasant”. This provides another proof that the last two indicators indeed do not measure the same concept of behavioural attitude as do the first three. Therefore, the last two were removed from the model.
Need for external control included four statements (Levinson et al., 2005), for example: “when it comes to my health, I prefer to be told what to do” and “When it comes to reducing my sugar-sweetened beverages consumption, I want an expert to tell me what to do” (1 = completely disagree, 7 = completely agree). A mean score was calculated (α = .89, M = 3.78, SD = 1.46), with a higher score indicating a higher NFC.
SSB consumption was measured by the average number of days per week of consumption: “on how many days per week do you drink sugar-sweetened beverages?” (0-7 days per week), and the average daily consumption of three beverage sizes: “on days that you drink sugar-sweetened beverages, how many: glasses, cans, and bottles do you drink?” (De Bruijn & Van den Putte, 2009; Kremers et al., 2007). To ensure that participants understand the concept SSB, it was defined in the questionnaire as “beverages with added sugars, for example, coke, lemonade, or fruit juices with added sugars.” The average daily consumption was multiplied by the amount of days and divided by seven, creating a daily average of SSB consumption in millilitres (M = 457.52, SD = 521.85). SSB consumption over 2000 ml daily was considered as an outlier and, therefore, adapted to a maximum of 2000 ml daily.
Control variables included age, gender, educational level, BMI (calculated by one’s weight and height), and suffering from diabetes. For an overview of the questionnaire, see Appendix C.
Data Analysis
Firstly, the manipulation check was done by conducting a χ2
-test, the equal distribution of the control variables across the conditions was tested by conducting χ2-tests and analyses of variance (ANOVAs), and correlations between the control variables and the dependent variables were calculated by conducting ANOVAs and regression analyses. Variables that were not equally distributed across conditions and correlated significantly with one or more dependent variables were included in the analyses as covariates.
Second, the hypotheses were tested by estimating the full model using structural equation modelling in AMOS 25. Source credibility, source resistance, behavioural attitude, and behavioural intention were included as latent constructs and estimated from their manifest indicators. The experimental conditions, SSB consumption, NFA, NFC, and the interaction terms of the moderators with the experimental conditions were included as observed
variables. The measurement model fit was assessed with a CFA. Items with low (< .60) factor loadings were removed. Afterwards, the structural model was assessed. Significant interaction effects were looked at more closely by performing a flood-light analysis, indicating at which levels of the moderator the effect of the independent variable is (in)significant (Spiller, Fitzsimons, & Lynch, 2013).
Results Randomisation Check
Significant differences were found between the conditions for education (χ2(6) = 20.55, p = .002): most of the lower educated participants were in the ‘no source’ condition. Moreover, a significant difference was found between the conditions for age; F(3, 319) = 4.51, p = .004. Participants in the ‘no source’ condition (M = 31.69, SD = 5.09) and in the ‘health
organisation’ condition (M = 31.33, SD = 5.02) were significantly older than participants in the ‘blogger and health organisation’ condition (M = 28.95, SD = 5.29). No significant differences were found between the conditions for gender (χ2(3) = 0.72, p = .868), diabetes (χ2
(6) = 2.20, p = .900), BMI (F(3, 316) = 1.73, p = .161), and baseline SSB consumption (F(3, 319) = 0.43, p = .731), indicating that most of the control variables were equally distributed across the conditions (aside from education and age). See Table A2 for the randomisation check for the Dutch and American samples.
Correlations with Dependent Variables
Since only variables that were not equally distributed across conditions and correlated significantly with the dependent variables were included in the analyses as covariates, only the significant correlations of education and age with the dependent variables are reported. See Table A3 for all correlations and an overview of the Dutch and American samples. Age correlated significantly with both the behavioural attitude (r = .13, p = .023) and SSB
consumption at T2 (r = -.14, p = .017). Since age was not randomly assigned to the conditions and it correlated with the behavioural attitude and with SSB consumption, it was included as a covariate in the analysis as a sensitivity test. However, adding age to the model did not affect the results. Therefore, for the model’s parsimoniousness, the reported results are without the covariate.
Manipulation Check
To test whether participants recognised the blog’s source as intended, a χ2-test was conducted for exposure at T0 and T1. For both exposures, the manipulation seemed to have worked with over 50% of the participants correctly recognising the source (see Table A4 for T0 and Table A5 for T1). The correct recognition rate ranged between 51.2-85.2%. Twice, there were significant differences between the conditions in terms of the source perception (T0: χ2(9) = 313.02, p < .001; T1: χ2(9) = 376.25, p < .001), indicating that the manipulation was successful. Therefore, the manipulation was used and effects found were attributed to the manipulation.
CFA Measurement Model
All variables seemed to be acceptable in terms of skewness and kurtosis and all
relationships were linear, thus, they could be tested using structural equation modelling. No problems with multicollinearity were found.
The model fit was assessed using a χ2
-test that, when not significant, indicates a good fitting model. However, the χ2
-test is sensitive to the sample size, therefore, the comparative fit index (CFI) and the root mean square error of approximation (RMSEA) were used as additional measures of model fit (Kline, 2011). CFI values over .90 are considered acceptable and values over .95 are considered good. RMSEA values under .08 are considered acceptable and values under .05 are considered good (Byrne, 2013). Full-information maximum
likelihood was used to estimate the model fit parameters. An initial measurement model with unit-loading indicators was found to have a reasonable model fit with a good CFI and an acceptable RMSEA value: χ2
(365) = 794.44, p < .001, CFI = .96, RMSEA = .06 90%CI [.06, .07]3.
Discriminant validity was assessed through factor correlations. A high correlation was found between source resistance and source credibility (r = .96). Between-factor correlations over .85 are considered high and may cause multicollinearity (Kline, 2011). However, source credibility and source resistance measure theoretically different concepts since they have different working mechanisms; source credibility is defined in terms of perceived source expertise and trustworthiness (Ballantine & Yeung, 2015), and source resistance is caused by a perceived threat to one’s freedom (Burgoon et al., 2002). Therefore, the factors cannot be merged and they were analysed separately. Other correlations ranged from -.39 to .82, demonstrating the distinct latent factors in the model (Kline, 2011).
Convergent validity was assessed by examining factor loadings. Factor loadings under .60 are considered problematic (Kline, 2011). Four factor loadings (source credibility:
“biased/unbiased” and “subjective/objective,” behavioural attitude: “complicated/simple” and “unpleasant/pleasant”) were found to be under r = .60 and were removed. This resulted in a
3 The initial Dutch model had an acceptable model fit: χ2(365) = 604.11, p < .001, CFI = .96, RMSEA = .06 90%CI [.05, .07]. The initial American model had poor fit: χ2(365) = 799.61, p < .001, CFI = .88, RMSEA = .10 90%CI [.09, .11]
good fitting model: χ2
(255) = 452.83, p < .001, CFI = .98, RMSEA = .05 90%CI [.04, .06]4. All other loadings on the intended latent constructs were significant, ranging from .68 to .95, indicating acceptable convergent validity (Kline, 2011). Thus, the measurement model
adequately measured the factors, indicating that the structural model can be further examined.
Structural Model
The structural model yielded a poor model fit with an acceptable CFI value but an unacceptable RMSEA value: χ2
(285) = 1012.06, p < .001; CFI = .93; RMSEA = .09 90%CI [.08, .10]5. When exploring the modification indices, it seemed that the disturbance terms of source resistance and source credibility had a high shared variance. Adding a correlation between these disturbance terms improved the model fit (χ2
diff(1) = 479.34, p < .0016) and resulted in a well-fitting model with good CFI and RMSEA values: χ2(284) = 532.73, p < .001; CFI = .97; RMSEA = .05 90%CI [.05, .06]7.
Hypotheses Testing
H1-4, which proposed that a blogger (H1-2) and a health organisation (H3-4) would lead to a lower SSB consumption through a higher source credibility (H1, H3) or a higher source resistance (H2, H4), a more positive behavioural attitude, and a higher behavioural intention, were not accepted. Results of the adapted model (see Figure 2) indicated that there were no effects of the blogger on source credibility (b* = -.30, p = .412), nor on source resistance (b* =.24, p = .504). In addition, no effect was found of the health organisation on source
credibility (b* = .09, p = .808), nor on source resistance (b* = .24, p = .503). No effect was
4 The adapted CFA Dutch model resulted in a good fitting model: χ2(255) = 390.10, p < .001, CFI = .98, RMSEA = .05 90%CI [.04, .06]. The adapted CFA American model resulted in an acceptable CFI value but a slightly high RMSEA value: χ2(255) = 506.50, p < .001, CFI = .93, RMSEA = .09 90%CI [.08, .10].
5 The Dutch structural model resulted in an acceptable model fit: χ2(285) = 679.76, p < .001, CFI = .93, RMSEA = .08 90%CI [.08, .09]. The American structural model resulted in an unacceptable model fit: χ2(285) = 753.70, p < .001, CFI = .87, RMSEA = .12 90%CI [.11, .13].
6 The added disturbance terms correlation significantly improved the Dutch model fit: χ2
diff (1) = 261.57, p < .001 and the American model fit: χ2
diff (1) = 563.47, p < .001.
7 The adapted measurement model resulted in a good Dutch model fit: χ2(284) = 418.19, p < .001, CFI = .98, RMSEA = .05 90%CI [.04, .06] and an acceptable American model fit, but with a slightly high RMSEA value: χ2
found of source credibility on behavioural attitude (b* = .19, p = .599) nor of source resistance on behavioural attitude (b* = -.40, p = .273). As expected, a significant positive effect was found of behavioural attitude on behavioural intention (b* = .47, p < .001), indicating that a more positive behavioural attitude led to a higher behavioural intention. However, no effect was found of behavioural intention on SSB consumption (b* = .07, p = .209).
H5-6, which proposed that the effects mentioned in H1-4 would be moderated by the NFA (H5) and the NFC (H6), were not accepted. NFA did not moderate the effect of the blogger on source credibility (b* = .35, p = .309), nor on source resistance (b* = -.13, p = .705), and it did not moderate the effect of the health organisation on source credibility (b* = -.27, p = .403) nor on source resistance (b* = -.06, p = .850). NFC did not moderate the effect of the blogger on source credibility (b* = .12, p = .643), nor on source resistance (b* = -.24, p = .349), and it did not moderate the effect of the health organisation on source credibility (b* =
.31, p = .269), nor on source resistance (b* = -.42, p = .118). The final model explained 0.5% of the variance in SSB consumption, indicating a small to no effect size.
Mediation analysis. To test whether the effects of the blogger and organisation and their
interactions with NFA and NFC were indeed mediated by source credibility, source
resistance, and the TPB constructs, the initial model was compared with models in which the independent and interaction variables directly affected behavioural attitude and SSB
consumption. Allowing the blogger to directly influence behavioural attitude significantly increased model fit, χ2diff (1) = 4.16, p = .0428. The blogger had a significant negative effect on behavioural attitude (b* = -.11, p = .040), indicating that the presence of a blogger led to a more negative behavioural attitude, and this effect was not mediated by source resistance or source credibility. All other effects on behavioural attitude and SSB consumption were non-significant, indicating that the found effects were mediated.
Dutch and American samples. Answering SQ1, the Dutch and American samples were
compared regarding H1-6. For the American model, the results were similar to the combined sample reported above, explaining 0.1% of the variance in SSB consumption. See Table A6 for an overview of the results. In contrast, different results were found for the Dutch model, explaining 2.8% of the variance in SSB consumption, indicating a small effect size.
Therefore, the differences between the Dutch and combined sample are discussed further. As for the combined sample, a significant positive effect was found of the behavioural attitude on the behavioural intention (b* = .38, p < .001), indicating that a more positive behavioural attitude led to a higher behavioural intention. In addition, contrasting the findings of the combined sample, a significant negative effect was found of behavioural intention on SSB consumption at T2 (b* = -.17, p = .018), indicating that a higher behavioural intention led to lower SSB consumption.
8 Model fit after adding the effect of the blogger on behavioural attitude was: χ2(283) = 528.57, p < .001; CFI = .98; RMSEA = .05 90%CI [.05, .06].
Moreover, NFC marginally significantly moderated the effect of the blogger on source credibility (b* =.67, p = .051) and significantly moderated the effect of the blogger on source resistance (b* = -.67, p = .037). To further explore these interaction effects, a flood-light analysis was performed. Further exploring the interaction effect of the blogger and NFC on source credibility yielded no significant results. The effect of the blogger on source credibility remained insignificant on all levels of the NFC. However, further exploring the interaction effect of the blogger and NFC on source resistance yielded significant results. For participants with a low NFC (< 2.68), there was a significant positive effect of the blogger on source resistance (.71 < b* < .36, .034 < p < .050), indicating that the blogger led to a higher source resistance for participants with a low NFC. However, this effect became insignificant for participants with a higher NFC (> 2.68; b* > .71, p > .050).
Discussion Discussion of Results
This study aimed to identify how a blogger (versus no blogger) and health organisation (versus no health organisation) of a HLB affect reducing SSB consumption among Dutch and American young adults. Also, it examined how the underlying mechanisms (source credibility and source resistance) and moderators (NFA and NFC) play a role in this effect.
No effect was found of the blogger and health organisation on source credibility nor on source resistance, in contrast to the expectations and previous studies (e.g., Jenkins &
Dragojevic, 2011; Klimmt et al., 2009; Neubaum & Krämer, 2015). There are several possible explanations for these insignificant results. Firstly, it may be explained by the target group which was not specifically focused on blog-readers. Blog-readers follow the blog for a longer time and build a (one-sided) trust relationship with the blogger (Colliander & Erlandsson, 2015). Therefore, the blogger can be perceived as more credible, there can be less resistance towards the source, and the message persuasiveness can be higher for blog-readers following
the blog for a longer time. However, it is unknown whether this relationship exists in the current study. HLB-readers are mostly young, highly educated women (Fox & Jones, 2009), indicating that readers have different characteristics than non-readers. Apart from blog-readers’ age, gender, and educational level, it is unknown what other (personal) characteristics differentiate blog-readers from the general population. Therefore, it is important to understand who reads HLBs and repeat this study, in future research, with a sample of blog-readers or to compare regular blog-readers with non-blog-readers. This might change the results and would provide scholars and healthy living bloggers with a deeper understanding as to what aspects of the blog influence behaviour change among readers and whether a fit between the HLB and the reader must be ensured for an effective behaviour change.
Secondly, the lack of results can be explained due to the extreme scores of the relatively high source credibility (M = 5.36, SD = .98 on a seven-point scale) and low source resistance (M = 2.64, SD = 1.11 on a seven-point scale) observed in this study. This may indicate that a ceiling of high source credibility and low source resistance has been reached and therefore, the source did not affect the mediators. Furthermore, low source resistance may indicate that participants did not perceive the message as a persuasive attempt. This is a promising point for healthy living bloggers: if the reader is unaware of the persuasive attempt, less resistance is experienced (Knowles & Lin, 2004), which leads to more health behaviour change (Moyer-Gusé, 2008). Persuasion knowledge, the degree to which one recognises a persuasion attempt, depending on the subject of the message (Van Reijmersdal et al., 2016), may play a role in the effect on source resistance and could explain the insignificant results. It is recommended that researchers explore the role of persuasion knowledge as a mediator in the future in order to understand when a blogger and an organisation increase source resistance. It is expected that lower persuasion knowledge would lead to a lower resistance and higher persuasion (Van Reijmersdal et al., 2016).
In contrast to the expectations (Neubaum & Krämer, 2015), a mediation analysis showed that there was a direct, significant, negative effect of the blogger on behavioural attitude, indicating that the blogger led to a more negative behavioural attitude. Since this effect was not mediated by source resistance and source credibility, it must be studied further in terms of underlying mechanisms.
In line with the expectations (Ajzen, 1991), the behavioural attitude had a significant positive effect on the behavioural intention, indicating that a more positive behavioural attitude led to a higher behavioural intention. However, no significant effect was found of the behavioural intention on SSB consumption at T2 for the combined, as well as the American, sample. For the Dutch sample, a negative effect was found of behavioural intention on SSB consumption at T2, indicating that a higher behavioural intention led to a lower SSB
consumption at T2 for Dutch participants only. These differences can be explained by the differences of SSB consumption in T0, where the American sample had a significantly higher SSB consumption (M = 689.39, SD = 620.94) than the Dutch sample (M = 369.67, SD = 399.69). The non-significant results for the American sample can be explained by the well-known intention-behaviour gap (Sheeran, 2002). Due to other personal and external factors, the behavioural intention is often not translated into actual behaviour. Daily behaviours such as SSB consumption become habits that do not require much intentional effort and they predict behaviour better than behavioural intentions (De Bruijn & Van den Putte, 2009; Kremers et al., 2007; Sheeran, 2002). Use of implementation intentions can assist with overcoming habitual behaviour such as SSB consumption, by providing context (i.e., a time and place) of behaviour (Sheeran, 2002). Implementation intentions are formed as “I intend to do X when situation Y arises” and focus on a more specific goal compared to behavioural intentions (Sheeran, 2002). Therefore, in future research it is important to study
providing participants the opportunity to form their own implementation intentions (i.e., “I intend to drink water instead of coca cola after lunch at work”) and study their effects on SSB consumption. If SSB consumption is indeed predicted mostly by habits, habits may explain the differences between the samples. SSB consumption is more difficult to change with a stronger habit (i.e., higher SSB consumption among the American versus Dutch population).
Lastly, in contrast to the expectations (Resnicow et al., 2008; 2014), no moderation effects of the NFA and NFC on the effect of the blogger and health organisation on the source
credibility or source resistance were found. Thus, the effect seemed to be similar for people with all levels of NFA and NFC. However, for the Dutch sample, two moderation effects were found: a marginally significant moderation effect of the NFC on the effect of the blogger on source credibility, and a significant moderation effect of the NFC on the effect of the blogger on the source resistance. After further exploring the moderation effects using a flood-light analysis, it seemed that for participants with a low NFC, there was a positive effect of the blogger on source resistance, indicating that the blogger led to higher source resistance for participants with a low NFC. This effect was insignificant for participants with a high NFC. No differences were found for different levels on the NFC for the effect of the blogger on source credibility.
Practical Implications and Future Research
The findings of this study have several practical implications for health communication professionals and healthy living bloggers and scientific implications for future research. Apparently, the presence of a blogger or a health organisation did not matter for the source credibility and source resistance, aside from Dutch participants with a low NFC who had a higher resistance towards the blogger. However, the blogger had a negative direct effect on the behavioural attitude. This implies that the presence of a blogger, as presented in the setting of this study, might produce boomerang effects (i.e., performing the unwanted
behaviour; Burgoon et al., 2002). Thus, a blogger might not be the right source for a behaviour change message in this format. The other insignificant effects imply that the source, as presented in this study, may be unimportant for behaviour change. Instead, the website on which it is published may be more important. Since blogs are often followed closely by their readers (Ballantine & Yeung, 2015; Colliander & Erlandsson, 2015), it is assumed that blogs are more often read than health organisations’ websites, which are usually consulted for specific information. It is important to study the source of health information in terms of where it is published. While prior studies have compared health information on blogs and websites (e.g., Hu & Sander, 2010; Neubaum & Krämer, 2015), none have compared the exact same text on the blog and the website, and none were studied in the context of SSB consumption. Therefore, health information about SSB must be studied comparing the effectiveness of an existing HLB and a health organisation’s website. Results would indicate whether the source plays an important role for reducing SSB consumption. Furthermore, since the blog was fictitious, the same study should be conducted with a well-known blogger or celebrity. Celebrities often have a high source credibility and are often used by marketers to promote products (Erdogan, 1999). Similarly, celebrities may be used for health
communication purposes in such a way that they write a HLB following their own healthy lifestyle (change), which may have a higher impact on readers than a fictitious blog. Future research should focus on studying the differences between the persuasiveness and initiation to behaviour change of a celebrity versus a lay-person, such as a blogger.
Since clear differences were found between the Dutch and American sample, it is implied for health communication professionals that the two populations must be approached
differently in order to achieve behaviour change. For example, the effect of source resistance on the behavioural attitude was significant for the Dutch participants with a low NFC, but not for the American sample. This implies that Americans perhaps overcome resistance
differently, for example, by a boomerang effect (Burgoon et al., 2002). Thus, the underlying mechanisms must be tested more extensively before creating (different) messages for both populations.
Strengths and Limitations
An important strength of this study is the use of two panels from two countries (i.e., the Netherlands and the U.S.) for cross-validation of the data. The study showed that, while some aspects were comparable between the samples, others differed. This makes it more likely that the results are mostly generalisable to other (Western) countries. However, the differences should be looked at more closely in future research. Moreover, this was a longitudinal study with three measurement moments conducted over two weeks. Since one exposure has minimal effects on behaviour change (Boiarsky et al., 2013), it included two blog-posts exposures and an opportunity to measure SSB consumption after two weeks. Thus, a main strength of the current study is that we cannot only draw conclusions regarding the
behavioural attitude and intention, but also regarding the behavioural outcome. Some limitations of this study must be taken into account as well. Firstly, SSB
consumption was self-reported, which may be over- or underestimated by participants (De Bruijn & Van den Putte, 2009). However, to reach large audiences, the SSB consumption of each participant cannot be measured objectively by observations. The question of SSB consumption required participant to remember only the last week, thus, they were not
expected to think back and so, the over- or underestimation should be minimal. Secondly, the HLB’s content discussed specifically SSB consumption, which is perhaps not an interesting topic for everyone. This may reduce attention to the message (Kremers et al., 2007) and therefore, having no effect on SSB consumption. However, neither attention nor topic interest were measured. Lastly, lifestyle change does not occur within two weeks, since it is a process that can take months (Boiarsky et al., 2013). Most longitudinal studies assess the behaviour
change after several months or even years (Malik et al., 2006). Nonetheless, two weeks can show the start of a behaviour change and indicate its direction, which was initially one of the greatest strengths of this study.
Despite the limitations, this study shows that while the blogger had a negative effect on behavioural attitude, the organisation probably does not influence the message
persuasiveness. The behavioural attitude positively influenced the behavioural intention, and eventually led to lower SSB consumption among the Dutch population, giving some hope for a healthier society and a decrease in obesity rates. As the World Health Organisation (2017a) stated, the individual is not the only one responsible for his/her health, these healthy lifestyle behaviour changes must be facilitated on a societal level by making healthier beverage choices more widely available.
References
Ahluwalia, R. (2000). Examination of psychological processes underlying resistance to persuasion. Journal of Consumer Research, 27(2), 217-232.
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior & Human Decision Processes, 50(2), 179–211.
Anderson, E. L., Steen, E., & Stavropoulos, V. (2017). Internet use and problematic internet use: A systematic review of longitudinal research trends in adolescence and emergent adulthood. International Journal of Adolescence and Youth, 22(4), 430-454.
doi:10.1080/02673843.2016.1227716
Ballantine, P. W., & Yeung, C. A. (2015). The effects of review valence in organic versus sponsored blog sites on perceived credibility, brand attitude, and behavioural
intentions. Marketing Intelligence & Planning, 33(4), 508-521. doi:10.1108/MIP-03-2014-0044
Boepple, L., & Thompson J. K. (2014). A content analysis of healthy living blogs: Evidence of content thematically consistent with dysfunctional eating attitudes and behaviors. International Journal of Eating Disorders, 13(4), 362–367. doi:10.1002/eat.22244 Boiarsky, G., Rouner, D., & Long, M. (2013). Effects of responsibility attribution and
message source on young adults' health attitudes and behaviors. Journal of Health Communication, 18(7), 881-894. doi:10.1080/10810730.2012.757389
Boncinelli, F., Gerini, F., Pagnotta, G., & Alfnes, F. (2016). Warning labels on junk food: Experimental evidence. International Journal of Consumer Studies, 1-8.
doi:10.1111/ijcs.12312
Brehm, J. W. (1966). A theory of psychological reactance. New York: Academic Press. Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance: A theory of freedom and
Brown, P., & Levinson, S. C. (1987). Politeness: Some universals in language usage. Cambridge, UK: Cambridge University Press.
De Bruijn, G. J., & Van den Putte, B. (2009). Adolescent soft drink consumption, television viewing and habit strength. Investigating clustering effects in the Theory of Planned Behaviour. Appetite, 53(1), 66-75. doi:10.1016/j.appet.2009.05.008
Burgoon, M., Alvara, E., Grandpre, J., & Voulodakis, M. (2002). Revisiting the theory of psychological reactance: Communicating threats to attitudinal freedom. In J.P. Dillard & M. Pfau (Eds.), The Persuasion Handbook: Developments in Theory and Practice. California, Sage Publications.
Byrne, B. M. (2013). Structural equation modeling with AMOS: Basic concepts, applications, and programming. New York, NY: Routledge.
Caplette, M. E., Provencher, V., Bissonnette-Maheux, V., Dugrenier, M., Lapointe, A., Gagnon, M. P., …, Desroches, S. (2017). Increasing fruit and vegetable consumption through a healthy eating blog: A feasibility study. JMIR Research Protocols, 6(4), e59. doi:10.2196/resprot.6622
Choi, J., Miao, L., Almanza, B., & Nelson, D. (2013). Consumers’ responses to restaurant inspection reports: The effects of information source and message style. Journal of Foodservice Business Research, 16(3), 255-275. doi:10.1080/15378020.2013.810536 Colliander, J., & Erlandsson, S. (2015). The blog and the bountiful: Exploring the effects of
disguised product placement on blogs that are revealed by a third party. Journal of Marketing Communications, 21(2), 110-124. doi:10.1080/13527266.2012.730543 Deci, E. L., & Ryan, R. M. (2000). The" what" and" why" of goal pursuits: Human needs and
the self-determination of behavior. Psychological Inquiry, 11(4), 227-268. doi:10.1207/S15327965PLI1104_01
Eastin, M. S. (2001). Credibility assessments of online health information: The effects of source expertise and knowledge of content. Journal of Computer-Mediated Communication, 6(4). doi:10.1111/j.1083-6101.2001.tb00126.x
Erdogan, B. Z. (1999). Celebrity endorsement: A literature review. Journal of Marketing Management, 15(4), 291-314. doi:10.1362/026725799784870379
Faul, F., Erdfelder, E., Buchner, A., & Lang, A. G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41, 1149-1160. doi:10.3758/BRM.41.4.1149
Fishbein, M., & Cappella, J. N. (2006). The role of theory in developing effective health communications. Journal of Communication, 56, s1-s17. doi:10.1111/j.1460-2466.2006.00280.x
Fox, S., & Jones, S. (2009). The social life of health information. Washington, DC: Pew Internet and American Life Project. Retrieved from
http://www.pewInternet.org/~/media//Files/Reports/2009/PIP_Health_2009.pdf Fransen, M. L., Verlegh, P. W., Kirmani, A., & Smit, E. G. (2015). A typology of consumer
strategies for resisting advertising, and a review of mechanisms for countering them. International Journal of Advertising, 34(1), 6-16. doi:10.1080/02650487.2014.995284 Gallagher, K. M., & Updegraff, J. A. (2012). Health message framing effects on attitudes,
intentions, and behavior: A meta-analytic review. Annals of Behavioral Medicine, 43, 101-116. doi:10.1007/s12160-011-9308-7
Green, M. C. (2006). Narratives and cancer communication. Journal of Communication, 56, 163–183. doi:10.1111/j.1460-2466.2006.00288.x
Harvard (2012). Sugary drinks and obesity fact sheet. Retrieved from
Hovland, C. I., & Weiss, W. (1951). The influence of source credibility on communication effectiveness. The Public Opinion Quarterly, 15(4), 635-650.
Hu, Y., & Sundar, S. S. (2010). Effects of online health sources on credibility and behavioral intentions. Communication Research, 37(1), 105-132. doi:0.1177/0093650209351512 Hummel, K., Nagelhout, G. E., Willemsen, M. C., Driezen, P., Springvloet, L., Mons, U., …,
De Vries, H. (2015). Trends and socioeconomic differences in policy triggers for thinking about quitting smoking: Findings from the International Tobacco Control (ITC) Europe Surveys. Drug and Alcohol Dependence, 155, 154-162.
doi:10.1016/j.drugalcdep.2015.07.678
Jenkins, M., & Dragojevic, M. (2011). Explaining the process of resistance to persuasion: A politeness theory-based approach. Communication Research, 40(4), 559-590.
doi:10.1177/0093650211420136
Jordan, A., Piotrowski, J., Bleakley, A., & Mallya, G. (2012). Developing media interventions to reduce household sugar sweetened beverage consumption. The Annals of the
American Academy of Political and Social Science, 640(1), 118-135.
doi:10.1177/0002716211425656
Klimmt, C., Hefner, D., & Vorderer, P. (2009). The video game experience as “true” identification: A theory of enjoyable alterations of players’ self-perception. Communication Theory, 19, 351-373. doi:10.1111/j.1468-2885.2009.01347.x Kline, R. B. (2011). Principles and practice of structural equation modeling (3rd ed.). New
York, NY: Guilford Press.
Knowles, E. S., & Lin, J. A. (2004). The importance of resistance to persuasion. In E. S. Knowles & J. A. Linn (Eds.), Resistance and Persuasion. Mahwah, New Jersey: Lawrence Erlbaum Associates.
