The effect of using ethical branded content on customers’ engagement, the role of COVID-19 panic

The effect of using ethical branded content on customers’

engagement, the role of COVID-19 panic

Author: yidai WU

S2948095

The effect of using ethical branded content on customers’

engagement, the role of COVID-19 panic

Author: yidai WU

S2948095

University of Groningen, Faculty of Economics and Business, M.Sc. Marketing

Master Thesis

Completion date: 11th of January 2021

Abstract:

With the increasing importance of branded content in companies’ modern business strategy, the drivers of the effectiveness of branded content have been brought under discussion. This paper focuses its research on the causal effect of ethical branded content on customer engagement of the advertisement, conditional on the role of health involvement. Besides, the mediation effect of comments credibility between the relationship of ethical branded content on customer engagement of the advertisement is discussed. Moreover, the moderation effect of stress on the relationship between ethical branded content and comments credibility are taken into consideration as well. The data was obtained by a 2×2 between subject experiments. The survey of the experiments is created by Qualtrics and delivered online via a personal link. In total, 259 valid responses are being used in the analysis. The output of the regression analysis shows that ethical branded content has positive effects on comments credibility and customer engagement of the advertisement even though health involvement is controlled. The comments credibility shows a positive effect on customer engagement of the advertisement and mediates the relationship between ethical branded content and customer engagement of the advertisement. However, there is no evidence that stress would enhance the positive effect of ethical branded content on comments credibility. Thus, no moderation effect was found.

Preface:

Table of Contents

1.Introduction ... 5

2. Theoretical Framework ... 7

2.2.1 Customer engagement ... 8

2.2.2 Ethical content in advertisement ... 8

2.2.3 Comments credibility ... 9

2.2.4 Stress ... 12

3. Experiment design and survey ... 14

3.1 Design of the experiment ... 14

3.2. Pre-test for the posts design and results ... 16

3.3 Survey data collection ... 19

3.4 Construct Measurement ... 20

3.4.1 Measurement scale ... 20

3.4.2 Factor analysis and reliability check ... 20

3.5 Manipulation re-check ... 23

3.6. Correlation matrix and regression test for control variable ... 24

3.7 Hypothesis testing ... 25

4. Output analysis ... 27

4.1 VIF testing ... 27

4.2.1 Ethical branded content and customer engagement of the advertisement (H1) ... 27

4.2.2 Ethical branded content and comments credibility (H2) ... 27

4.2.3 Comments credibility and customer engagement of the advertisement (H4) ... 27

4.2.4 The mediating effect of comments credibility (H3) ... 28

4.2.5 The moderating effect of stress level on the relationship between ethical branded content and comments credibility (H5). ... 29

4.2.6 Moderated mediation ... 30

5. Conclusion and discussion ... 31

5.1 Conclusion and discussion ... 31

5.2. Management implication, limitation and suggestion for further research ... 33

6. Reference ... 35

1.Introduction

The message of branded content is usually displayed in the form of an integration of advertising and article content (Harms et al. 2017). Opposite to the traditional banners or pop-ups on TV or web which gives customers a strong feeling of persuasion (Friestad et al. 1994), branded content is often embraced naturally by the design and style of the platform that it is placed in, encouraging consumers to seek for information about the advertised product itself voluntarily (Harms et al. 2017). Consequently, advertised brands communicate to their target groups more effectively thanks to the application of different types of platforms, which would bring potential value to all parties involved, e.g., advertised brands, the media platforms and consumers. For instance, branded content expands the traditional advertising content and brings additional value for customers (Tutaj et al. 2012). Besides, the utilization of the branded content also brings an income stream to the service provider such as publishers and platforms (Probst et al. 2013). According to Risselada et al. (2014), branded content has a stronger sharing influence than traditional ones thanks to the ‘word-of-mouth’ effect. Berger et al. (2012) found that the sharing effect is stronger when the online content arouses strong positive and negative emotions compared to those which only lead to low emotions. Due to the large benefits to different parties concerned, branded content has been gaining more and more importance in the field of marketing. According to Perrin (2019), advertisers are expected to spend almost around 70% of the total display budget on native ads in 2020 and the percentage will keep on increasing in the future.

Besides the ethical information, stress may be an important factor that influences customers' engagement of the advertisement. Perceived feelings of stress are found to be positively related with level of depression and anxiety (Finkelstein et al. 2007). The feeling of stress comes when consumers perceive an imbalanced relationship between them and the environment, at the time when consumers feel a lack of resources or when their wellbeing is threatened (Lazarus & Folkman, 1984, p 21, Maier & Wilken, 2014). Researchers found that when the consumers experience a feeling of fear or stress, they are more willing to accept the recommendation and take action to change (Fennis et al. 2015).

This thesis focuses on analyzing the causal effect of using ethical information in branded content on customer engagement of the advertisement in the food industry. Besides, a mediation effect of ‘comments credibility of ethical advertisements’ on ‘customer engagement of the advertisement’ will be analyzed. Researchers found the more ethical consumers perceive the radio story, the more credible they feel the story is (Wei et al. 2008). Montano et al. (2015) found the more the consumers believe the advertisement, the more they are willing to share it to others. Thus, a mediation effect of comments credibility between the relationship of ethical content on customer engagement will be tested as well. Furthermore, Fennis et al. (2015) found that consumers are more willing to follow the recommendation and change their behavior when they experience a feeling of stress. Thus, stress (being or not being exposed to the Corona news) is assumed to moderate the relationship between ethical content on comments credibility.

hypothesis will be discussed. Then, in section 3, the experiment design, data collection and constructs measurement used in the survey for this thesis will be presented. Furthermore, in section 4, the data descriptive and the regression output of the model will be presented and discussed. Finally, in section 5, the conclusion, discussion and limitations of this thesis will be provided.

2. Theoretical Framework

Figure 2.1 shows the conceptual model with the variables ethical branded content, comments credibility, customer engagement of the advertisement and stress.

F

is predicted that the relationship between ethical content and comments credibility is influenced by customers’ feelings of stress, namely, after being exposed to the Corona news, consumers are more stressed and thus, they are more likely to believe the credibility of the comments to the ethical information in the advertisement.

In addition to the main variables, the variable health involvement is used as a control variable since it will also impact the relationship between ethical content, comments credibility and customer engagement of the advertisement. For instance, eating fish has been promoted by companies and governments as healthy eating as common sense. Researchers found the more health conscious the consumers, the more often they consume fish (Olsen et al. 2003). Thus, consumers with higher health involvement are supposed to like ethical branded content better than those with lower involvement of health since ethical products are considered healthier (Williams et al. 2001). Besides, researchers found consumers seeing the comments from expert sources will believe the online advertising more credible than ones who did not (Pjesivac et al. 2018). Since consumers with higher health involvement usually have more knowledge on nutrition and the quality of the product (Simpson et al. 1998), a positive effect of health involvement on comments credibility is suspected as well.

2.2 Hypothesis

2.2.1 Customer engagement

Customer engagement is used in this model to indicate the effectiveness of branded content. In order to build and maintain deeper, more meaningful and more sustainable interactions with customers, a high customer engagement is desired by the company (Unit, Economist Intelligence, 2007). A higher rate of customer engagement represents a higher effectiveness of the branded content, e.g., higher customers’ willingness to like, comment and share the branded content advertised.

2.2.2 Ethical content in advertisement

are healthier and safer for themselves. Also, consumers believe ethical products produce less pollution to the environment and avoid child labor during production. For instance, in the food industry, the ethical information is often related to the word ‘organic’ as within the organic sector, some production procedures are required to follow requirements which are even stricter than the ethical aspects covered in EU regulations (Zander et al. 2010). What’s more, researchers have found that people choose to buy ethical products or behave good because they expect to gain some tangible or intangible things in return, e.g., status, respect etc. According to Griskevicius et al (2010), the status motive theory suggests consumers are more likely to shop green products in public than on private occasions because they want to be seen when they behave good. Therefore, we expect that there exists a positive relationship between the ethical content and consumers engagement of the advertisements. Consumers are more willing to like, comment, or share an advertisement with ethical information because they have the motivation to live healthy or want to maintain a positive self-image by doing so.

H1: Branded content with ethical information has a larger positive effect on consumers engagement of the advertisement than branded content with no ethical information.

2.2.3 Comments credibility

feeling of reliability. Thus, it is suspected that the use of ethical information will increase consumers’ perception of the online reviews’ credibility. Consequently, the second hypothesis predicts:

H2: Branded content with ethical information has a larger positive effect on comments credibility than branded content with no ethical information.

Besides, comments credibility is predicted to have a positive effect on customer engagement of the advertisement. Firstly, the online comment, being considered as a novel type of word-of-mouth information on the product and usage experience, keeps on increasing its importance in customer engagement of the advertisement (Chen et al. 2008). Since the online comments are posted by the individual customers to show their personal experience and opinion about the product (Zhou et al. 2016), they have been found to be more credible for customers than the retailer self-made information (Bickart et al. 2001). Second, Garbarino et al. (1999) found that the more credible consumers feel about the product, the more likely they will give positive comments and stay with the company in the long run. Researchers have also found that consumers are more willing to give positive comments about a product or a company when they feel the employees are reliable (Gremler et al. 2001). It is predicted that the more credible consumers feel about the comments, the more likely they will like, share or comment the advertisement. Thus, the fourth hypothesis predicts:

H4: comments credibility has a positive effect on customer engagement of the advertisement.

indicated they don’t believe the products with ethical labels had different taste, though, they still report the chocolate with ethical labels tastes better because of this psychological effect. Second, according to the reasoned action theory, a person’s attitude is one of the determinants to the intention of that behavior (Montaño et al. 2015). The customers’ willingness to share or like the branded content can be partly explained by consumers’ high expectation on ethical products. Since they expect the organic products to be safer and healthier than non-organic products (Kavaliauske et al. 2014), they are more likely to believe the comments below, and thus, more willing to like and share the advertisement to others. Third, Berman et al. (2015) discusses the braggart’ dilemma in their paper. Bragging or advertising about one’s prosocial behavior shows a negative influence when the prosocial behavior is already known by the public and perceived to meet an individual's selfish goals such as showing off. However, they argue that in the situation when one publicly displays generosity but at the same time try to elicit the same prosocial behavior from others i.e., giving positive comments and recommending it, people will appreciate the recommendation more than attack it because they are more likely to believe the credibility of the reviews. As a result, consumers are expected to be more willing to show a positive attitude to the reviews and engage in the content of the advertisement with ethical content. Lastly, the social conformity theory (Cialdini et al. 2004) suggests three core drivers that influence consumers’ choice to conform to surroundings, i.e., informational, normative social influence and persuasion. According to this theory, consumers choose to like and share the advertisements because they believe the majority is right or because they want to be liked by their peers and keep a good self-concept by doing so. Since the ethical content increases consumers’ perception of the content credibility (Nelson et al. 2015), consumers are expected to be more likely to like or share the advertisement because they think it's right and good to engage in it.

To conclude, a partial mediation effect of comments credibility is expected between the relationship of ethical content on customer engagement of the advertisement. The use of ethical information in the branded content tends to increase consumers believes of the comments to the advertisement. The higher believes of the positive comments will further increase their willingness to engage in the advertisement. Thus, the third hypothesis predicts as follow:

2.2.4 Stress

The level of stress is hypothesized to influence the relationship between the use of ethical content and comments credibility. The consumer becomes stressful when an imbalanced relationship occurs and consumer feels a lack of resources or his/her well-being is threatened (Lazarus & Folkman, 1984, p. 21, Maier & Wilken, 2014). Maier et al. (2014) conducted their experiment based on construal level theory to test consumers evaluation of the product features’ importance. The construal level theory mainly states that people’s perception of an object depends on his/her psychological distance to it. The closer the objects are to you. the more concrete you will think of it. However, the more distance between you and the object, the more abstract you will perceive it (Trope et al. 2010). Maier et al. (2014) find in their research that when feeling stressed, consumers are more likely to put their attention on low level construal characteristics which do not acquire further cognitive effort and perceive them to be more important. For example, when evaluating a camera, consumers with stress tend to put their attention on detailed information, e.g., weight, rather than on the overall usage of the camera. It is supposed after being exposed to the Corona information, consumers' stress level is increased. As a result, consumers tend to put more attention on information in advertisements which provides them a direct and concrete link to overcome this stressful situation as they feel the stressful situation is close to them. Since researchers have found ethical products are believed to be healthier and safer by the consumers (Williams et al. 2001), it is predicted that under the circumstance of feeling stressed about COVID-19, higher comments credibility of the advertisement with ethical information would be perceived by consumers compared to the case under the situation with no stress.

supposed that consumers will increase their intention to find preventive solutions. Thus, consumers are more likely to believe the comments of advertisements with ethical information.

The fear and anxiety mechanism suggests that with the life-threatening situation, public attention to the advertisement will increase because it is more significant for survival (Robinson, 1998). The feeling of anxiety and stress can be generated unconsciously as implicit memory during the pre-attentive stage when consumers process information (Robinson,1998). Though scholars argue consumers' further attitude towards the advertisement will change for the reason that the deeper processing of information will lead to longer storage of the information in memory (Fennis et al. 2015), some researchers found that the consumers’ willingness to accept the recommendation increases with their increasing fear (Witte et al. 2000). Thus, it is suspected that consumers are more willing to believe and accept the comments after they are exposed to the COVID-19 news when they feel worried about health. Besides, the feeling-as-information theory suggests that consumers tend to use their mood, emotions and other subjective experience as input to make other choices when they are not aware their mood and emotion are correlating with their choices (Schwarz et al. 2011). They found people reported themselves on sunny days more satisfied with life than on rainy days, while they feel more depressed on rainy days than sunny days. Also, Russell. (1980) found in the Affect Circumplex Model that when consumers feel positive, they tend to keep this mood. However, when they feel negative, they tend to take actions to change it. Thus, based on these theories, it is expected that when feeling stressed after seeing the Corona news, consumers are more likely to believe the comments of the advertisement because they want to take action to change.

It is expected after consumers being exposed to the Corona news, they are more likely to believe comments because they want to put action to avoid the threat and be healthy. What’s more, the positive relation between branded content and comment credibility is strongest under the circumstance of feeling worried after being exposed to COVID-19 news, while at the same time, there is ethical information presented in the advertisement. Under such a case, consumers are most likely to notice the ethical information in the advertisement and believe the comments because they want to take actions to avoid getting coronavirus themselves. Thus, the fifth hypothesis predicts:

3. Experiment design and survey 3.1 Design of the experiment

In this thesis, an experimental design is conducted and a questionnaire linked to the design is followed in order to get the data for further analysis. The manipulation of two variables, namely the usage of ethical information in branded content and the application of stress arousing information, leads to four conditions which are presented by the figures from 3.1 to 3.4. To be more specific, a 2 (ethical information in branded content - 1 = Yes vs. 0 = No) ×2 (Stress arousing information - 1 = Yes vs. 0 = No) between subject experimental design was conducted.

The pictures used in the manipulation were taken from the original Instagram account from brand ‘Innocent’ and Dutch News.nl. First, in order to prepare the posts for the experiment manipulation of the ethical information, the text on the sign from the original picture was changed to ‘This is a smoothie of

mixed organic fruit’ and the text on the bottle was changed to ‘Mixed organic fruit’ in figure 3.1 to

represent the case with ethical information in branded content. On the other hand, to manipulate the case with no ethical information in branded content, the text on the sign was adjusted to 'This is a

smoothie of mixed fresh fruit’ and the text on the bottle was adjusted to ‘Mixed fresh fruit’. What’s more,

in order to get rid of the bias from the original brand name ‘Innocent’, which would increase consumers’ perception on the ethical level, a fake name ‘Fresh’ altered the original name of the drink. Also, all the comments stay the same on both occasions. All other possible biases are controlled for in Figure 3.1 and 3.2; only the text on the bottle and the text on the sign beside the bottle are different.

Figure 3.1: advertisement with ethical information

Figure 3.3: news with stress evoking Corona information

Figure 3.4: news without stress-evoking Corona information

3.2. Pre-test for the posts design and results

or 3.2 randomly. The answer is based on a 7- point Likert scale. Next, the respondents are required to answer the question ’To what extent do you see stress - arousing information?’ with regard to text in figure 3.3 or 3.4 at random. The answer is also based on a 7 Likert scale. The text news and related questions are provided to the participants directly after they finish the previous question on the level of perceived ethical information as if they don’t need to do anything.

stress - arousing (Yes) stress - arousing (No) Ethical information

Yes 10 10

No 10 10

Table 3.1 Pretest of the design

The Levene’s test for equality of variance should be taken into consideration to decide whether or not the equal variance should be assumed when interpreting the results. The p - value of the test for equality of variance with regard to perceived ethical level is 0,588, which is way larger than 0,05. Thus, we accept the null hypothesis that the variances of the two groups are equal, i.e., the distribution of the perceived ethical level for contents with no ethical information is in similar shape to the distribution of the perceived ethical level for contents with ethical information. Thus, the first line in the corresponding output table is going to be interpreted.

Similarly, the p - value of the test for equality of variance with regard to stress level is 0,620, which is way larger than 0,05. Thus, we also accept the null hypothesis of the equal variance and choose to read the first line in the corresponding output table as well.

Similarly, on average, compared to the stress level in the group with no stress arousing information in the news, the consumers rate 4.5 score higher on stress level when the news contains stress-arousing information. The result is also significantly different between the 2 groups since the p-value is much smaller than 0,05 (p = 0,000).

The result of the manipulation shows that the differences of the means between compared groups are significant and thus, the manipulation check is successful. The posts can be added to the survey to create the data set for further analysis.

Besides, the same manipulation check questions will be included in the survey and the result will be reported later in section 3.5.

Perceived ethical information level (based on 1-item question) ethical information

N Mean S. D Sig (2 tailed), 95% confidence interval 0(no) 20 2,70 0,979 0,000 1(yes) 20 5,90 0,968 Stress level (based on 1-item question) Fear arousing information

N Mean S. D Sig (2 tailed), 95% confidence interval

0(no) 20 1.60 0,883 0,000

1(yes) 20 6.10 0,912

3.3 Survey data collection

The survey was created by Qualtrics and distributed via online links to friends and family members. Some respondents helped further forward their survey to classmates or workmates in China. As a result, a majority of the respondents are students in various universities or freshmen white collar workers in companies in China.

Before starting the formal survey, an introduction of the survey was presented to respondents, stating the goal of the survey and that the data being collected will be completely anonymous. Besides, respondents are informed that they can quit the survey anytime if they don’t want to participate. Afterwards, a short instruction was presented to participants: ‘please read the following 2 posts carefully and answer the following questions.’ The participants are then exposed to one of the four conditions composed by figure 3.1 to figure 3.4. After being exposed to one of the four conditions, participants are required to fill in the survey about the questions related to the construct which will be discussed in detail in part 3.4 later in this thesis. On the last page of the survey, a thank you was given to the participants for their time and effort filling in the survey.

In total, 259 valid responses are used for the data analysis. There are 49 values missing for ‘age’, 17 missing values omitted for ‘vegetarian’, 17 values missing in ‘gender’. Though there are missing values, the cases related are still captured for analysis because these variables are only used for describing the sample. The distribution of respondents over the four conditions in the experiment is shown in table 3.3.

Ethical Fear N

No Yes 54

No No 75

Yes Yes 57

Yes No 73

Table 3.3. Distribution of valid respondents in 4 conditions

3.4 Construct Measurement

3.4.1 Measurement scale

Table 3.4 shows the items for the construct measurement. All the items are measured based on a 7 Likert scale. Besides, the factor loading and Cronbach’s Alpha of each item are presented in the table as well. The control variable Health Involvement is measured by 3 items in the survey. Furthermore, at the end of the survey, respondents are required to provide their background information by indicating their gender in terms of ‘male, female and other’, whether they are vegetarian or flexitarian and filling in their age.

3.4.2 Factor analysis and reliability check

In order to test whether the items loaded on the respective construct are highly related to the construct, factor analysis is applied to each of the constructs. There are several reasons why a separate analysis is applied to each of the constructs instead of a combined factor analysis applying to all constructs: Factors analysis is used to reduce a large set of variables into small set of uncorrelated factors (Gijsenberg, 2020), namely, it is used to search for orthogonal solutions (factors do not correlate among each other). Whereas, in the conceptual model. We hypothesize a relationship between the dependent variable and the mediator, moderator and other variables. Therefore, conducting a combined factor analysis for all constructs might confound the results.

Next, in order to check whether a separate factor analysis can be applied to each of the constructs, 3 criteria have to be met: the KMO value > 0,5, the Bartlett’s test of Sphericity has to be significant and the communities has to be between 0.4 to 1 (Gijsenberg, 2020). The results show that the KMO of each construct showed a value above 0,5. The Bartlett’s Test of Sphericity of all constructs are found to be significant (p=0,000). Besides, the community value of each item is above 0,4. It indicates that a factor analysis can be applied to each of the separate constructs. The results in Table 3.4 show that all factor loadings are highly above 0,5, indicating all the items should be kept to their constructs (Gijsenberg. 2020).

factors found are strong enough. What’s more, the result of ‘Alpha if item deleted’ shows that regarding all items, the alpha of their related construct would all go down if any of the items was not in the test. This indicates that all questions are good since deleting any of the single questions would lower the overall alpha.

Construct source Items Factor

loading Cronbach’s Alpha Comments

credibility MacKenzie, S. B., & Lutz, R. J. (1989)

How convincing/unconvincing 0,816 0,735 how believable/unbelievable 0,809

how biased/unbiased they feel about the comments 0,800 Customer engagement of the advertisement

Wu, T. Y., & Atkin, D. J. (2018). Nikolinakou, A., & King, K. W. (2018). Adjusted from above

After reading the branded content, how willing would you be to post your comment on this branded content?

0,795 0,754

How willing would you be to share this branded

content? 0,851

How likely would you be to click ‘like’ for this branded

content? 0,817

perceived ethical

information Brunk, K. H (2012)

the product respects moral norms 0,827 0,748 the product is a socially responsible product 0,853

the product avoids damaging behavior at all costs

during production 0,765 Stress level

Reynolds, C. R., & _{Richmond, B. O.} (1978)

After reading the statement,

I feel I have trouble making up my mind 0,819 0,877 I feel I worry about what is going to happen 0,827

I feel nervous 0,809

I feel I worry a lot 0,832 I feel I worry about something bad will happen to me 0,806 health

involvement Steptoe, A., Pollard, T. M., & Wardle, J. (1995)

It is important to me that this smoothie keeps me

healthy 0,821 0,745

It is nutritious 0,822

It is good for my skin/teeth/hair/nailset 0,799

3.5 Manipulation re-check

Similar to the pretest reported in section 3.3, the two 1-item questions are included in the official survey in Qualtrics. For both 1 - item questions, the p - value of the test for equality of variances are larger than 0,05, the null hypothesis of the equal variances is accepted and thus, the first line in the output table was interpreted. The respective results are shown in table 3.5. The results are in line with the pretest. The results show that there both exist significant differences on the means of perceived ethical information and stress level between the compared two groups.

Perceived ethical information level (based on 1-item question)

ethical information N Mean S. D Sig (2 tailed), 95% confidence interval 0 (no) 129 3,40 1,53 0,000 1 (yes) 130 5,01 1,36 Stress level (based on 1-item question) Fear arousing information

N Mean S. D Sig (2 tailed), 95% confidence interval

0 (no) 148 3,17 1,62 0,000

1(yes) 111 4,91 1,56

Table 3.5. manipulation of perceived ethical information and stress level based on 1-item question

significantly different between the compared two groups when the second line in the output table was interpreted. Perceived ethical information level (based on created construct)

ethical information N Mean S. D Sig (2 tailed), 95% confidence interval 0 (no) 129 3,93 1,18 0,000 1 (yes) 130 4,73 1,17 Stress level (based on created construct) Stress arousing information

N Mean S. D Sig (2 tailed), 95% confidence interval

0 (no) 148 4,09 1,41 0,002

1 (yes) 111 4,60 1,18

Table 3.6. manipulation of perceived ethical information and stress level based on created constructs

3.6. Correlation matrix and regression test for control variable

Health Involvement Comments credibility Customer engagement of the advertisement Health Involvement 1 0,710** (0,000) 0,531** (0,000) Comments credibility 0,710** (0,000) 1 0,684** (0,000) Customer engagement of the advertisement 0,531** (0,000) 0,684** (0,000) 1 Table 3.7. Pearson Correlation **correlation is significant at the 0.01 level (2-tailed).

Customer engagement of the

advertisement Comments credibility

Health Involvement 0,551**

(0,000) 0,718** (0,000)

Table 3.8. Regression test for health involvement **significant at the 0.01 level

3.7 Hypothesis testing

The hypothesis of this thesis will be tested first separately and then being combined in one analysis based on Hayes PROCESS model 7 in SPSS. The aim of analyzing all relevant variables in one model is to check whether the result of the separate analysis will hold when all the variables are included in one model and the theory is tested as one. Besides, considering the fact that independent variables in the conceptual model consist of both categorical and metric variables, Ancova analysis is considered for testing H1, H2, H5 (Malhotra, 2010). Hypothesis 4 is analyzed by a linear regression. The mediation effect H3 is tested by 3 linear regressions based on Baron and Kenny (1986): First regress dependent variable on independent variable. Next regress mediator on independent variable and then regress dependent variable on mediator and independent variable.

Interaction term p-value H1 ethical*HI 0,000 H2 ethical*HI 0,000 H5 ethical*HI stress*HI ethical*stress*HI 0,003 0,108 0,375 Table 3.9. Test for homogeneity of regression slope

H1: CE= 𝛽! + 𝛽"ethical + 𝛽#HI + 𝜖 CE = customer engagement of the

advertisement

CC = comments credibility HI = health involvement

Ethical = ethical branded content (0 = No, 1 = Yes)

Stress = stress (0 = No, 1 = Yes) 𝜀= error term H2: CC= 𝛽! + 𝛽"ethical + 𝛽#HI + 𝜖 H3: CE= 𝛽! + 𝛽"ethical + 𝛽#HI + 𝜖 (1) CC= 𝛽! + 𝛽"ethical + 𝛽#HI + 𝜖 (2) CE= 𝛽!+ 𝛽"ethical + 𝛽#CC + 𝛽%HI + 𝜖 (3) H4: CE= 𝛽! + 𝛽"CC + 𝛽#HI + 𝜖

H5: CC= 𝛽! + 𝛽"ethical + 𝛽#stress + 𝛽%ethical*stress +

𝛽_&HI + 𝜖

4. Output analysis 4.1 VIF testing

Before starting the regression analysis, the existence of multicollinearity was checked. The aim of applying a multicollinearity test is to see whether there exists causal relationship among independent variables in each model. In table 4.1, the results show that the VIF values in each model are all lower than 3, which meet the criteria of 4 (Gijsenberg, 2020). Therefore, no multicollinearity was found.

4.2.1 Ethical branded content and customer engagement of the advertisement (H1)

Column (1) in table 4.1 shows the result of regression analysis for hypothesis 1. Model 1 represents the original impact of using ethical branded content on customer engagement of the advertisement. Under the situation that the health involvement is controlled when analyzing the effect of ethical branded content on customer engagement of the advertisement, using ethical content increases on average 0,642 scores on customer engagement of the advertisement than the content with no ethical information. This positive effect is significant, which is in line with H1. Besides, the control variable health involvement has a significant positive effect on customer engagement of the advertisement (Coeff = 0,490, p =0,000), suggesting the higher health involved a customer is, the higher engagement of the advertisement he/she would be.

4.2.2 Ethical branded content and comments credibility (H2)

Column (2) in table 4.1 shows the regression analysis for hypothesis 2. When controlling the effect of health involvement to test the impact of ethical information on comments credibility, the coefficients of both explanatory variable and covariate are all positive and significant. The result is in line with the hypothesis 2 that the use of ethical information in branded content would increase the credibility of the comments. What’s more, health involvement again plays an important role.

4.2.3 Comments credibility and customer engagement of the advertisement (H4)

this case, the control variable health involvement has no effect on customer engagement of the advertisement since the coefficient is insignificant (p = 0,133).

(1) DV: CE (2) DV:CC (3_1) DV: CE (3_2) DV:CC (3_3) DV: CE (4) DV: CE (5) DV:CC Ethical branded content 0,642 (0,000) 0,322 (0,006) 0,642 (0,000) 0,322 (0,006) 0,467 (0,000) 0,383 (0,015) (VIF) 1,064 1,063 1,064 1,063 1,095 0,1927 Comments credibility 0,588 (0,000) 0,630 (0,000) (VIF) 2,075 2,015 Stress 0,287 (0,084) (VIF) 2,121 Ethical branded content*Stress - 0,128 (0,579) (VIF) 2,902 Health involvement 0,490 (0,000) 0,688 (0,000) 0,490 (0,000) 0,688 (0,000) 0,087 (0,186) 0,101 (0,133) 0,670 (0,000) (VIF) 1,064 1,063 1,064 1,063 2,022 2,015 1,119 R square 0,337 0,518 0,337 0,518 0,502 0,473 0,526 Adjusted R square 0,331 0,514 0,331 0,514 0,496 0,469 0,518

Table 4.1 The regression output of hypothesis (unstandardized coefficient)

4.2.4 The mediating effect of comments credibility (H3)

The mediation effect was tested based on the 3 equations of hypothesis 3 from Baron and Kenny (1986) in table 3.10. The results are listed from column (3_1) to (3_3) in table 4.1.

Column (3_2) shows the result of the regression with comments credibility as the criterion variable and ethical branded content as a predictor, testing the effect of path a. The result shows a positive significant effect of ethical branded content on comments credibility when health involvement is controlled (coefficient =0,322, p = 0,006). Again, control variable health involvement shows a positive significant effect (coefficient =0,688, p = 0,000). Comments credibility increase with the level of health involvement.

Column (3_3) shows the result of the regression with customer engagement of the advertisement as the criterion variable and ethical branded content, comments credibility as predictors, testing the direct effect (path c’) and the effect of path b. The coefficients are significant and positive when health involvement is controlled. However, health involvement shows no effect because the p - value is insignificant (p=0,186).

The details of mediating effect are presented in figure 4.1. Since c≈c’+ab, the requirement of total effect = direct effect + indirect effect is met. Besides, due to the fact that both direct and indirect effect of ethical branded content on customer engagement of the advertisement is significantly larger than 0, partial mediation effect of comments credibility is found.

Figure 4.1 mediating effect analysis

4.2.5 The moderating effect of stress level on the relationship between ethical branded content and comments credibility (H5).

moderation effect does not exist when the health involvement is controlled. However, the control variable health involvement still shows a significant positive effect on comments credibility.

4.2.6 Moderated mediation

Finally, the relationships in the conceptual model are analyzed all together using Hayes Process Model 7 in SPSS based on a number of 5000 bootstrap samples and 95% confidence interval. The control variable health involvement is included in analysis because it was found to have a significant effect on comments credibility and customer engagement of the advertisement. The results are consistent with the results of previous separate analysis and are indicated in figure 4.2. Besides, though there is a significant effect of control variable health involvement on comments credibility (Coeff = 0,6704, p= 0,0000), no significant effect of health involvement on customer engagement of the advertisement is found (Coeff = 0,0871, p= 0,1862).

The direct effect of ethical branded content is found to have a positive effect on customer engagement of the advertisement (Coeff = 0,4671, p = 0,0002), confirming H1. Besides, H2 is confirmed because the coefficient of ethical branded content on comments credibility is significantly positive (Coeff = 0,3266, p = 0,0056). Meanwhile, comments credibility is found to have a positive effect on customer engagement of the advertisement (Coeff = 0,5883, p = 0,0000), thus, H4 is confirmed. With regard to the indirect effect, conditional indirect effect is found as 0 loads outside the lower and upper bound of the confidence interval (0,0267, 0,4358) when stress stays at 1 S.D. below its mean. Thus, H3 is confirmed as well.

Figure 4.2. Moderated mediation analysis based on Hayes Process Model 7

5. Conclusion and discussion 5.1 Conclusion and discussion

However, interestingly, opposite to H5, a negative sign was found on the interaction term in moderation analysis. Though the coefficient of the interaction is found negative, the p-value is insignificant. Thus, it can be only concluded that no moderation effect of stress on the relationship between ethical branded content and comments credibility is found (H5). Being exposed to the Corona news does not influence customers’ perceived comments credibility on ethical branded content. One potential reason might be because of China's successful control of COVID-19 after being the first hit in Jan 2020 (Burki. 2020). Chinese have high compliance with the government rule on mask wearing etc. As a result, the most hit area Wuhan released the lockdown on 7th April, 2020 (Zhong et al. 2020), let alone other parts with few cases. During the time the data was collected, the coronavirus has been well controlled and the society has been back to the normal from locked down for more than 8 months. Thus, being exposed to news about the coronavirus might not affect the respondents’ evaluation on comments credibility of ethical branded content. Another reason might be that consumers feel ethical products are useless to prevent the COVID-19 so they give up taking actions to avoid the threat. Witte et al, (2000) found when consumers are faced with strong fear and low-efficacy messages, they are most likely to ignore or even have defensive responses instead of believing the content and changing their behavior. In the experiment, it might be the case that people have been perceiving COVID-19 as a big threat and thus, normal health advice such as healthy eating is not perceived as an effective way to avoid the threat.

Hypothesis Results

H1: Branded content with ethical information has a larger positive effect on

consumers engagement of the advertisement than branded content with no ethical information.

Confirmed

H2: Branded content with ethical information has a larger positive effect on comments credibility than branded content with no ethical information.

Confirmed

H3: The relationship between ethical information and customer engagement is

partially mediated by comments credibility. Confirmed

H4: comments credibility has a positive effect on customer engagement of the advertisement.

Confirmed

H5: Stress level enhances the positive effect of ethical content on comments credibility

Rejected

Besides, as for the control variable, a positive and significant effect of health involvement is found on comments credibility. The finding is in line with the hypothesis. Whereas, surprisingly, no significant effect of health involvement on customer engagement of the advertisement is found. One potential reason might be that since the survey was handed out via personal network and the large portion of the respondents are Chinese, thus, their involvement of the advertisement in English is relatively low.

5.2. Management implication, limitation and suggestion for further research

This research provides more insights to the management strategy of native advertising. First, in order to increase customer engagement of branded content, managers are encouraged to use ethical information in branded content. Second, with regard to the comments credibility, it is important for managers to make positive comments visible as customer engagement of the advertisement increases with comments credibility. Third, managers are advised to classify the customers based on their health conscious level and expose ethical branded content with positive comments to those with a high health conscious level.

Additional to the management implications, there are existing several factors that could bias the results of the experiment: First, the survey was distributed online randomly depending on the personal network. Thus, there is not a specific target group in the distribution of the survey. For example, since a large portion of the respondents are students in various universities in China, their understanding of the English language is biased because English is not their native language. Furthermore, Goodrich et al. (2014) found that customer engagement of online advertisement differs in various cultures. Thus, the results of the experiment can be quite different if the same questions are answered by Dutch students. As a result, the experiment's external validity is influenced.

Another issue is that COVID-19 is well under control in Chinese society. They have been back to the normal life for more than eight months since the lockdown. Since in the experiment, the Corona news is used to induce respondents’ stress, this manipulation of stress might make people in China feel less stressful than people in the UK, where people are still struggling with the coronavirus and going into a second national lockdown (BBC.2021).

6. Reference

Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of personality and social psychology, 51(6), 1173.

BBC News. (2021). Covid: What are the lockdown rules across the UK? Retrieved from BBC: https://www.bbc.co.uk/news/explainers-52530518

Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal of marketing research, 49(2), 192-205.

Berman, J. Z., Levine, E. E., Barasch, A., & Small, D. A. (2015). The Braggart's dilemma: On the social rewards and penalties of advertising prosocial behavior. Journal of Marketing Research, 52(1), 90-104.

.

Bickart, B., Schindler, R.M., 2001. Internet forums as influential sources of consumer information. J. Interact. Mark. 15 (3), 31–40.

Brunk, K. H. (2012). Un/ethical company and brand perceptions: Conceptualising and operationalising consumer meanings. Journal of business ethics, 111(4), 551-565.

Burki, T. (2020). China's successful control of COVID-19. The Lancet Infectious Diseases, 20(11), 1240-1241. Chen, Y., & Xie, J. (2008). Online consumer review: Word-of-mouth as a new element of marketing communication mix. Management science, 54(3), 477-491.

Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annu. Rev. Psychol., 55, 591-621.

Fennis, B. M., & Stroebe, W. (2015). The psychology of advertising. Psychology Press.

Friestad, M., & Wright, P. (1994). The persuasion knowledge model: How people cope with persuasion attempts. Journal of consumer research, 21(1), 1-31.

Garbarino, E., & Johnson, M. S. (1999). The different roles of satisfaction, trust, and commitment in customer relationships. Journal of marketing, 63(2), 70-87.

Gijsenberg, M.J. (2020). Marketing Research Methods MRM01: Factor Analysis [PowerPoint slides]. University of Groningen

Gijsenberg, M.J. (2020). Marketing Research Methods MRM05: Ancova [PowerPoint slides]. University of Groningen

Goodrich, K., & De Mooij, M. (2014). How ‘social’ are social media? A cross-cultural comparison of online and offline purchase decision influences. Journal of marketing communications, 20(1-2), 103-116. Gremler, D. D., Gwinner, K. P., & Brown, S. W. (2001). Generating positive word-of-mouth communication through customer-employee relationships. International Journal of Service Industry Management. Griskevicius, V., Tybur, J., & Van den Berg, B. (2010). Going green to be seen: Status, reputation, and conspicuous conservation. Journal of Personality and Social Psychology, 98(3), 392-404

Groves, P. M., & Thompson, R. F. (1970). Habituation: a dual-process theory. Psychological review, 77(5), 419.

Harms, B., Bijmolt, T. H., & Hoekstra, J. C. (2017). Digital native advertising: practitioner perspectives and a research agenda. Journal of Interactive Advertising, 17(2), 80-91.

Kavaliauske, M., & Ubartaite, S. (2014). Ethical behaviour: Factors influencing intention to buy organic products in Lithuania. Economics and management, 19(1), 72-83.

Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer.

MacKenzie, S. B., & Lutz, R. J. (1989). An empirical examination of the structural antecedents of attitude toward the ad in an advertising pretesting context. Journal of marketing, 53(2), 48-65.

Maier, E., & Wilken, R. (2014). The impact of stress on consumers' willingness to pay. Psychology & Marketing, 31(9), 774-785.

Malhotra, N. K. (2010). Marketing Research: An Applied Orientation Sixth Edition Pearson Education. McKnight, D. H., Choudhury, V., & Kacmar, C. (2002). Developing and validating trust measures for e-commerce: An integrative typology. Information systems research, 13(3), 334-359.

Montaño, D. E., & Kasprzyk, D. (2015). Theory of reasoned action, theory of planned behavior, and the integrated behavioral model. Health behavior: Theory, research and practice, 70(4), 231.

Nelson, M. R., & Park, J. (2015). Publicity as covert marketing? The role of persuasion knowledge and ethical perceptions on beliefs and credibility in a video news release story. Journal of Business Ethics, 130(2), 327-341.

Nikolinakou, A., & King, K. W. (2018). Viral video ads: Emotional triggers and social media virality. Psychology & Marketing, 35(10), 715-726.

Nowak, K. L., & McGloin, R. (2014). The influence of peer reviews on source credibility and purchase intention. Societies, 4(4), 689-705.

Olsen, S. O. (2003). Understanding the relationship between age and seafood consumption: the mediating role of attitude, health involvement and convenience. Food quality and Preference, 14(3), 199-209. Perrin, N. (2019). US Native Advertising 2019, Display Budgets Keep Flowing to New Formats. Retrieved from emarketer: https://www.emarketer.com/content/us-native-advertising-2019.

Poister, T. H., & Thomas, J. C. (2011). The effect of expectations and expectancy confirmation/disconfirmation on motorists’ satisfaction with state highways. Journal of Public Administration Research and Theory, 21(4), 601-617.

Probst, F., Grosswiele, L., & Pfleger, R. (2013). Who will lead and who will follow: Identifying Influential Users in Online Social Networks. Business & Information Systems Engineering, 5(3), 179-193.

Reynolds, C. R., & Richmond, B. O. (1978). What I think and feel: A revised measure of children's manifest anxiety. Journal of abnormal child psychology, 6(2), 271-280.

.

Risselada, H., Verhoef, P. C., & Bijmolt, T. H. (2014). Dynamic effects of social influence and direct marketing on the adoption of high-technology products. Journal of Marketing, 78(2), 52-68.

Robinson, M. D. (1998). Running from William James' bear: A review of preattentive mechanisms and their contributions to emotional experience. Cognition & Emotion, 12(5), 667-696.

Russell, J. A. (1980). A circumplex model of affect. Journal of personality and social psychology, 39(6), 1161.

Sabate, F., Berbegal-Mirabent, J., Cañabate, A., & Lebherz, P. R. (2014). Factors influencing popularity of branded content in Facebook fan pages. European management journal, 32(6), 1001-1011.

Samson, A., & Voyer, B. G. (2014). Emergency purchasing situations: Implications for consumer decision-making. Journal of Economic Psychology, 44, 21-33.

Schwarz, N. (2011). Feelings-as-information theory. Handbook of theories of social psychology, 1, 289-308.

.

Steptoe, A., Pollard, T. M., & Wardle, J. (1995). Development of a measure of the motives.underlying the selection of food: the food choice questionnaire. Appetite, 25(3), 267-284.

Taylor, D. G., Strutton, D., & Thompson, K. (2012). Self-enhancement as a motivation for sharing online advertising. Journal of Interactive Advertising, 12(2), 13-28.

Trope, Y., & Liberman, N. (2010). Construal-level theory of psychological distance. Psychological review, 117(2), 440.

Tutaj, K., & Van Reijmersdal, E. A. (2012). Effects of online advertising format and persuasion knowledge on audience reactions. Journal of Marketing Communications, 18(1), 5-18.

Unit, Economist Intelligence. "Beyond loyalty: meeting the challenge of customer engagement, part 2." Economist (2007).

Van Ryzin, Gregg G. 2004. Expectations, performance, and citizen satisfaction with urban services. Journal of Policy Analysis and Management 23:433–48.

Wei, M. L., Fischer, E., & Main, K. J. (2008). An examination of the effects of activating persuasion knowledge on consumer response to brands engaging in covert marketing. Journal of Public Policy and Marketing, 27(1), 34–44.

Williams, P. R., & Hammitt, J. K. (2001). Perceived risks of conventional and organic produce: pesticides, pathogens, and natural toxins. Risk analysis, 21(2), 319-330.

Witte, Kim, and Mike Allen. (2000) A meta-analysis of fear appeals: Implications for effective public health campaigns. Health education & behavior 27.5: 591-615.

Zander, K., & Hamm, U. (2010). Consumer preferences for additional ethical attributes of organic food. Food quality and preference, 21(5), 495-503.

Zhong, Raymond. "China Ends Wuhan Lockdown, but Normal Life Is a Distant Dream". New York Times. Retrieved 7 April 2020.

4 ➔ Consumers believe ethical

products are healthier and have higher standards (williams et al.2001 & Zander et al. 2010) ➔ Status motive theory

(Griskrevicius et al. 2010)

➔ Consumers perceive the story telling more credible when it contains ethical information (Nelson et al. 2015 & Wei et al.

2008). ➔ A positive relation between

comments credibility and visual characteristics’ credibility(Nowak et al. 2014)

➔ Expectancy -confirmation theory (Poister et al. 2011, Van Ryzin et al.

2014, Lotz et al. 2013) ➔ Reasoned action theory (Montano et

al. 2015, Kavaliauske et al. 2014) ➔ Braggart’s dilemma (Berman et al.

2015) ➔ Social confirmation theory (Claldini

et al. 2004) ➔ Construal level theory (Maier et al.

2014)

➔ Dual process theory (Groves et al. 1970)

➔ Fear and anxiety mechanism (Robinson, 1998) ➔ The feeling as information theory

(Schwarz et al. 2011) ➔ Affect circumplex model (Russell,

1980)

➔ Online comments made by individuals are found to be more credible than retailer-made information (Bickart et

al. 2001) ➔ The more credible consumers feel

about the product, the more likely they give positive comments and stay with the company in long run (Garbarino et

7

8

13

14