D
OES INFLUENCER
MARKETING LEAD TO
PURCHASE
?
–
A
STUDY FROM THE CONSUMER
’
S
PERSPECTIVE
2020: VT2020CE12 Examensarbete – Civilekonom Företagsekonomi Sharon Arzaghi Madeleine HolmAcknowledgments
It has been very interesting to study this subject especially since influencer marketing is an expanding marketing strategy in today’s society.
We would like to thank our supervisor Magnus Roos for guiding us through the entire process and always being available when needed. We would not have been able to do this without his commitment and encouragement. Furthermore, we would also like to thank our friends and family for their support throughout this entire process.
Borås 2020-06-07
_________________________ _________________________ Sharon Arzaghi Madeleine Holm
Svensk titel: Leder influencer marketing till köp? En studie från konsumentens perspektiv Engelsk titel: Does influencer marketing lead to purchase? A study from the consumer’s
perspective
Utgivningsår: 2020
Författare: Sharon Arzaghi och Madeleine Holm Handledare: Magnus Roos
Abstract
Since the increased popularity of social media, new marketing strategies have been developed by companies and marketers in order to advertise more effectively. This has concluded in a new marketing strategy originated from celebrity endorsement, called influencer marketing, which is implied by various brands today and often used on the social media platform
Instagram. Influencer marketing is a strategy where a brand invests in an individual who has a big influence over its followers, with the purpose to recommend a brand or a product. The issue however is that there are thousands of influencers, and companies have to be selective when choosing who to collaborate with. Furthermore, there are studies examining factors which have an effect on attitudes and purchase intention, but very few study purchase behaviour.
The purpose of this paper is to study which factors of an influencer have an effect on consumers’ purchase behaviour. This will be done by examining commerciality,
entertainment and factors included in source credibility towards purchase. There was also an interest to investigate if certain personality types of consumers were more prone to make a purchase. Because of this, extraversion and self-monitoring were also included in relation to purchase behaviour. Furthermore, two types of exposure will also be examined to see their effect on purchase.
For this study a quantitative approach was applied with the use of a survey questionnaire which resulted in 524 valid answers. The results showed that commerciality, attractivity and similarity had a significant effect on purchase, whereas trustworthiness, knowledge and entertainment did not. Furthermore, personality traits are also a determinant factor as high levels of extraversion and self-monitoring show to have an impact on purchase. Lastly, purchase seems to increase if an individual spends more time on Instagram or follows more influencers, meaning exposure also had a significant effect.
In conclusion, our results show clear differences in the dominating factors which have been previously researched towards purchase intention. Attractivity is the factor which has most influence over purchase, closely followed by commerciality.
Sammanfattning
Uppkomsten av sociala medier har förändrat människors sätt att söka och samla information. Detta har påverkat företags val av strategi för att nå ut till sina kunder på bästa möjliga sätt. Den strategi som kommit att bli dominerande bland företag på den sociala plattformen Instagram är influencer marketing. Denna strategi innebär att företag investerar i en individ som har stort inflytande över sina följare för att marknadsföra en viss produkt eller tjänst. Detta görs i syfte att påverka individens medvetenhet och uppfattning om företag men även dess köpintention för att potentiellt utföra ett köp. Influencer marketing innebär dock att en vanlig person utan att behöva vara känd kan ha ett stort inflytande över sina följare vilket resulterat i tusentals olika influensers. På grund av detta måste företag vara selektiva vid valet av influenser för att marknadsföringsstrategin ska vara effektiv. Tidigare studier har undersökt vilka faktorer som har en inverkan på individens attityd och köpintention men endast ett fåtal undersöker köpbeteende.
Syftet med denna studie är att undersöka vilka faktorer av en influenser som påverkar konsumenters köpbeteende. Detta kommer att göras genom att studera faktorerna kommersialitet, underhållning och de faktorer som ingår i modellen source credibility
gentemot köp. Studien syftar även till att undersöka om individer med en viss personlighet är mer benägna att utföra ett köp, därav inkluderades extraversion och self-monitoring i relation till köpbeteende. Vidare undersöks effekten av två typer av exponering i förhållande till köp.
För studien användes en kvantitativ insamlingsmetod som bestod av en enkät där 524 giltiga svar samlades in. Resultaten visade att faktorerna kommersialitet, attraktivitet och likhet hade en signifikant effekt på köp. Däremot visade faktorerna trovärdighet, kunskap och
underhållning ingen signifikant effekt. Från resultaten framgick det även att personlighetstyp är en avgörande faktor där hög grad av extraversion och self-monitoring har inverkan på köp. Dessutom uppvisar resultaten att köp ökar om en individ spenderar mer tid på Instagram eller om hen följer fler influensers, vilket innebär att exponering har en betydande effekt.
Slutligen visar våra resultat tydliga skillnader avseende de dominerande faktorer som tidigare undersökts i relation till köp. Faktorn attraktivitet har störst inflytande över köp följt av kommersialitet.
Nyckelord: Social media, influencer marketing, Instagram, Source credibility, köpbeteende,
Table of content
1 Introduction ... - 1 -
1.1 Background ... - 1 -
1.2 Problem formulation ... - 1 -
1.3 Purpose ... - 2 -
1.4 Research questions and hypotheses ... - 2 -
1.5 Delimitations ... - 3 -
1.6 Relevance of the subject ... - 3 -
1.7 Thesis outline ... - 3 - 2 Previous research ... - 4 - 2.1 Social media ... - 4 - 2.2 Endorsement ... - 4 - 2.3 Influencer marketing ... - 4 - 3 Theoretical framework ... - 6 - 3.1 Source credibility ... - 6 -
3.2 Advertising entertainment value ... - 6 -
3.3 Extraversion ... - 7 -
3.4 Self-monitoring ... - 8 -
3.5 Exposure and context ... - 8 -
4 Method ... - 9 -
4.1 Research approach ... - 9 -
4.2 Research method ... - 9 -
4.3 Research design ... - 10 -
4.4 Data collection ... - 10 -
4.4.1 Primary & secondary data ... - 10 -
4.4.2 Sampling ... - 11 -
4.4.3 Survey ... - 11 -
4.4.4 Survey design ... - 11 -
4.4.5 Pilot study ... - 12 -
4.4.6 Dependent and independent variable ... - 12 -
4.5 Analysis method ... - 13 - 4.6 Quality standards ... - 13 - 4.6.1 Reliability ... - 13 - 4.6.2 Validity ... - 14 - 4.7 Source criticism ... - 15 - 4.8 Ethics ... - 16 - 4.9 Method criticism ... - 16 -
5.1 Descriptive analysis ... - 18 -
5.2 Correlation analysis ... - 19 -
5.3 Regression analysis ... - 20 -
6 Discussion and conclusions ... - 22 -
6.1 Introduction ... - 22 -
6.1.1 Which factors of an influencer has an effect on a consumer’s purchase behaviour? .. - 22 -
6.1.2 Does a consumer’s personality type have an effect on purchase behaviour? ... - 23 -
6.1.3 Does exposure have an effect on consumer’s purchase behaviour? ... - 23 -
6.2 Implications and conclusions ... - 24 -
6.3 Future research ... - 24 -
References ... - 25 -
1 Introduction
1.1 Background
Marketing has come a long way from what it used to be twenty years ago. Traditional marketing consisting of TV, radio and newspaper are no longer the dominant channels used for advertising. The emerging of internet has created lots of new opportunities for companies to market themselves and differentiate from the traditional marketing methods. It has also generated new ways for consumers to gain information from various online sources (Lou & Yuan 2019). Statista (2020a) estimated the total of active internet users to be 4.54 billions of people all around the world, whereas 3.8 billions of these were active social media users. This creates a big platform online, enabling companies to reach out to a wider audience through new methods. As a result of various opportunities created by the internet, it has become more important for marketers to carefully design the message they want to send and also the choice of channel where the message shall be sent. Considering that consumers are exposed to a huge amount of advertising on a daily basis, they have also become skilled at sorting out messages that does not catch their attention (Hwang 2010). The focus of advertisement has shifted from giving specific information about a product or service, to managing new ways of delivering it and trying to control how it is perceived by the consumer (Brown & Hayes 2008, p. 11). Lou and Yuan (2019) explain that it has become common to use various influential
individuals as spokespersons to communicate a certain advertising message. These individuals are often celebrities who are compensated to advertise and talk about a certain brand, product or service. This is called celebrity endorsement and is often used by companies to add
recognition to their brand (Jin & Phua 2014). Nowadays however, endorsement is used with another type of influential individuals which are called social media influencers. These individuals have become an online personality by creating and posting user-generated content on social media (Freberg, Graham, Mcgaughey & Freberg 2011; Lou & Yuan 2019).
A social media influencer is an opinion leader who can influence its followers’ opinions and purchasing decisions when sharing a certain review or opinion about a product or service (Dhanesh & Duthler 2019). The use of these individuals in marketing is called influencer marketing which in recent years have come to be used by many marketers to transmit advertising messages to its target audience. The exponential growth of influencer marketing can be noticed by the increased budget allocated for this purpose. Statista (2018) estimated a global spending of 5.67 billion dollars in 2018 and estimates that this number will increase to 8.08 billion dollars at the end of year 2020. Since influencers have the capacity to reach out to a large audience, companies can take advantage of the benefits this entails such as exposure, engagement and brand awareness (Huang & Sarigöllü 2011). If high brand awareness is achieved, the confidence toward the brand will increase and therefore also the trust (Lu, Chang & Chang 2014).
1.2 Problem formulation
In order for influencer marketing to be effective the brand and the influencer who cooperates should project the same values. Social media influencers usually have some knowledge or expertise within certain subjects which has been developed by their interests in these areas. Depending on which field the influencer specialises in, it will create and post content on its
service which does not correlate with its image and values, it may result in negative effects instead. The consumer may perceive the source of information and the brand as unreliable which further affects the consumers’ attitudes and purchase intentions (Djafarova & Rushworth 2016; Lou & Yuan 2019).
Social media influencers publish both non-sponsored and sponsored posts where the influencer receives compensation such as free products, free services or monetary compensation (Stubb & Colliander 2019). Depending on the number of followers and engagement the influencer has on its social platform the brand determines the form of compensation (Dhanesh & Duthler 2019). However, a common issue with sponsored posts that involves any kind of compensation is that the consumer may perceive the communicated message as less credible (De Veirman & Hudders 2019). According to the source credibility model other factors that may influence the credibility of a source is; knowledge,
trustworthiness and attractivity (Ohanian 1990).
Moreover, there are different reasons why people follow influencers. This could simply be because of the entertainment they provide through their channel or that they feel a sense of connection to the influencer because of similarity (Sokolova & Kefi 2020). These factors may be the reason why people decide to follow a certain influencer, but it is difficult to draw a connection to purchase behaviour as influencer marketing is a new phenomenon and not as established as celebrity endorsement. Many companies may perceive a large amount of following as vital when choosing which influencers to collaborate with, but studies show that other factors are just as important (De Veirman, Cauberghe & Hudders 2017).By only observing the influencer’s following, marketers and companies may spend their resources in an ineffective manner. This principle is of most importance for smaller companies who do not have the economic resources to choose the biggest influencers to collaborate with. Moreover, prior studies have examined factors in relation to purchase intention, leaving a gap in
literature for purchase behaviour.
1.3 Purpose
The purpose of this study is to examine which factors of an influencer have an effect on purchase behaviour, as well as if certain personality types are more prone to make a purchase. These factors include; trustworthiness, knowledge, attractivity, commerciality, similarity and
entertainment which will be studied in relation to influencer marketing. Furthermore,
exposure and personality traits such as self-monitoring and extraversion will also be examined
to see if they have an impact.
1.4 Research questions and hypotheses
R1: Which factors of an influencer has an effect on a consumer’s purchase behaviour?
H1. A high level of trustworthiness (H1a), knowledge (H1b), attractivity (H1c),
commerciality (H1d), similarity (H1e) and entertainment (H1f) will lead to purchase
R2: Does a consumer’s personality type have an effect on purchase behaviour?
H2. A high level of extraversion will lead to purchase
H3. A high level of self-monitoring will lead to purchase
H4: An individual who follows more influencers on Instagram is more likely to make a
purchase
H5: An individual who spends more hours on Instagram is more likely to make a purchase
1.5 Delimitations
As this study had to be conducted under a certain time frame with limited resources, we decided to focus on six specific factors which are well known and established within science of marketing. Since the purpose also touches on the issue of personality traits, we decided to limit it to extraversion and self-monitoring. The reason for this is because prior studies have shown that these factors have an impact on purchase behaviour. The target group was limited to only include woman between the ages of 16 to 35, who are regular social media users. Furthermore, a criterion for the target group was that these women also had to be frequent Instagram users, as the study focuses on influencers on this platform. In order to be as precise as possible, an influencer was defined as a third-party endorser who has become famous through their social media and has over 10.000 followers (Boerman 2020; Dhanesh & Duthler 2019).
1.6 Relevance of the subject
Influencer marketing is a new marketing strategy which has emerged over the recent years. This originates from the same principle as celebrity endorsement, which means there are similar aspects between these two strategies. Influencer marketing however is more precise in terms of the channels used for exposure. Since the majority of influencers are ordinary people who created their platform, information regarding their values may be lacking. Companies will therefore be enforced to have a great amount of trust for these influencers when engaging a collaboration. Furthermore, since there is a huge number of influencers, companies also have to be very selective when choosing who to collaborate with. By knowing which factors of influencers the consumers mostly value in terms of purchase behaviour, companies can be more effective when choosing who to collaborate in order to allocate their resources the best way possible.
1.7 Thesis outline
The study is disposed according to the following structure: chapter two concerns previous research which is aimed to provide a narrower focus for the subject of the study, as well as give a prior understanding. The third chapter explains the theoretical framework were relevant literature, theories and concepts are explained. This is also the base of the study which guided the data collection. Chapter four concerns the methodology, where the chosen research approach, research design and data collection is explained. The fifth chapter presents the results of the empirical data collection. Lastly chapter six includes the discussion and conclusions of the study, as well as suggestions of future research.
2 Previous research
2.1 Social media
Social media is a combination of different platforms which are mobile and web-based technologies (Koetsier 2020). On these platforms communities and individuals interact by creating, sharing, discussing and modifying content which is generated from the user, usually called user-generated content (Kietzmann, Hermkens, McCarthy & Silvestre 2011). Social media is also used by various actors such as social networkers, business firms and
governmental organizations as a communication tool (Kim & Ko 2012). The possibility to interact through social media platforms has also caused an increase in the usage of social media amongst celebrities and social media influencers to communicate and engage in conversations with theirfollowers(Chung & Cho 2017). The most common platforms for influencing are Instagram, YouTube, Snapchat and Facebook (Dhanesh & Duthler 2019; Torres, Augusto & Matos 2019).
2.2 Endorsement
Endorsement is a marketing tool which is used by marketers and firms to impact the effectiveness of advertising in terms of purchase intention, brand recognition and purchase behaviour (Spry, Pappu & Cornwell 2011). Celebrities and social media influencers are examples of endorsers that are often used in advertising (Hu, Zhang & Wang 2019). A celebrity is an individual who has become famous through music, movies or sports without having a history in advertising activities(Jin, Muqaddam & Ryu 2019). A social media influencer however is a third-party endorser who has become famous through their social media. Endorsement by social media influencers is considered to be a new type of
endorsement compared to traditional celebrities (Freberg, Graham, McGaughey & Freberg 2010).
According to Boerman (2020) social media influencers can be divided into three levels which are; micro, meso and macro. Micro influencers have up to 10,000 followers, while meso influencers have a following of 10,000 to a million on their social media platform. The last level, macro, concerns influencers with over one million followers and are usually established celebrities which are known internationally such as, Kylie Jenner (Boerman 2020). However, the most common endorsement used nowadays falls into the category of social media
influencers and not traditional celebrities. This since consumers tend to feel a stronger connection and perceive these as more trustworthy (Jin, Muqaddam & Ryu 2019).
2.3 Influencer marketing
Influencer marketing is a marketing strategy where a brand invests in an individual who has an influence over its followers to promote and recommend a brand or product (Brown & Hayes 2008, p. 11; Lou & Yuan 2019; Stubb & Colliander 2019). Social media influencers are opinion leaders who have a large number of followers on their social media platform. These influencers are trusted by their followers, and therefore their recommendations are of importance when evaluating productsto make purchase decisions (Boerman 2020; Lou & Yuan 2019; Stubb & Colliander 2019). The influencer usually creates and/or promotes the branded content through its own personal social media channel to its audience and reviewsthe brandby giving information about the product and often its own personal experience
regarding it (Lou & Yuan 2019; Stubb & Colliander 2019). The large network a social media influencer has enables the brand that invests in the individual to reach a wide audience which encourages engagement as likes and comments (Boerman 2020). Brands and marketers use
influencer marketing to stimulate brand awareness and purchase intentions which is easier to do through an influencer’s brand generated content since it seems to be more authentic and organic (Lou & Yuan 2019).
The content of a social media influencer consists of both sponsored and non-sponsored posts which makes it difficult for consumers to recognise it as advertising (Boerman 2020; Stubb & Colliander 2019). A sponsored post implies that the influencer receives some type of
compensation for recommending a certain product or service. A non-sponsored post however is the influencer’s original content which has not been sponsored in any way to be commercial (Stubb & Colliander 2019). To make a differentiation between the two types of posts the social media platform Instagram introduced a standardized built-in disclosure which the influencer has to add to the sponsored post (Instagram 2020). When posting content that is not sponsored, the influencer also has an option to disclose this to emphasise impartiality. To disclose a collaboration the influencer can use hashtags as #paidad and #sponsored to mark the post and increase recognition for influencer marketing (Boerman 2020; Stubb &
Colliander 2019). However, Hwang and Jeong (2016) explain that when a message is recognised as advertising it has a negative effect on the consumer’s attitude. When a consumer recognises a persuasive message, they may want to resist the information that is provided (Stubb & Colliander 2019). The perceived credibility of the social media influencer decreases and therefore the sponsored posts may receive less attention and engagement (De Veirman & Hudders 2019).On the other hand, consumers consistently reevaluate information based on prior experiences and knowledge (Stubb & Colliander 2019). This could entail that a sponsored post may also have positive effects if the consumer’s attitude towards the
influencer is positive and s(he) believes in the content of the recommendation. If this is achieved the consumer may have the intention to purchase a certain product (Lu, Chang & Chang 2014). Furthermore, posts that are not sponsored may be perceived as simple reviews, which in itself may not create a purchase intention, whereas sponsored posts are made with the intention to generate a purchase or consumption (Spry, Pappu & Cornwell 2011).
3 Theoretical framework
This section aims to present the literature and theories which have been used in this study. Theories and models are explained, which make the foundation for the research questions and hypothesis, as well as how the data was collected and analysed.
3.1 Source credibility
Source credibility is a model which refers toa source of information who communicates a message, as well as if the receiver perceives the source to be unbiased, credible and
trustworthy. The factors included in source credibility also determines if the receiver accepts the message. Ohanian (1990) has developed the model from former theories by Hovland & Weiss (1953) and McGuire (1958). The model is composed of three dimensions which determines the credibility of a source; expertise, trustworthiness and attractiveness (Ohanian 1990). Expertise concerns how the receiver perceives the communicator of a source to make valid assertions. Knowledgeable is one of the factors included in expertise. Moreover,
trustworthiness implies how confident and reliable the source is when communicating these
assertions. Attractiveness concerns the effectiveness of the communicated message, which depends on the respondent’s similarity, familiarity and likeability towards the source. Ohanian (1990) explains that similarity and likeability are used as factors of identification and
interpersonal attraction between the source and the receiver of the message. Lastly familiarity concerns how familiar the source is to the receiver (Ohanian 1990). These factors are essential when communicating different messages which is why they have a significant effect on how things are perceived. Furthermore, it is explained by Ohanian (1990) that these factors should consciously be taken into consideration in terms of marketing as they have an impact on attitudes and consumers’ purchase intentions. This theory is of importance when evaluating an influencer since they regularly communicate messages to their audience.
Model 1: The Ohanian Model of Source Credibility
3.2 Advertising entertainment value
According to the advertising value model developed by Ducoffe (1995) advertising can be entertaining, informative or irritating. The different factors affect the advertising value and further the consumer’s attitude. If a consumer determines the advertising as valuable, the attitude is most likely to be positive. The value of entertainment is determined by how
entertaining, enjoyable and pleasing the advertisement is. Ducoffe (1995) also argues that consumers value entertainment because of its capacity to fulfil needs such as diversion and enjoyment. Moreover, informativeness and entertainment are the strongest factors which affect the consumers’ beliefs towards advertising. It has also been shown that these
dimensions are connected. Advertising which is found as entertaining by the consumer is also more likely to be evaluated as informative. Ducoffe (1995) explains that a possible reason for this is that if a consumer primarily does not find the advertising as entertaining, it may be ignored. Entertainment is therefore an important factor to consider when creating and communicating advertisement.
Model 2: Attitudes towards advertising model (Ducoffe 1995)
3.3 Extraversion
Eysenck’s (1946) introversion-extraversion personality theory involves two types of dimensions of an individual’s personality; extraversion and introversion. It is however explained that this is a model with two extreme counterparts, and that most people fall somewhere in between these. There are however characteristics connected to each dimension in the model. Eysenck explains that people who are introverts tend to do tasks slowly with more precision, executing it more correctly whereas those who are more extraverted do not have the same type of patience and tend to do tasks quicker which might come out inaccurate. Furthermore, it is also explained that introverts often underrate themselves and their
performances in all types of situations, while simultaneously having high aspirations. On the contrary, extroverts tend to have lower aspirations, and often overrate themselves. Extraverted individuals also show greater reactions to sensory stimulation than introverts, meaning they are more influenced by external factors (Eysenck 1946).
A similar definition is given from the five factor-model commonly known as the Big Five which includes the factors; extraversion, conscientiousness, openness, agreeableness and
neuroticism. This model was created and developed by independent researchers over a period
of many years (Digman 1990). Extraversion is connected to a number of traits which includes feeling positive emotions by interacting with others. The model also describes extraverted individuals as active, sociable and talkative (Costa & McCrae 1992). Theories regarding personality traits are of importance for businesses and brands when evaluating consumers as this affects their behaviour and actions.
3.4 Self-monitoring
Another theory which can be connected to consumers’ personality and behaviour is self-monitoring. The self‐monitoring theory explains social behaviour based on motivational, behavioural and situational factors (Gangestad & Snyder 1982). This theory is divided into high self-monitoring individuals and low self-monitoring individuals. The classification is done through the use of a self-monitoring scale which is based on scores classifying individuals into categories. High self-monitoring individuals are identified by their high scores on the scale and implies that an individual's behaviour is more or lesssituationally guided. These individuals also tend to enter and spend time in social situations. Low self-monitoring individuals are identified by their low scores on the scale which implies that the individual does not respond to situational factors at the same extent as high self-monitors. The behaviour of these individuals is rather guided from inner sources. This indicates that those who have a high level of self-monitoring are more likely to adapt to their surroundings and follow various norms and trends (Gangestad & Snyder 1982).
3.5 Exposure and context
In terms of advertising, contextual factors and exposure are always considered since they are traced back to human nature. Various theories explain these two, including the repetition-variation hypothesis. This theory illustratestwo strategies of advertising variation that can be applied to influence the consumer, which are; cosmetic variation and substantive variation. Cosmetic variation implies that the message of an ad is the same while the features of the ad as layout, background, colour and graphics might differ. On the contrary substantive variation concerns changes in the content of the message, whereas the visual content remains the same or at least similar. Contextual factors are necessary to study since the point in time when the consumer forms an opinion during an advertisement is crucial to its outcome (Schumann, Petty & Clemons 1990). On the other hand, Zajonc (1968) developed the mere-exposure effect which explains how people’s attitudes towards something can change and a liking can be developed solely because of familiarity. This is connected to exposure and illustrates that attitudes can be changed depending on how often an individual is exposed to something. In his study he discovered that when an individual firstly is exposed to a novel stimulus, the first emotion is fear. As this exposure continues a sense of familiarity is developed which instead leads to a liking.
4 Method
In this chapter we will describe the different parts of our method and how we decided to conduct the study. The chapter ends with a reflection of the chosen methodology.
4.1 Research approach
A study can adopt either a qualitative or quantitative approach depending on the purpose of the study and the information that needs to be obtained. A qualitative approach is used when the researcher focuses on words and wants to understand the perspective of a particular individual, object or phenomenon (Jacobsen 2002, p.38; Rienecker & Stray Jørgensen 2017, p. 174). A quantitative approach however is used when the researcher wants to quantify the collected data in order to find relations between variables and further analyse and present the results statistically (Bryman & Bell 2017, pp. 395-397; Rienecker & Stray Jørgensen 2017, p. 173). A problem when conducting a quantitative research may be that the researcher is more limited since the used method e, g., a survey cannot be modified once it has been distributed. There are also limitations within the surveys as the questions often are brief and do not provide an explanation of the participants’ answers (Bryman & Bell 2017, pp. 624-629).
However, a quantitative approach is more suitable when the research question is aimed towards relations between different variables, and the goal is to generalise the results across a larger audience. This is possible because a quantitative approach enables the researcher to include a larger selection of people to participate in the study (Jacobsen 2002, pp. 146-147). Since the purpose of this study was to examine different factors and if these could lead to a purchase, we had to collect data that was quantifiable. By doing this we were able to process and analyse the data through SPSS to observe different correlations and draw conclusions about which factors that played a role in the consumer’s decision to make a purchase. Due to our limited time and the chosen factors being obtained from existing theories based in qualitative studies, we had prior knowledge of the reasons why these affected people’s purchase decisions. This meant that there was no need to conduct another qualitative study to further understand the reasons behind these, making a quantitative approach the best possible choice for our purpose.
4.2 Research method
Qualitative and quantitative approaches are closely linked to two methods of reasoning which are deduction and induction. Bryman and Bell (2017, pp. 42-46) explain that a deductive process starts in existing theory and narrows it down into more specific hypotheses. These hypotheses set the foundation for how the data is collected. The results of the research will show if the hypotheses should be accepted or rejected, setting the ground for the last step in the process which is to reformulate the theories. On the contrary an inductive process usually has its starting point from e.g., observations which further influences the choice of theory used in the study. Both methods of reasoning have its limitations; a deductive process is limited to adhere to the chosen theory, whereas an inductive process is less standardised making it far more time consuming and no guarantee for a certain outcome. Furthermore, the researcher will only be able to draw conclusions from what has been observed (Alvesson & Sköldberg 2008, pp. 54-55; Bryman & Bell 2017, pp. 42-46).
For this particular study we emanated from already existing theories to search for gaps within the subject. Factors that lead to purchase intention has been studied by various researchers,
meaning a deductive approach was applied. The advantage of this approach is that we based the study on established theories, which influences the data collection and how the
information is interpreted. This increases objectivity in the study and decreases the chance of the researcher’s own opinions reflecting in the study. However, when using a deductive approach, you are naturally assuming that these prior theories are correct and still accurate, leaving little to no room for new findings (Davidson & Patel 2003, p. 23).
In order to ensure that these theories still were accurate, a pilot study was conducted where a small group of women were asked about this subject. This gave us an insight on how well the theory is applicable in today's society. Furthermore, this also helped us to consider the
different options available and select the correct theories for our research.
4.3 Research design
The different types of research design which can be used in a study are; cross-sectional, longitudinal, case study, comparative study and experimental design. The diverse types of research design determine how the researcher has to approach the collection and analysis of the data (Bryman & Bell 2017, pp. 72-86). In this study we collected the data through the use of a survey which implicates a cross-sectional study. The survey was equally distributed in a few groups, where it was available for two weeks. Reason for this was to gain as many responses as possible, enabling people to answer whenever they had time within this time span. Cross-sectional design is used then the researcher is interested in variations within a group. In order to use this design successfully, it is necessary to have a larger sample. This will increase the chance of finding variations between the variables and therefore make more accurate distinctions between the cases (Bryman & Bell 2017, p. 81). Furthermore, Bryman and Bell (2017, pp. 199-203) also explain the mean of the target group should be the same as the population’s.
The disadvantage with a cross-sectional method is that it only enables a study at a certain point in time, making it difficult to sometimes discover accurate causations (Bryman & Bell 2017, p. 81). However, in this particular study a research over time is not needed since progress in society needs a certain amount of time to become visible. Some of these theories have also been studied for many years and are primarily based on biological factors, making them accurate until something dramatically changes in the human behaviour.
4.4 Data collection
4.4.1 Primary & secondary data
Data can be gathered through primary and secondary sources, where a primary source implies information that has been collected for the first time and for a specific purpose. Data material which is collected through interviews, surveys or experiments are most often primary sources (Jacobsen 2002, pp. 152-153). A secondary source implies the use of already existing
information from prior research which has been processed. This could be scientific articles and books (Booth, Colomb & Williams 2004, p. 82; Eriksson & Wiedersheim 2011, p. 88).
The current study is based on both primary and secondary data where the collection of primary data has been done with the use of an online survey as well as different theories. Those theories who do not qualify as a primary source have been collected through books where they have been retold. We are aware that most secondary sources have collected data for another purpose than ours. This has resulted in interpretations of the secondary data to be able to connect it to this study and purpose.
4.4.2 Sampling
When conducting a study, a commonproblem that many researchers face is when they want to investigate their target group. Most often examining the entire population requires
extensive financial resources and time which is not an option for the majority of researchers. Because of this, it is common to use various sampling methods to create a sample which should be representative of the population. This makes it possible to generalise the results from the sample to the population (Bryman & Bell 2017, p. 180). There are two types of sampling methods; probability sampling and probability sampling. When using a non-probability sampling method some individuals in the population have a greater chance of being included in the sample. When using a probability sampling method every individual in the population has an equal chance of being included in the sample (Bryman & Bell 2017, p. 192).
We chose to study women between the ages of 16-35, since studies have shown that Swedish women use social media more than men (Svenskarna och internet 2018) and that women are more prone to shop online (Statista 2020b). The reason for this age interval was to include millennials without reaching the upper limit for this age group (McCormick 2016). On the other hand, the age 16 is the lowest age of individuals who can participate without parents’ permission (Vetenskapsrådet 2017). For this study we adopted a non-probability sampling method in form of a self-selection sample. This means that anyone who had access to the survey was able to answer. The main reason for this was because we did not have access to a register of our target group. Neither did we have the possibility of creating one due to
insufficient resources and lack of time. The next best possible option was to identify where our target group best could be found and conduct our study there. Since our target group is females aged between 16-35 who are active social media users, we decided to conduct the study on a social media platform. The reason for using Facebook is because there are several Swedish groups for women with over at least 50 000 members, which increases the
probability of acquiring answers from individuals within our target group. We chose the five biggest girl groups in Sweden which are; Honey and the Bees, Girltalk, Girlsquad, Pink Room and The Royal Ladies. By using these groups we ensured that the participants used social media, however we also had a disclaimer that one should only answer the survey if they had Instagram and followed at least one influencer. Since using a non-probability sample, the ability to make generalisations was restricted (Bryman & Bell 2017, pp. 180, 192-199).
4.4.3 Survey
Since this study aims to find relations between variables, we decided to conduct a quantitative research through a survey. When deciding which platform to use for the survey we compared the different options available and what they had to offer. Our first option was SurveyMonkey which overall had all the functions we needed and was aesthetically pleasing which could increase our response rate. However, in order to include more than ten questions in a survey a subscription was needed which included a cost. Because of this we decided to use Google Forms which had similar functions but with a simpler layout. There was also prior knowledge about Google Forms which facilitated the use of the program.
4.4.4 Survey design
Bryman & Bell (2017, pp. 330-333) explain that there are different scales and variables that can be used when creating a survey. The most optimal options are ratio- and interval scale. Interval scales have the same distance between the different alternatives. What separates an
makes it possible to calculate ratios. This makes the ratio scale the strongest level of measurement.
The survey used in this study consisted of 19 questions with both open-ended and close-ended questions, which were all obligatory. Initially we had a brief description of how we defined the term influencer in order to make the respondent answer the questions based on this definition. Since the study is aimed to examine females aged between 16-35, the two first questions in the survey were control questions. Both of these were close-ended questions where the respondents were asked to specify their gender and age. Furthermore, the
respondents were also asked to indicate their favourite influencer which was the only open-ended question in the entire survey. Remaining questions were all close-open-ended and related to the theory. Because of this, these questions had to be quantified and a numeric scale in form of an interval was created. This included questions where the respondents were asked to mark their opinion on both five-point- and seven-point scales. To see the used scales, see appendix 1.
4.4.5 Pilot study
Bryman and Bell (2017, p. 266) explain that pilot studies are done to make sure that the questions are correctly phrased and answered as intended. It is overall a secure way to minimise problems that can occur with surveys. When conducting a qualitative study there is always a researcher available to answer the respondents’ questions if something is unclear. This option is generally not available with a quantitative study, and it is therefore essential that the researcher minimises all potential problems before sending out the survey.
Because of this, a pilot study was conducted before the official survey. The individuals included for the pilot study were friends and acquaintances which fulfilled the criteria of our target group. These individuals were not included in the official survey in order to favour the representativeness of the study. When the pilot study was finished, we acquired answered surveys and personal notes from the respondents that helped us rephrase the survey in order to minimise misunderstandings.
4.4.6 Dependent and independent variable
A quantitative study is usually conducted when the researcher wants to quantify the data and explore different relations between variables. This includes variables that are dependent and independent. An independent variable is a variable which affects the dependent variable. Depending on the study, there can be multiple independent- or dependent variables. It is therefore important that the survey questions are correctly formulated in order to be able to measure the relations (Bryman & Bell 2017, pp. 69-70).
This study aims to gain an understanding of the variables that influence the choice to make a purchase from an influencer, which makes it necessary to distinguish between dependent and independent variables. Since we want to explore how the factors trustworthiness, knowledge,
attractivity, commerciality, similarity and entertainment have an effect on purchase, these
factors are our independent variables while purchase is our dependent variable. We also want to investigate the effect of the variables extraversion, self-monitoring and exposure on
purchase; therefore these variables are also independent. By testing the independent variables one by one, it is possible to investigate how the dependent variable is affected.
4.5 Analysis method
After collecting and saving the answers from Google forms, these were downloaded and exported to Excel. In Excel the invalid answers were sorted out. Thereafter the variables were coded into numbers which enabled us to transform them into scale values and prepared for SPSS. Since the respondents’ level of extraversion was measured through two questions, extraversion and introversion, one of these had to be reversed in order for us to compare them with each other. This was done by changing and reversing the introversion scale, meaning that a value of 5 was modified into 1, 4 was modified into 2 and so on. After this another column was created for the mean of these two factors, which became the official variable for
extraversion which was used for the analyses.
After modifying the data in excel, it was uploaded to SPSS where different analyses were conducted. The main analyses were made through regression and correlation. Regression is performed to identify the impact of one independent variable on the dependent variable. A regression analysis can be univariate, bivariate or multivariate. When conducting a bivariate analysis there is only one dependent variable and one independent variable. On the other hand, a multivariate analysis also includes control variables to make sure that there is not a third variable that may influence the relation between the independent and dependent variable. It also enables the researcher to predict the relations between more than two variables
(Bryman & Bell 2017, pp. 333-337; Doane & Seward 2016, pp. 493-495). For this study we used both bivariate and multivariate regression analysis.
Correlation analysis can be performed through Pearson’s r or Spearman’s rank correlation coefficient. These are similar to each other and often work equally as well but Pearson’s r is used when dealing with variables which are interval or ratio scales, whereas Spearman’s rank correlation coefficient is used when dealing with variables which are ordinal. Both of these correlation coefficients will have a value between 0 and 1, and can be negative or positive (Bryman & Bell 2017, pp. 338-341; Doane & Seward 2016, pp. 488-489). In this case Pearson’s r was used to determine correlations as this is most commonly used and more sufficient for interval variables.
Different descriptive analyses were also carried out to get an overall view of the data. Thereafter the independent variables were analysed through Pearson’s r in relation to the dependent variable to see possible connections. We also conducted a regression analysis to find predictions of how the independent variables affected the dependent variable, as well as to see if these were significant. VIF-values were also calculated to detect issues related to multicollinearity. Since independent variables often correlate to each other, this was an essential step in the analysis. Variance inflation factor (VIF) is a test that calculates the level of multicollinearity. Any value under 10 indicates that this problem has been avoided (Doane & Seward 2016, pp. 573-574).
4.6 Quality standards
4.6.1 Reliability
Reliability is a measurement which concerns the trustworthiness of a study. This is affected by examining the results of a study. It means that if the same type of study should be conducted at another time, the results should be similar to the original study. Therefore,
however be hard to achieve in studies that concerns individual’s thoughts and emotions, as this changes over time.
The reliability of this study is considered to be high, if it was to be conducted on a similar target group. Reason for this is because values differ depending on how old people are and where they are geographically located. There is therefore no guarantee that individuals of other backgrounds have the same mindset and will answer in a comparable way. This study is however based on theories which have been studied for many years, meaning that there is a level of stability as the results have been consistent over time. Some of the questions included in the survey have been formulated in the same way as the Swedish SOM-institute which conducts independent opinion polls (SOM-institutet 2019). This was primarily done to increase the reliability and replicability of the study since the questions will be interpreted in the same manner by the respondents, which further enables comparison to the Swedish
population. The factors included in the survey were all explained through our definition which was based in the theory as well. By defining the factors in the survey, we left little room for the respondents to interpret them and deviate from the intended definition. Should another researcher use the same factors and define them in another way the results may differ.
Bryman and Bell (2017, p. 266) also explain that conducting a pilot study beforehand may increase the reliability as it is an opportunity to test the survey and get feedback from
participants to improve those questions that are deficient. As recommended, a pilot study was conducted which gave an insight to how the questions were interpreted. A few of the
questions caused uncertainty and were therefore altered for the official survey.
The internal reliability concerns the researchers’ interpretation of how the analysis and empirical study should be conducted. This also includes interpretation of indicators which are meant to measure a concept (Bryman & Bell 2017, p. 174). As this study uses a quantitative method through SPSS, the capacity to interpret the empirical data is limited. Therefore, the internal reliability in this case is affected by how the researchers interpreted the theories and formulated the survey questions. As previously mentioned, some of the questions were taken from independent institutes and by the definitions explained in theory, leaving no room for internal interpretations. Hence why the internal reliability is considered high.
4.6.2 Validity
Validity is a concept that is closely linked to reliability. In order to achieve a high validity, it is essential that the reliability is high. This means that high validity cannot be achieved when the reliability is low. There are two main types of validity; external validity and internal validity. The internal validity concerns issues related to causality. This is a central principle when conducting a quantitative research which involves different variables. The issue in this case is whether the researcher is able to make accurate conclusions that a certain variable affects another, and that there are no other factors affecting these (Bryman & Bell 2017, p. 69). Furthermore, the use of cross-sectional design can decrease the internal validity as it may aggravate the ability to see causal correlations (Bryman & Bell 2017, p. 81).
The external validity concerns the issue of generalisation. In this case the use of sampling is essential to determine if generalisation is possible beyond the context of which the study has been conducted (Bryman & Bell 2017, p. 69). Since this study used a non-probability sample, a generalisation beyond this exact group is not possible. In turn, this decreases the external validity. However, since the main purpose is to see connections between the variables in this precise target group, it is not determinant for the entire study’s validity.
Furthermore, construct validity aims to understand if the measurement used for a concept actually measures what is intended (Bryman & Bell 2017, p. 176). This often concerns studies where various indicators are used to measure a concept. Indicators can either be direct or indirect where a direct indicator means that the concept has a clearer connection in terms of comprehension of the indicator (Bryman & Bell 2017, p. 170).
The survey of this study included both indirect and direct indicators. The indirect indicators were those that measured extraversion. In order to assure high validity, the indicators used to measure this concept were taken from the SOM-institute. There were also two questions regarding this in order to reassure high construct validity. Having two questions enabled us to compare the results in SPSS and assure that the intended concept was measured correctly. The other indirect indicator was used to measure self-monitoring. Due to limited time and trying to keep the survey short and consist only one question regarding this was included. This question was constructed based on Snyder and Gangestad’s (1982) theory regarding self-monitoring. However, the original study contained 25 questions and was aimed to only measure
self-monitoring. These questions in combination with each other work as one unit to measure the
whole concept and would not have been as effective if only one or two of these had been selected. Therefore, we created a question that would capture the core of this concept.
Since our target group regularly uses social media it was appropriate to conduct the study on one of these social media platforms, which in our case became Facebook. The survey also included control questions to further ensure that the obligatory criteria were fulfilled. By using questions from the SOM-institute we were able to compare their results to ours and therefore conclude that our data material was representable for our target group.
4.7 Source criticism
Source criticism is a process where different sources of information are evaluated. It is often used in scientific research where the person conducting the study has to collect information from various sources and needs to be certain that these are correct (Thurén 2005, pp. 15-16). According to Esaiasson (2012, pp. 281-285) there is a model that can be used to evaluate sources in a systematic manner. This includes four principles which are; authenticity, time
elapse, dependence and bias.
Authenticity indicates that the source should be truthful and credible, meaning it should be
what it claims to be. This is a very common issue, especially with internet sources as they might spread false information. Time elapse explains that a source has a higher credibility the more recent it is. This can also be connected to internet sources where it is important to check when the information last was updated. Dependence means that the information should not have been affected by other sources. This could be the case when information passes through multiple individuals. In order to avoid this, the researcher should always try to trace back the information to its original source. The last principle which is bias means that the information should be impartial and not affected by personal interests. These four principles are also explained by Thurén (2005, p. 13) meaning that they are common and widely used in researches.
Throughout the entire study these principles have been applied when collecting information. To ensure that we collected data that was authentic we searched for scientific articles that
stamp means that the researchers vouch for the article and its presented information. These types of articles are therefore more trustworthy and increase the trustworthiness of a study when being used (Rienecker & Stray Jørgensen 2017, p. 220). For this reason, peer reviewed articles have been used to the greatest possible extent in order to strengthen our arguments. The use of peer reviewed articles supports the principle of authenticity.
Some of the information have been from more recent sources and some from older. In a few cases we decided to use the newer sources to favour the principle of time elapse. This has been the case for statistical information that has been used in this study to support our arguments. The used sources are official institutes and organisations such as Statista (2020). By using these we decreased the chance of bias. Furthermore, some theories have been found through peer reviewed articles. In these cases, the presented information has been traced back to its original source and been collected directly from there, favouring the principle of
dependence.
Even though these principles of source criticism have been considered, we are aware that some sources still may be lacking in terms of different aspects, such as not being free of bias. In attempt to reduce possible errors connected to this, multiple sources regarding the same information has been used. This has particularly been the case when using scientific articles and other literature such as books.
4.8 Ethics
When conducting a study, ethical considerations have to be included since it affects how the data should be collected and handled. These considerations are crucial when managing data concerning individual and other personal information. Bryman and Bell (2017, s. 141) explain that these ethical rules include information, voluntarism, consent, integrity, negative effects, confidentiality and anonymity. First and foremost, it is important that the participants are aware of the purpose of the study and the included stages. It should also be known to the respondents that the participation is voluntary and could be cancelled at any time during the process. The rules regarding confidentiality and anonymity mean that no unauthorised individuals will have access to the data, and that it will only be used for scientific purposes. Lastly it is of importance that the researcher does not provide false or misleading information regarding the study in order to avoid any kind of damage towards the respondents (Ahrne & Svensson 2015, p. 29; Bryman & Bell 2017, p.141).
Since this study included a survey and collection of personal data, we consciously followed these ethical rules. When sharing the survey in the Facebook groups we included information regarding the purpose of study and the participation. The respondents were made aware that their participation would be anonymous and that no unauthorised personnel would have access to the data. Collecting answers through Google forms guaranteed anonymity since no personal data such as IP-address was saved. Neither did the questions included ask for any personal data that could be traced back to a certain person. Furthermore, instructions about the survey and what it included was also given, meaning that it was completely voluntary to participate and that they could cancel their participation at any time. No false or misleading information was provided, and the survey did not include any sensitive or harmful questions which could cause damage.
4.9 Method criticism
This study is based on the data collected from the survey, which means that the process of creating this and processing it is crucial for the results. Since the purpose concerns social
media, which is a known subject for most people today, there was an understanding of this beforehand. We were aware that this could affect the thought process as well as how certain phenomenon were perceived. This is according to Bryman and Bell (2017, pp. 60-62) a common issue as subjective thoughts and values may interfere with the scientific process. In order to use this prior knowledge to our advantage, theories were constantly researched to verify information and decrease our own subjective thoughts. The pilot study that was conducted was also done to eliminate these kinds of errors.
Through the pilot study feedback was received from participants which gave an insight to how things were perceived from another angle. This led to reformulation of certain questions. For example, it was brought to our knowledge that people underestimate the number of influencers they follow. Therefore, the answering alternatives were increased and made into larger intervals.
Because of the nature of this study, we had limited time and resources to collect the data. The collection was therefore made through a non-probability sample which normally is not the ultimate choice for these types of research. This could have affected the processing of data, whereas more accurate generalisations could have been done if the data was collected through a probability sample.
5 Results
This part of the study will illustrate the empirical data which was collected through our survey. The various analyses which were conducted will also be presented in this chapter in order to either reject or accept the hypotheses.
5.1 Descriptive analysis
Bryman and Bell (2017, p. 192) explain that attrition rate is an error source which does not concern the sampling process. Itis rather affected by the respondents who by various reasons are not able to or refuse to participate and answer correctly. Therefore, when conducting a study, the researcher should calculate the possible attrition rate and add this number to the total amount of answers required for the study. Since this study conducted a self-selection sample, we had less control over the number of participants and the attrition rate cannot be calculated. However, this could also have been an advantage since the participants themselves chose to participate, which increased the amount of completely answered surveys.
The survey was published in five different Facebook groups for women where they were available to answer for a period of two weeks. When this period ended the survey no longer accepted answers and the data was collected and saved. This included a total of 584 answers. Due to having a specific target group, 26 of these answers were removed as they were not between the ages of 16-35 or had made a mistake when writing their age making it invalid. Another 4 answers were excluded as they identified with another gender than female. 1 person did not have Instagram and 29 people did not follow an influencer and were therefore also removed. This concluded in 524 valid answers that were further analysed. The distribution of the ages varies where the youngest respondents are 16, and the oldest is 35. The mean age of the collected data is 23 years.
Furthermore, there was also a question regarding which factor the individuals valued the most when deciding to follow an influencer. 39.9% chose entertainment as the leading factor, followed by trustworthiness which was chosen by 29%. The third most popular option was similarity which was answered by 11.8%.
Furthermore, an independent-sample t-test was conducted to compare the degree of extraversion between our sample and a nationally representative sample of the same target group (females aged 16-35) in the Swedish population. There was a significant difference in extraversion between our sample (M = 3.47, SD = 0.95) and the Swedish population (M = 3.34, SD = 0.58); t = (717) = -2.15, p = 0.032 (two-tailed).
To favour the descriptive analysis there was also an open question asking which influencer the respondents liked the most. Therese Lindgren, which is a Swedish influencer, stood out as she was mentioned by 70 respondents. In second place came Helen Torsgården who was mentioned by 34 people, and closely after came Bianca Ingrosso who was chosen by 33 people. Other names that were mentioned included Antonija Mandir, Joakim Lundell & Jonna and Kenza.
5.2 Correlation analysis
Furthermore, analyses were conducted through Pearson’s to observe correlations between different variables. This was firstly done to find connections between the factors
trustworthiness, knowledge, attractivity, commerciality, similarity and entertainment. The table below (Table 1) illustrates the results. As shown, all factors correlate with each other, but not all are statistically significant. The strongest positive correlation is found between trustworthiness and knowledge. This shows that when one of these variables increases with one unit, the other also increases with 0.617 units. The only negative correlation is found between the factors trustworthiness and commerciality. This illustrates that when one of these variables increases with one unit, the other one decreases with 0.095 units, and vice versa. Because of all the low values, the correlation between trustworthiness and commerciality is the only one that is considered strong. Furthermore, since correlation does not show causality, it is not possible here to say that one factor affects another as there could be a third variable affecting these.
Table 1: Correlation between factors
Pearson’s r was also used to determine the correlation between extraversion and introversion. The values of introversion were reversed in order to reflect extraversion and were thereafter compared with the data of the extraversion question. The results demonstrated r = 0.553
5.3 Regression analysis
Since there were multiple independent variables, a test had to be conducted to detect possible issues with multicollinearity. This was ensured by calculating the VIF-values. The results showed values between 1 and 2. Doane and Seward (2016, p. 574) explain that a VIF-value of 1 means that there is no variance inflation whereas a value of 2 indicates a mild variance inflation. However, since the values are close to 1 and much lower than 10, this issue has been avoided.
Thereafter a multiple regression analysis of the different factors was conducted. This was firstly done with each independent variable individually against the only dependent variable (purchase). The results of this was saved and the different factors were ranked from most statistically significant to least. They were thereafter combined in a regression analysis where they were added gradually starting from the most statistically significant variable to the least significant. This is shown in Table 2. p>α = 0.05 is not considered significant and will be rejected. According to the results, when all of the factors are combined in relation to purchase the most significant factor is commerciality. This shows p = 0.006 and β = 0.125 meaning that an increase of one unit in commerciality leads to a 0.125 increase in purchase. The second most significant variable is attractivity with p = 0.010 and β = 0.127. Similarity is the last significant variable with p = 0.017 and β = 0.111. This means that the hypotheses H1c, H1d and H1e are accepted. The remaining factors entertainment, trustworthiness and knowledge have p>α = 0.05 meaning that the hypotheses H1a, H1b and H1f are rejected.
Table 2: Regression analysis of factors towards purchase
Furthermore, a regression analysis was conducted to identify if extraversion had an effect on purchase. The following regression chart shows that extraversion is statistically significant with p = 0.024. β = 0.100 meaning that if extraversion increases with one unit, purchase increases with 0.100 units. Hypothesis H2 is therefore accepted.
Table 3: Regression analysis of extraversion towards purchase
Similar results are presented when self-monitoring was tested against purchase. With p = 0.000 this variable is also statistically significant but has a stronger effect on purchase with β = 0.206. An increase of one unit in self-monitoring leads to a 0.206 increase in purchase. Hypothesis H3 is therefore also accepted.
Table 4: Regression analysis of self-monitoring towards purchase
Exposure showed to have an effect on purchase as well which is illustrated in the table below (Table 5). This demonstrated that an increase of one unit in the number of followed
influencers leads to an increase of 0.147 units in purchase. This is also statistically significant as p<α = 0.05 meaning that hypothesis H4 is accepted. Furthermore, the number of hours spent on Instagram per day was also tested against purchase. This variable is also statistically significant with β = 0.158 implying that for every extra hour someone spends on Instagram, the purchase increases with 0.158 units. Hypothesis H5 is accepted.
