The influence of online ad avoidance on search engine marketing Jelmer Remco Helderman June, 2015

(1)

And then there was AdBlock:

The end of SEA?

The influence of online ad avoidance on

search engine marketing

(2)

(3)

SEA?

The influence of online ad avoidance on search engine

marketing

Master thesis, MSc, Marketing Management

University of Groningen, Faculty of Economics and Business June 22, 2015

Author

Jelmer Remco Helderman Student number: 1795325 Beren 11 9714 DW Groningen +31(0)6-15213281 jelmerhelderman@gmail.com First supervisor

prof. dr. Peter C. Verhoef

Second supervisor

(4)

The rapidly growing importance of the online channel and its growing advertising spending are accompanied by a decrease in advertising elasticities. One explanation for this lower advertising impact comes from the area of advertising avoidance. More specifically, this study aims to gain knowledge about the influence of ad-blocking (browser plugins that automatically filter all kinds of online advertising) in the context of search engine marketing. Hence the research question is as follows:

What is the influence of ad-blocking on advertising outcomes such as satisfaction and click-through? And is it still effective to advertise in search engines when consumers are adopting ad-blocking software?

The proposed model is tested using a self -administered survey and experiment of 167 respondents. The results of this study show that there is initial proof that ad-block usage mediates the relationship between perceived intrusiveness of advertising (an antecedent of ad avoidance) and satisfaction. This implies that ad-block users are less satisfied, when they are confronted with advertisements, than non-ad-block users. Second, there is initial proof that ad-block usage mediates the relationship between perceived intrusiveness of advertising and click-through (when consumers are confronted with advertisements). This implies that ad avoidant people that use ad-blockers are less likely to click on advertisements than those who do not use ad-blockers when presented with advertisements on a search engine’s results page.

This study does not provide reasons to believe that advertising in search engines is not effective anymore due to ad-blocking software. However, the author can provide management with insights for the future. In this study, the main reasons for the usage of ad-blocking technologies are

(5)

clicking on the advertisement consumers do not feel betrayed. Advertisers should also try to diminish the privacy concerns consumers might have, so that they are less inclined to install ad-blocking software. This can be done by ensuring consumers that they are in control of their personal information and informing them about how their information is stored in databases.

(6)

Nowadays the internet is becoming more and more important. When I started my master marketing I found that not much attention to online marketing was given in the curriculum.

I believe every marketer now and in the future should be aware of the growing potential of the online channel. Therefore a personal goal was to compensate for this ‘gap’ and gather knowledge about the online counterpart of marketing before finishing my master. This thesis in that sense is the final step. Moreover this thesis marks the end of an important phase in my life. Spending my time as a student in Groningen brought me a large amount of experiences which made me discover my limits and abilities on both intellectual as more personal levels.

First, I want to address a special word of thanks to my parents and of course my stepfather for their support. I also want to thank my girlfriend Rianne for her love, understanding and motivation during the sometimes hectic weeks that I spend in the library. Without these four people I would have never pushed myself to where I am today. Especially after a time of being in a very bad physical shape. Naturally I also would like to thank my friends for the amazing time we had and giving me the energy to succeed. Finally, I would like to thank my supervisors Peter Verhoef and Evert de Haan for their constructive feedback. In particular Evert for his motivation and time he invested in my thesis.

(7)

The ever growing importance of the online channel and its growing advertising spending are

accompanied by a decrease in advertising elasticities. Based on literature review explanations come from the area of advertising avoidance, advertising relevance and perceived privacy concerns. However the option to opt -out for online advertisements, embodied by ad-blockers, is proposed to influence the relationship between on the one hand ad avoidance, privacy concerns, ad relevance and satisfaction, click-through on the other. Using a self -administered survey and experiment (N= 167), the proposed model is tested for mediation with regression analysis. The main findings suggest that there is initial proof that ad-block usage mediates the relationship between perceived

intrusiveness, measuring ad avoidance, and the dependent variables (satisfaction and click-through). Given the increased spending in search engine marketing this thesis concludes with providing insights for advertisers and marketers acting on their behalf.

Keywords: Ad-block, ad avoidance, privacy concerns, ad relevance, click-through, satisfaction, SEA,

(8)

2. Research Context: Search Engine Marketing ... 3

2.1 Search Engine Advertising ... 3

2.2 Search Engine Optimization ... 5

3. Theoretical Framework ... 5

3.1 Ad-blocking ... 5

3.2 Online advertising avoidance ... 7

3.3 Privacy concerns ... 8 3.4 Ad relevance ... 8 3.5 Satisfaction ... 9 3.6 Advertising effectiveness ... 10 3.7 Control variable ... 11 4. Methodology ... 12 4.1 Data collection ... 12 4.1.1 Experiment ... 13 4.1.2 Survey ... 14 4.2 Data analysis ... 15

4.2.1 Factor and Reliability analysis ... 15

4.2.2 Mediation analysis ... 17

5. Results ... 18

5.1 Mediation Analysis 1: Satisfaction ... 18

5.2 Mediation analysis 2: Click-through ... 23

6. Discussion ... 28

6.1 Conclusions and implications ... 28

6.2 Limitations and future research ... 31

References ... 32

APPENDIX A ... 39

APPENDIX B ... 40

APPENDIX C ... 43

(9)

1

1. Introduction

According to a recent forecast, advertisers around the world will spend $234.75 billion on digital and mobile advertising in 2015, accounting for 39.6% of total media spending (eMarketer 2014). This digital and mobile advertising spending is expected to rise to an astonishing 57.9% of total media spending in 2018. Dinner, Van Heerde and Neslin (2014) note that the growth in the online channel occurs simultaneously with the growth in online advertising spending. While advertising budgets and online budgets in specific increase, advertising elasticities decrease. Assmus, Farley, and Lehmann (1984) were among the first to provide empirical generalizations on advertising elasticity (the effect of an increase or decrease in advertising on a market). In the 25 years subsequent to the research done by Assmus, Farley, and Lehmann it has been reported that the advertising elasticities decrease. This lower advertising response in recent times is because of increased competition, ad clutter and the arrival of the internet as an alternative information source and the consumer’s ability to opt out of television commercials (Sethuraman, Tellis, and Briesch 2011; p. 460). Another explanation for the lower advertising impact, and focus of the current study, comes from the area of advertising

avoidance. Ad clutter for instance is believed to be one of the major drivers of advertising avoidance (Anderson and Gans 2011; Cho and Cheon 2004; Speck and Elliot 1997). Anderson and Gans (2011) note that content providers rely on advertisers to pay for content. And that when consumers adopt ad-avoidance technologies (like TiVo and ad-blockers) that allow them to view content without ads, “consumers thereby siphon off the content without paying the price (Anderson and Gans 2011, p. 1).” The price consumers pay for content provision is the advertising clutter. Many personalized advertising thus seems to be unwelcome resulting in advertising avoidance (e.g. by blocking online advertisements, filtering email or subscribing to do-not-call/mail or track programs). Although estimates differ, Baek and Morimoto (2012) report that more than half of all American households avoid unwanted commercial messages by installing some sort of ad-blocking software on their computers. The most downloaded browser extension of all time is AdBlock Plus, which according to the company, is downloaded more than 300 million times and is available on all browsers on desktops, and is upcoming on mobile (Android and iOS) devices. A recent report of PageFair and Adobe (2014) declares that ad-blocking is going mainstream and argues that there are approximately 144 million active ad-block users globally (4.9% of all internet users). Currently in more popular literature the notion is held that this rapid growing adoption of online advertisement blocking poses a threat for advertisers and the business model that popular websites rely on (PageFair 2013;

(10)

2

As a consequence these users avoid or dismiss all kinds of advertisements irrespectively of whether or not they use ad-blocking software. Secondly, when ad-block users are exposed to advertising this might actually have a stronger effect.

The topic of advertising avoidance has been well-researched over the past decades (Abernethy 1990; Bellman, Schweda and Varan 2010) however this research has focused mainly on traditional media such as newspaper, television and radio. A clear definition of advertising avoidance, used by several authors, was developed by Speck and Elliot (1997, p. 61): “All actions by media users that

differentially reduce their exposure to ad content.” In the last fifteen years advertising avoidance was also related to the general online environment (Baek and Morimoto 2012; Cho and Cheon 2004; Edwards, Li and Lee 2002) and to social networking websites in specific (Kelly, Kerr and Drennan 2010). Much of the authors mentioned above focus more on display and banners advertisements. This study contributes to academic literature because it focuses on keywords generated ads which are more specific than general untargeted spam ads (e.g. banners ads). This study will focus on one aspect of internet marketingthat is called search engine marketing (SEM), consisting of search

engine optimization (SEO) and search engine advertising (SEA). According to Rutz and Trusov (2011)

search engine advertising is the leading tool for consumer acquisition within internet marketing. There are two types of results when using a search engine, organic search results and sponsored search results. Organic results refer to the (not paid for) engines search results that appear because of their relevance to a certain keyword. Companies can take several actions to improve their position on the organic search listings which are referred to as search engine optimization (hereafter SEO). These actions can take the form of improving the content of the company its website or trying to improve the ranking in the search engines results. SEO is aimed at getting more relevant visitors and conversions on a website.

(11)

3

Although the empirical findings of Yang and Ghose (2010) show that there is a positive

interdependence between organic and sponsored search advertising they did not account for the use of online ad avoidance tools like AdBlock, which completely offset the influence of SEA.

Therefore the main contribution of this study is to combine research on organic and sponsored advertising with research on consumer avoidance and ad-blocking of online advertising. The research questions of this study are:

Chapter 2 will now first discuss the context of search engine marketing and provide explanations and definitions. In chapter 3 the theoretical framework is presented. This entails an overview of the relevant literature on advertising avoidance and ad-blocking and its proposed impact on advertising. Chapter 4 will built on the previous and provide the research methodology including data collection and data analysis. In chapter 5 the results are discussed. Finally in chapter 6 the conclusions, limitations and suggestions for future research are discussed.

2. Research Context: Search Engine Marketing

2.1 Search Engine Advertising

How does the mechanism of SEA work? When consumers search on the web, the phrases or words they enter in a search engine are referred to as keywords. Advertisers have to pay search engines like Google or Yahoo! to appear next to the regular (organic) search results for that keyword in a search engine (Agarwal, Hosanagar and Smith 2011; Rutz and Bucklin 2011).

(12)

4

FIGURE 1

Search results example for Google.

Advertisers pay a search engine on a cost per click (CPC) basis. The CPC is determined by a second price auction mechanism. As Agarwal, Hosanagar and Smith (2011) explain this second price auction also establishes the position of the advertisement and the advertiser pays an amount that is equal to the minimum bid needed to secure that position. The ultimate goal of SEA is not to gain as many clicks as possible but to gain more relevant visitors and through this higher conversion rates on the company’s website. A conversion is any action of a consumer on the advertiser’s website that is valuable to the advertiser (Zenetti et al. 2014). In general (and in this thesis) a conversion refers to a website visitor that turns into a paying customer. In addition to that, a conversion can for instance also include a consumer requesting information, filling in a contact form, downloading a document and so on. Advertisers often use the conversion rate as performance indicator which is calculated by dividing the number of conversions by the total clicks on the ad during a time period.

Some important research finds that advertisers who are now spending huge amounts of their budgets in bidding (in sponsored search results) may place to emphasis on securing top positions (Agarawal, Hosanagar and Smith 2011). In their empirical work on SEA, Ghose and Yang (2009) found that keywords that hold better positions (e.g. the top) on the search engine results page will lead to higher click-through or conversion rates. However these keywords do not generate the most revenue or profit per se, which is also confirmed by Agarawal, Hosanagar and Smith (2011). This stream of research finds that the highest profits are made for keywords ranked in the middle.

(13)

5

2.2 Search Engine Optimization

Companies use search engine optimization in order to increase their visibility in the organic search results. Back when search engines like Google did not have the advertising function, the only way to attract potential customers to a website was by means of search engine optimization. So why use SEO nowadays? As is the case with SEA, search engine optimization is aimed on generating more relevant customers and conversions on the company’s website. The major difference is that with ‘white hat’ SEO this goal is reached by creating a high landing page quality which can boost the company’s organic ranking for certain keywords (Ghose and Yang 2009). This is important because organic rankings are determined by very complex and secret algorithms owned by the search

engines. And landing page quality and the relative importance to other links have shown to influence these algorithms positively. Ghose and Yang (2009) have found that when the landing page is of high quality this will increases conversion rates and decreases the advertisers cost per click.

According to Berman and Katona (2013) companies should use SEO because consumers tend to trust organic links more than sponsored links. Yang and Ghose (2010, p.618) also emphasize the

importance of SEO by mentioning there is empirical evidence for a ‘second opinion’ or

‘reinforcement effect’. The second opinion effect arises when an advertiser’s link is present in both paid and organic listings. And the consequence of this effect is that more consumers click on the advertiser’s link.

3. Theoretical Framework

In this chapter an overview of relevant literature is given on ad-blocking, online advertisement avoidance, privacy concerns and advertisement relevance. In addition to this, literature on satisfaction and click –through is discussed. Subsequently the hypotheses are presented. The resulting conceptual model can be seen in figure 3. Based on research in marketing and advertising the author theorizes that consumers in an online environment engage in ad-blocking because of ad avoidance, privacy concerns and irrelevance of ads.

3.1 Ad-blocking

(14)

6

As Vratonjic et al. (2012) describe, ad-blocking tools are made out of two mechanisms in order to block ads. The first mechanism prevents the loading of certain elements whose URLs match so called filter rules that are used to classify elements as advertisement. The second mechanism hides page elements that match a cascading style sheets selector. In practice this means that users only have to install an extension to their web browser and they will not be confronted with any type of advertising online anymore. When taking into account the definition of advertising avoidance as proposed by Speck and Elliot (1997), installing ad-blocking software is the ultimate action users can take to reduce exposure to advertising content online. To illustrate this let us return to the example of entering the keywords ‘hotel Groningen’ in Google. As opposed to the previous scenario this time the consumer has installed an ad-blocking extension in his or her web browser. As can be seen in figure 2, no advertisements can be found and the consumer only relies on the organic search results.

FIGURE 2

Search results example for Google with ad-blocking software.

Ad-blocking also raises concerns with advertisers and content providers who rely on advertisers for their existence. And indeed the quick diffusion of ad-blocking (currently most likely under consumers that exhibit high ad avoidance) technologies changes the mix of consumers that view advertisements (Anderson and Gans 2011). These authors use the analogy of the rise in ads broadcasted in the United States and argue that greater penetration of ad-blocking may lead content providers to raise advertising clutter. Consequently the volume of advertising that is aimed at those who not use ad-blocking is increased. Johnson (2013) states that when advertisers pay per-click, the increasing use of ad-blocking technologies will reduce the motivations of firms to engage in advertising. This has to do with the fact that advertising campaigns also have fixed costs that can be divided over fewer

(15)

7

3.2 Online advertising avoidance

The topic of advertising avoidance is already on the research agenda for several decades. However much of this work is focused mainly on traditional media (e.g. Abernethy 1990; Bellman, Schweda and Varan 2010; Speck and Elliot 1997). The body of research on advertising avoidance in an online environment is less well established. Duff and Faber (2011) propose that two of the most common reasons why people avoid advertising are intrusiveness and interference with another activity. Cho and Cheon (2004) propose that there are three antecedents of online advertising avoidance: perceived goal impediment, perceived ad clutter and prior negative experience. The argument underlying perceived goal impediment and ad clutter arises from the notion that consumers might be more goal-directed when they go online and thus perceive advertisements as more intrusive than other (offline) media ads (Cho and Cheon 2004, Edwards, Li and Lee 2002). And indeed Cho and Cheon (2004) show that online advertisements hinder consumers in their search, distract them and are being perceived as disruptive (translated as perceived goal impediment). Perceived goal

impediment showed to be the most important antecedent in explaining internet ad avoidance. Another important factor in explaining internet ad avoidance is perceived ad clutter. Consumers perceive ad clutter when they are exposed to an excessive amount of advertising. This may lead to irritation and avoidance. In an online environment this excessive component is also present (e.g. annoying pop-up ads, banners and advertorials) and might be a reason why consumers install ad-blocking extensions in their browsers. The third antecedent of internet avoidance according to Cho and Cheon (2004) is prior negative experience. These prior negative experiences can be described as situations where the consumer is confronted with deceiving stimuli, leading them to inappropriate websites (e.g. pop-up ads or banners that claim you have won a large amount of money). A previous negative experience has the effect that consumers will avoid the source of the negative experience (Cho and Cheon 2004).

Attitudes drive our thoughts, influence our feelings and affect people’s behavior (Hoyer, MacInnis and Pieters (2013). Cho and Cheon (2004) included the attitudinal consumers’ responses (cognitive, affective and behavioral) to advertising .Thus the attitudinal response to online advertising , that is advertising avoidance, can manifest itself in: cognitive avoidance, behavioral avoidance and affective avoidance. In their study, Cho and Cheon (2004) showed that the affective component of ad

avoidance is the most significant in internet ad avoidance, followed by the cognitive and behavioral components. A negative cognitive response occurs when a consumer holds more negative

(16)

8

Finally, behavioral ad avoidance originates from actions people take that are different from lack of attendance, such as scrolling down and clicking away from ad pages to avoid banner ads or purging pop-up ads (Cho and Cheon 2004). In this study ad-blocking is considered to be a consequence of the behavioral component of advertising avoidance since it entails a consumer action that is other than lack of attendance (like cognitive avoidance). In other words ad-blocking is an attitudinal response to online advertising that manifests itself in a behavioral way. Because ad-blocking is expected to arise from advertising avoidance this study proposes that ad-blocking is driven by perceived goal

impediment, perceived ad clutter and prior negative experience. Hence the first hypothesis is:

H1: Increased online ad avoidance will increase ad-block usage.

3.3 Privacy concerns

According to the report of PageFair and Adobe (2014) the main reasons why consumers install ad-blocking software are a complete lack of desire to view any advertising, privacy concerns and

intrusiveness of ads. Whereas some users find the ongoing display of advertisements on the internet annoying, other users are more concerned about their privacy (Vratonjic et al. 2012). A widely cited definition of information privacy is provided by Westin (1967, p. 7): “the claim of individuals, groups, or institutions to determine for themselves when, how, and to what extent information about them is communicated to others.” Concerns about this privacy arise when internet users perceive that there is a high degree of collection and use of their personal information by websites (Malhotra, Kim and Agarwal 2004; Hong and Thong 2013). Adding to this, Hann et al. (2008) argue that privacy is a key concern for consumers and argue that this is one of the main reasons why they use advertising avoidance techniques. Firms in recent times are constantly improving their ability to discover details about individuals in order to provide them with advertising (e.g. by tracking of web browsing or targeting). Although this may benefit the advertising firms, consumers do not benefit to the same extent and find this advertising annoying and perceive increased targeting as infringement on their privacy (Johnson 2013). However, Smith, Dinev and Xu (2011) apply some nuance to this and state that there is evidence of a privacy paradox. Individuals on the one hand state they have privacy concerns but on the other behave in ways that contradict their statements.

H2: Increased privacy concerns will increase ad-block usage.

3.4 Ad relevance

(17)

9

Specifically it is found that when ad relevance is high for the product advertised within the context of the website this will evoke more positive attitudes towards advertisements (Lee and Mason 1999). Kim and Sundar (2010) also have empirically found that ad relevance has a pivotal role in the attitudes consumers have towards websites and advertisements in a portal context (e.g. search engines). Furthermore, Van Doorn and Hoekstra and (2013) argue that customized advertising, from a consumer perspective, has the advantage that of higher relevance and fit. Although they do note that the personalization of ads based on identity and transactions is perceived as being intrusive by consumers. Therefore the customization of online advertising offers a “double-edged sword” that may positively influence purchase intentions, but it also will increase the perceived intrusiveness which has a negative effect on purchase intentions (Van Doorn and Hoekstra 2013, p. 341). As mentioned before one of the main reasons why consumers install ad-blocking software is the complete lack of desire to view any advertising and to move as many ads as possible from websites (PageFair and Adobe 2014). This lead the author to assume that ad-block users may be a different segment in the population that dismisses all kinds of advertising since they perceive advertising as irrelevant. Therefore the third hypothesis is:

H3: Perceived relevance of advertisements will decrease ad-block usage.

3.5 Satisfaction

A substantial part of the research done on (dis)satisfaction has focused on how consumers evaluate offerings on utilitarian and hedonic dimensions. Satisfaction thus has to do with feelings that result from positive evaluations made by consumers or from happiness with their decision (Hoyer, MacInnis and Pieters 2013). McKinney, Yoon and Zahedi (2002, p. 298) define overall satisfaction as “an

affective state representing an emotional reaction to the entire Web site search experience.” Furthermore, prior research has shown that website design influences online customer satisfaction (e.g. Schlosser, White and Lloyd 2006; Szymanski and Hise 2000). According to Szymanski and Hise (2000) a good designed website is fast does not include ad clutter and is easy-to-navigate. Clear becomes that convenience also is at play when people are browsing on the web. A website search experience that has less ad clutter is convenient and fast therefore should be evaluated by

(18)

10

H4: Ad-block users will be more satisfied, compared to non-ad-bock users, when they are not confronted with advertisements.

3.6 Advertising effectiveness

In a social media environment Kelly, Kerr and Drennan (2010) pose that when consumers ignore or dismiss advertising, advertising dollars are flushed down the drain and the future of social

networking websites as a lucrative advertising medium is in danger. A recent report of PageFair (2013) also points out that when consumers completely dismiss a huge proportion of online

advertising, they might be endangering the business model on which popular websites rely. However as mentioned before this study is questioning the endangerment of search engine advertising

because a proportion of internet users might be uninterested in advertisements anyway. This implies that there is heterogeneity among these users that avoid or dismiss all kinds of advertisements regardless of whether or not they use ad-blocking software. Lo, Hsieh and Chiu (2014) find that people may not click on keyword advertising but argue that it is still included in the path of eye movement. Search engine advertising thus can be unconsciously processed in the consumer’s memory. This thought is related to pre -attentive processing and the mere exposure effect. Zajonc (1968) is viewed as the first to describe the mere exposure effect and established that attitudes towards stimuli became more positive when these were repeated. Thus although click-through rates may be drastically decreased, advertisers should not only focus on clicks but also on the content of their advertisements (Lo, Hsieh and Chiu 2014). Since 2011 AdBlock Plus has kept a ‘whitelist’ of websites that can still expose internet users, who have installed the browser extension, to advertisements (the company even makes money from it). However only non –intrusive advertisements can be whitelisted. This implies that even ad-blockers can be confronted with

advertisements on the internet. Therefore the author finds it interesting to see if ad-block users react different from non-adblock users when they are confronted with advertisements. As a consequence, in this study ad-block users are compared with non-ad-block users when they are exposed to

advertisements on a search engine’s results page. Hence the following hypothesis is proposed:

H5: Ad-block users are less likely to click on advertisements than non-ad-block users when confronted with advertisements.

(19)

11

Since determining the best online metrics is beyond the scope of this study the author will only include click-through rates to determine SEA effectiveness. Remember that a click-through occurs when the user clicks on the advertisement and is taken to the advertiser’s website.

Finally, it is proposed that the drivers of ad-blocking also have a direct negative effect on the outcome variables satisfaction and advertising effectiveness. Consumers are probably dissatisfied with their online experience when they avoid online advertising, have privacy concerns and perceive advertisements as irrelevant. And this will also have negative effect on click-through rates as they do not click on the sponsored advertisements. Hence the final hypotheses are:

H6: The drivers of ad-block usage will have a direct negative effect on satisfaction. H7: The drivers of ad-block usage will have a direct negative effect on click-through rates.

3.7 Control variable

One socio- economic variable is also included in this research to control for the influence of other independent variables on the dependent variables. This control variable is the level of education. The author chose for this control variable because previous studies have shown that when people are less educated they tend to report lower behavioral or mechanical advertisement avoidance as compared to higher educated people (Rojas-Méndez, Davies, and Madran 2009; Zufryden, Pedrick and Sankaralingam 1993). Shavitt, Lowrey and Haefner (1998) for instance find that higher educated consumers rely less on advertising than lower educated consumers.

(20)

12

4. Methodology

In this chapter the author in paragraph 4.1 will describe how the data was collected and provide some demographic information about the sample. Thereafter the procedure is explained for the experiment and the survey. In paragraph 4.2 the principal component analysis and the reliability analysis are discussed where after the different analysis to test the hypotheses are explained.

4.1 Data collection

(21)

13 TABLE 1 N= 167 # % Gender Male Female 78 89 46,7% 53,3% Education High school MBO Bachelor Master 25 27 89 26 15% 16,2% 53,3% 15,6% Occupation Student Working Unemployed Retired Self-employed Other 70 67 7 1 17 5 41,9% 40,1% 4,2% 0,6% 10,2% 3% Browser info Chrome Firefox MSIE Safari 66 73 7 21 39,5% 43,7% 4,2% 12,6% Ad-block user Yes No 80 87 47,9% 52,1% Mean St. Deviation Age 28,58 9,54 Demographics 4.1.1 Experiment

(22)

14

The respondents were regarded to be in one of the four situations as can be seen in table 2. Whether or not the respondent uses or has the intention to use an ad-blocker was self-declared in the survey afterwards. In the first situation the respondent is already using or has the intention to use an ad-blocker and is presented with advertisements on the search engines results page. As is proposed in the conceptual model this has the most negative effect on satisfaction and click –through. In the second situation the respondent is already using or has the intention to use an ad-blocker and is not presented with advertisements on the search engines results page. The author expects this to be a more neutral situation. Third, the respondent is not using or does not have the intention to use an ad-blocker and is presented with advertisements on the search engines results page. Finally, the respondent is not using or does not have the intention to use an ad-blocker and there are no ads displayed on the search engines results page.

TABLE 2

Is given ads

Yes No

Has an ad-blocker Yes -- +

No +/- +

Different scenarios in experiment

In the experiment click/click through rates were measured as well as an examination on which position (e.g. 1st, 2nd link) the participants click. This important because it is the consumer who is carrying out a search query and not the search engine. And because literature shows that position does have an effect on click-through rates (e.g. Ghose and Yang 2009). Impressions were not measured since the treatment group had different conditions than the control group.

4.1.2 Survey

The experiment was followed by a (self-administered) survey online in order to determine the drivers of ad-blocking and to measure satisfaction. In this survey consumers answered questions about satisfaction of the search engine, ad-blocking, online ad avoidance, privacy concerns and ad relevance (see Appendix B). Furthermore, questions were added about whether or not the

(23)

15

multi-item scales from Schumann, Wangenheim and Groene (2014) where used. Satisfaction was measured with single-item satisfaction (1 = very dissatisfied, 7 = very satisfied) adopted from Reichheld (2003) and multi-item scales (on a scale from 1= very dissatisfied to 10= very satisfied) adopted from Keiningham et al. (2007) and these were adjusted to the context of this study. The level of education was asked via a direct question with four categories (see Appendix B).

In order to identify and eliminate potential problems the questionnaire was pretested on a small sample (N=20) of respondents that were drawn from the same population as the respondents from the actual survey. In this test the respondents were asked some additional questions about how they perceived the duration of the experiment and survey and if they understood all questions. This resulted in deleting the perceived lack of incentive items that were also proposed by Cho and Cheon (2004). The rationale behind this was that respondents either did not understand the questions or commented that no incentive is given in a sponsored search context.

4.2 Data analysis

In order to test the conceptual model, mediation was tested through regression analysis. First the data had to be prepared for analysis. Second, the satisfaction scores had to be on the same scale, since item A1 was on a 7-point scale and item A2 on a 10-point scale. This was done by

standardization, meaning that the mean 𝑋 was subtracted from each score and then was divided by the standard deviation, or 𝑧𝑖 = (𝑋𝑖− 𝑋)/𝑠 . Also several items had to be correctly coded. This

resulted in the recoding of items B 11, 13, 14, 16 and 17 in the opposite direction (see Appendix B).Then a factor analysis was conducted to determine if there were too many items for further analysis. Finally the conceptual model was tested through (logistic/linear) regression analysis.

4.2.1 Factor and Reliability analysis

After data collection a factor analysis (principal component analysis) was conducted to determine if there were indeed three factors that explain ad-blocking. In this thesis orthogonal rotation (varimax) was used. First, the Kaiser-Meyer-Olkin (KMO) statistic was checked to assess whether factor analysis was appropriate. The KMO-statistic was used to compare the magnitudes of the compared

(24)

16

TABLE 3

Rotated Component Matrix Component Cronbach’s Alpha Items

1 2 3 4

Perceived Intrusiveness

1. Internet ads make it harder to browse Web pages.

,749

,894 2. Internet ads slow down Web

page downloading.

,559 3. Internet ads make Internet

navigation difficult.

,780 4. Internet ads disrupt my viewing

of Web pages.

,733 5. Internet ads disrupt the

reception of desired content.

,779 6. Internet ads intrude on my

search for desired information.

,756 7. Internet ads distract me from the editorial integrity of Web pages.

,765 8. I think the amount of advertising

on the Internet is excessive.

,643 9. I think the amount of advertising

on the Internet is irritating.

,633

Prior Negative Experience

11. I am dissatisfied with my decision to click Internet ads.

,774

,687 13. I am not happy with my earlier

decision to click Internet ads.

,770 16. Clicking Internet ads does not

help me improve my personal performance.

,503

17. I think that my Internet ad use does not improve my productivity.

,611

Privacy Concerns

1. I am concerned about threats to my personal privacy.

,804

,83 2. Consumers have lost all control

over how personal information is used.

,659

3. I am concerned that a person can find private information about me on the Internet.

,873

4. I am concerned that information I submit on the Internet could be misused.

,835

Ad Relevance 1. I will receive useful information through online ads.

,828

,86 2. Online advertisements will be

interesting to me.

,814 3. Online advertisements will be

worth paying attention to.

,889 4. I will see online ads that are

relevant to me.

,743

(25)

17

Literature stated that ad avoidance was explained through three factors: perceived goal impediment, perceived ad clutter and previous negative experiences. However in the factor analysis these three factors where rearranged. The items that were measuring perceived goal impediment and ad clutter loaded on the same factor. This however was not entirely unexpected because both have the same underlying argument according to Cho and Cheon (2004). Namely, that consumers might be more goal –directed when online and thus perceive advertisements as more intrusive. Item B10 was deleted because it had a very low communality (0,277 < 0,4). Item 14 was problematic because it loaded weakly on factors two and four, and was therefore dropped. Items 12, 15 and 18 were problematic as well because they did not clearly belong to the ad avoidance factors and were also dropped. After reducing the number of items by means of a factor analysis all factors showed enough internal consistency reliability because they were higher than 0,6. The factor scores were saved for the regression analyses. This has the advantage that these factor scores are standardized and are not correlated with each other, preventing multicollinearity problems.

4.2.2 Mediation analysis

The previous hypotheses and the conceptual model (see figure 3) propose a mediation effect. In order to test if mediation is in place the following four conditions must hold according to the Baron and Kenny (1986) test: (1) the effect of the independent variables (ad avoidance, ad relevance and privacy concerns) on the dependent variable satisfaction/click –through has to be significant(c-path). (2) The effects online ad avoidance, ad relevancy and privacy concerns on ad-block usage have to be significant (a-path). (3) The effect of ad-block usage on click-through and satisfaction has to be significant(b-path). (4) And finally (the c’-path) when ad-block usage is included, the effects of the independent variables on the impact of advertising have to be non –significant or at least lower its effect.

In the first mediation analysis satisfaction was the dependent variable. The constant term is 𝛼0, the

coefficient is 𝛼𝑘 and 𝜀 is the error term. The adverb PerInt denotes perceived intrusiveness. NegExp

stands for prior negative experience. AdRel is the perceived relevance of advertisements. PrivCon are the privacy concerns people have. Ad-block measures whether or not people already use

(26)

18

be conducted (binary logit model).

𝑝

In the a-path denotes the probability that people use ad-blockers. Hence the following regression equations were used:

(C) 𝑺𝒂𝒕𝒊𝒔𝒇𝒂𝒄𝒕𝒊𝒐𝒏 = 𝜶𝟎+ 𝜶𝟏𝑷𝒆𝒓𝑰𝒏𝒕 + 𝜶𝟐𝑵𝒆𝒈𝑬𝒙𝒑 + 𝜶𝟑𝑨𝒅𝑹𝒆𝒍 + 𝜶𝟒𝑷𝒓𝒊𝒗𝑪𝒐𝒏 + 𝜶𝟓𝑬𝒅𝒖 + 𝜺

(A)

𝐥𝐧(_𝟏−𝒑𝒑 ) = 𝜶𝟎+ 𝜶𝟏𝑷𝒆𝒓𝑰𝒏𝒕 + 𝜶𝟐𝑵𝒆𝒈𝑬𝒙𝒑 + 𝜶𝟑𝑨𝒅𝑹𝒆𝒍 + 𝜶𝟒𝑷𝒓𝒊𝒗𝑪𝒐𝒏 + 𝜶𝟓𝑬𝒅𝒖

(C’) 𝑺𝒂𝒕𝒊𝒔𝒇𝒂𝒄𝒕𝒊𝒐𝒏 = 𝜶𝟎+ 𝜶𝟏𝑷𝒆𝒓𝑰𝒏𝒕 + 𝜶𝟐𝑵𝒆𝒈𝑬𝒙𝒑 + 𝜶𝟑𝑨𝒅𝑹𝒆𝒍 + 𝜶𝟒𝑷𝒓𝒊𝒗𝑪𝒐𝒏 + 𝜶𝟓𝑨𝒅 −

𝒃𝒍𝒐𝒄𝒌 + 𝜶𝟔𝑨𝒅𝑺𝒆𝒆𝒏 + 𝜶𝟕𝑨𝒅 − 𝒃𝒍𝒐𝒄𝒌 ∗ 𝑨𝒅𝑺𝒆𝒆𝒏 + 𝜺

In the second mediation analysis the dependent variable was click –through and also the c’-path was tested with a logistic regression. However the same equation for the a-path as in mediation analysis one was used since it is the same for both analyses. The constant term is 𝛽0, the coefficient is 𝛽𝑘 and

𝑝

denotes the probability that people click –through. Since only the situation where respondents were confronted with advertisement was used (N=80), no moderator effect was proposed. The last two equations were as follows:

(C) 𝐥𝐧(_𝟏−𝒑𝒑 ) = 𝜷𝟎+ 𝜷𝟏𝑷𝒆𝒓𝑰𝒏𝒕 + 𝜷𝟐𝑵𝒆𝒈𝑬𝒙𝒑 + 𝜷𝟑𝑨𝒅𝑹𝒆𝒍 + 𝜷𝟒𝐏𝐫𝐢𝐯𝐜𝐨𝐧 + 𝜷𝟓𝑬𝒅𝒖

(C’) 𝐥𝐧(_𝟏−𝒑𝒑 ) = 𝜷𝟎+ 𝜷𝟏𝑷𝒆𝒓𝑰𝒏𝒕 + 𝜷𝟐𝑵𝒆𝒈𝑬𝒙𝒑 + 𝜷𝟑𝑨𝒅𝑹𝒆𝒍 + 𝜷𝟒𝐏𝐫𝐢𝐯𝐜𝐨𝐧 + 𝜷𝟓𝑨𝒅 − 𝒃𝒍𝒐𝒄𝒌 +

𝜷𝟔𝑬𝒅𝒖

5. Results

In this chapter the results of this empirical research will be presented and discussed. Paragraph 5.1 will describe the first mediation analysis. Hereafter in paragraph 5.2 the second mediation analysis with click –through as dependent variable will be discussed. Finally an overview of the

confirmed/rejected hypotheses within the conceptual model is given in figure 6.

5.1 Mediation Analysis 1: Satisfaction

C-path

As was discussed in paragraph 4.2.2, in order to test if mediation is in place four conditions must hold. The first step in the analysis discusses the direct effect of de independent variables on the dependent variable (satisfaction), or the C path. This first step consists of a multiple regression analysis, in order to measure the effect of the independent variables on the dependent variable (satisfaction). As can be seen from table 4 the independent variables explain approximately five percent of the variance in satisfaction. The overall model is significant (p= 0,020 < 0,05). Two

(27)

-19

0,126), in support of hypothesis 6. This means that when a person is more ad avoiding this will lead to a decrease in satisfaction. Also in support of hypothesis 6, there was a significant positive relation between ad relevance and satisfaction, β= 0,135, p=0,032 < 0,05. This implies that when a person perceives advertisements as being relevant will lead to an increase in satisfaction. When an advertisement is then perceived as not relevant this has a negative effect on satisfaction. Privacy concerns (p= 0,363 > 0,05) and prior negative experience(p= 0,639 > 0,05) did not significantly contribute in explaining the variance in satisfaction, rejecting hypothesis 6. Hypothesis 6 proposed that ad avoidance, privacy concerns and ad relevance had a direct negative effect on satisfaction. This hypothesis can thus be partially supported. Testing the effect of the dependent variables on the dependent variable (C) may have low power so it is essential to provide an estimation of the indirect AB effect.

TABLE 4

Independent variables B Std. Error Sig.

Perceived intrusiveness -0,126 0,062 0,044

Prior negative experience -0,030 0,063 0,639

Ad Relevance 0,135 0,062 0,032

Privacy Concerns -0,057 0,062 0,363

Control variable

Education -0,272 0,136 0,047

Model Summary: Adjusted R2 0,051

ANOVA: Sig. 0,020

Output regression C path A-path

(28)

20

(p= 0,073 > 0,1) and has a positive effect (β= 0,30). This means that when people are more concerned about their privacy they tend they are more likely to decide to use ad-blockers, marginally in support of hypothesis 2. There was not enough evidence to state that ad relevance (p= 0,935 > 0,05) does contribute in explaining why people use ad-blockers, rejecting hypothesis 3.

TABLE 5

Perceived intrusiveness 0,423 0,172 0,014

Prior negative experience 0,295 0,169 0,082

Ad Relevance -0,014 0,168 0,935

Privacy Concerns 0,304 0,169 0,073

Education 0,989 0,369 0,007

Model Summaries: Cox and Snell 0,112 Nagelkerke 0,150

Omnibus: Sig. 0,001

Hit rate 64,1

Output logistic regression B-path

The third step in the analysis discusses the effect of the mediator variable on the dependent variable controlling for the moderator variable. As can be seen in table 6, the overall model is marginally significant (p= 0,053 < 0,1). Hypothesis 4 suggests that ad-block users will be more satisfied, compared to non-ad-bock users, when they are not confronted with advertisements.

TABLE 6

Prior negative experience -0,030 0,064 0,633

Ad Relevance 0,140 0,063 0,027 Privacy Concerns -0,058 0,063 0,360 Ad-block Usage 0,182 0,178 0,307 Ads Shown 0,116 0,175 0,506 Ad-block*Ads Shown -0,344 0,250 0,170 Control variable Education -0,275 0,141 0,053

Model Summary: Adjusted R2 0,045

ANOVA: Sig. 0,053

(29)

21

When no advertisements are shown to the respondents (ads-shown = 0), the relationship between ad block usage (see table 6) and satisfaction is not significant (p= 0,307 >0,1). The results show there was not enough evidence to accept this notion. Therefore hypothesis 4 is rejected. However the significant interaction effect of the variables ad-block usage and ads shown indicates a moderator effect. This interaction effect is significant when we use one sided test and a confidence level of 90%. Thus when advertisements are shown (ads-shown = 1), the effect of ad-block usage on satisfaction is marginally significant (p/2= 0,085 <0,1) and is negative (β= -0,344).

C’-path

Ideally the author would have opted for a bootstrapping procedure as proposed by Preacher and Hayes (2004) to analyze the effect of the independent variables on the dependent variable controlling for the mediator and moderator. Because a bootstrapping procedure was not feasible (the mediator is a dichotomous variable) a Sobel test was conducted. The Sobel test estimates the indirect effect of (path) ab in one single step, as depicted in the equation: √𝑏𝑎𝑏 2𝑠𝑒𝑎+ 𝑎2𝑠𝑒𝑏.

However note that the Sobel test is very conservative and might have low power when the sample size is smaller than 200. In this thesis the sample size is smaller than 200 (N= 167) implying that estimates should be used with caution. Remember that the a-path was estimated with logistic regression, creating a problem. In ordinary regression the scale is constant across equations whereas in logistic regression it is not (MacKinnon and Dwyer 1993). In order to make the regression

coefficients comparable across equations and to use them in a Sobel test, transformations were made with the following equations from Herr.

(1) Comparable coefficients: a = a * SD(X)/SD (M') (2) Standard Error (comp a) = SE (a) * SD(X)/SD (M')

In table 7 the resulting unstandardized regression coefficients are given and transformed into the unstandardized regression coefficients for the association between the independent variables and the mediator (a) and the standard error (Sa).

(30)

22

After transforming the coefficients from the a-path, also some modifications had to be made to the coefficients of the b-path. From table 6 we can derive that when no ads are shown to respondents, ad-block usage does not have a significant effect on satisfaction. When advertisements were shown to the respondents there was a marginal significant effect (p/2 = 0,085 < 0,1). To obtain the

coefficient (b) for the association between the mediator and the dependent (including the

dependent variables) the regression coefficients from ad-block usage and the interaction effect are summated: 0,182 + (-0,344)= -0,162. The standard error is calculated with the following formula: 𝜎̂𝑎𝑦

𝑎𝑥 = √𝑣𝑎𝑟(𝛽̂1) + 𝑍

2_(𝛽̂

3) + 2𝑍𝑐𝑜𝑣(𝛽̂1𝛽̂3). Resulting in (0.032+0.03+ (0.013*2))^0.5= 0.297.

Based on the previous regression analyses, perceived intrusivenesswas the only variable that was both significantly contributing to explaining satisfaction (c path) and positively attributes on the decision whether or not to use ad-blocking to the mediator variable (a path). In table 8 the results of the Sobel test are given where the dependent variable is satisfaction, the independent variable is perceived intrusiveness and the mediator is ad-block usage.

TABLE 8

Independent variable Test statistic Std. Error p-value

Dependent variable: Satisfaction

Output Sobel test 1

(31)

23

TABLE 9

Hypotheses Supported/Not

supported

Classification 1: “Increased online ad avoidance

will increase ad-block usage.”

Supported** Increased perceived intrusiveness** and prior negative experiences* will increase ad-block usage.

2: “Increased privacy concerns will

increase ad-block usage.”

Supported* Increased privacy concerns do increase ad-block usage

3: “Perceived relevance of

advertisements will decrease ad-block usage.”

Not Supported Advertisements that are perceived as relevant do not decrease ad-block usage.

4: “Ad-block users will be more

satisfied, compared to non-ad-bock users, when they are not

confronted with advertisements.”

Not Supported There was only a marginal significant, negative effect on satisfaction when advertisements were shown.

6: “The drivers of ad-block usage

will have a direct negative effect on satisfaction.”

Partially supported**

Perceived intrusiveness and ad relevance (when perceived as not relevant) have a direct negative effect on satisfaction but prior negative experienceand privacy concerns do not.

*p<0,1 **p<0,05

Summary results and hypotheses

5.2 Mediation analysis 2: Click-through

Heat maps

(32)

24

FIGURE 4 FIGURE 5

Advertisements shown No advertisements shown

Recall that in the previous mediation analysis the influence of ad avoidance, perceived ad relevance and privacy concerns on ad-block usage (a path) was already tested. This paragraph therefore continues with the c, b and c’-paths. As was previously stated, respondents are confronted with either organic and sponsored links or only organic links. Therefore, the author only tested for the situation when respondents are confronted with advertisements and do (or do not) use ad-blockers. In the next mediation analysis only the cases (N=80) where respondents are presented with

advertisements (see situation 1 in Appendix A) are used.

C path

(33)

25

regression analysis is conducted. As can be seen in table 10, the overall model is significant according to the Omnibus test (p=0,015 < 0,05). The overall fit of the model by the Cox & Snell R square = 0,162 and Nagelkerke = 0,216 and the classification table shows that 66,3% of the cases, are correctly classified. The significance of the estimated coefficient is based on the Wald’s statistic. In support of hypothesis 7, perceived intrusivenessshowed to be significant (p= 0,031 < 0,05) and has a negative effect on clicking on advertisements (β= -0,553). This implies that when people are ad avoidant this attributes negatively on the decision whether or not to click on advertisements (click-through). On a 90% confidence level, ad relevance showed to be significant (p= 0,074 > 0,1) and has a positive effect (β= 0,453), in favor of hypothesis 7. When people perceive advertisements as being relevant they this has a positive effect on the decision to click on these advertisements (when advertisements are then perceived as not being relevant this effect is negative). Prior negative experience (p= 0,140 > 0,05) and privacy concerns (p= 0,113) do not contribute in explaining why people click on sponsored advertisements, therefore the data did not provide evidence to accept hypothesis 7. Hypothesis 7 is then partially accepted.

TABLE 10

Ad Relevance 0,453 0,254 0,074

Education 0,079 0,523 0,879

Hit rate 66,3

Output logistic regression B path

(34)

26

less likely in deciding to click on an advertisement (implying a negative effect on click –through), marginally in support of hypothesis 5.

TABLE 11

Ad Relevance 0,477 0,258 0,065

Ad-block usage -0,806 0,547 0,140

Education 0,285 0,550 0,604

Hit rate 62,5

Output logistic regression C’path

As was the case in paragraph 5.1 a Sobel test is conducted to test for mediation and the same procedure is followed here. The a-path is modified according to equations 1 and 2 from paragraph 5.1. To make the b-path comparable across equations as well, the following equations are used. (3) Comparable coefficients: b = b * SD(M)/SD (Y'')

(4) Standard Error (comp a) = SE (b) * SD(M)/SD (Y'')

The resulting unstandardized regression coefficients are transformed into the unstandardized regression coefficients for the association between the independent variables and the mediator (a) and the standard error (Sa), as can be seen in table 12..

TABLE 12 Independent variable B logistic regression Std. Error SD a sa Ad-block usage -0,806 0,547 0,501 -0,217 0,147

Transformed coefficient b-path

According to the Baron and Kenny (1986) test only for the independent variable perceived

(35)

27

path). In table 13 the results of the Sobel test are given where the dependent variable is click – through for advertisements, the independent variable is perceived intrusiveness and the mediator is ad-block usage.

TABLE 13

Independent variable Test statistic Std. Error p-value

Dependent variable: Click-through

Output Sobel test 2

When the mediating variable is included, the significance level of the independent variable level drops to being insignificant. This means that when ad-block usage is included and advertisements are shown the level of ad perceived intrusiveness is insignificant, implying full mediation. People who are ad avoidant and use ad-blockers are thus less likely to click on advertisements than those who do not use ad-blockers when presented with advertisements in a sponsored search environment. In the second mediation analysis there was not enough proof to state that the level of education influences click-through. Figure 6 summarizes the confirmed/rejected hypotheses within the conceptual model.

FIGURE 6



Not significant,



Accepted,



/



Partially accepted,

Conceptual model with confirmed/rejected hypotheses

H1: Increased online ad avoidance will increase ad-block usage. H2: Increased privacy concerns will increase ad-block usage.

(36)

28

H3: Perceived relevance of advertisements will decrease ad-block usage.

H4: Ad-block users will be more satisfied, compared to non-ad-bock users, when they are not confronted with advertisements.

H5: Ad-block users are less likely to click on advertisements than non-ad-block users when confronted with advertisements.

H6: The drivers of ad-block usage will have a direct negative effect on satisfaction. H7: The drivers of ad-block usage will have a direct negative effect on click-through rates.

6. Discussion

The goal of this research was to investigate the influence of ad-blocking on click-through and satisfaction as well as to determine whether it is still effective to advertise in search engines. This concluding chapter outlines both the theoretical and practical implications. Finally, the limitations and suggestions for future research are discussed.

6.1 Conclusions and implications

The current research contributes to literature by doing empirical research in the context of search engine marketing to provide further insights on the relationship between online ad avoidance and satisfaction/click –through. Different in this research is that the author is accounting for the use of ad-blocking. Therefore this research tested the influence of ad-blocking on satisfaction and click-through in the context of search engine marketing. Recall that the research questions of this study are as follows:

First, the influence of ad-blocking on the outcome variables satisfaction and click-through is discussed. The results of the empirical research showed that there is initial proof that when

(37)

29

On the contrary not enough evidence was found that ad-block users, when no advertisements were shown, where more satisfied with search results than non-ad-block users.

Ad-block usage also showed to have a marginal negative influence on click-through. This is however not unexpected since, in the second mediation analysis, ad-block users are confronted with

advertisements. It is not unlikely that ad-block users tend to disregard advertisements when they see them, because ad-block users perceive advertisements as intrusive, have a prior negative

experiences or privacy concerns. Cho and Cheon (2004) have previously described this intentional ignorance as a negative cognitive response. The relationship between perceived intrusiveness and click-through is mediated by ad-block usage (however this was only tested for the situation where respondents were presented with advertisements). This means that ad avoidant people who are using ad-blockers are less likely to click on advertisements than those who do not use ad-blockers, when presented with advertisements on a search engine’s results page.

Second, to answer the question whether or not it is still effective to advertise in search engines when people are using ad-blocking software, the antecedents of ad-block usage are now discussed and managerial implications are given. As expected this study confirms that increased online ad avoidance explains why people decide to use ad-blockers. And this finding confirms the proposed theory in this thesis that ad clutter, perceived goal impediment (factored as perceived intrusiveness), and prior negative experiences are amongst the reasons to use ad-blockers. In addition to this, Kelly, Kerr and Drennan (2010) pose that there are also mechanical means that make ad avoidance

automatic. Hence it came as no surprise that ad-blocking is considered to be such an automatic technology. The online advertising industry can benefit from this outcome by making advertisements less intrusive so that consumers feel less urge to install ad-blockers. Especially for larger companies this is important because otherwise getting on ‘the white list’ of companies like AdBlock Plus is not an option. Ad avoidance was also negatively related to satisfaction and click –through. Namely, perceived intrusiveness was found to negatively influence both outcome variables. This has two implications. The first implication for advertisers is that they should try to avoid making goal impeding advertisements. In other words advertisers should produce advertisements that help consumers with their search for information, products, services or anything else. The second implication for advertisers is that advertisers should try to reduce their advertising so that it is perceived as less excessive by consumers.

(38)

30

Here the author agrees with Malhotra, Kim and Agarwal (2004) that firms (or marketers that act on their behalf) should ensure that their consumers can easily verify what kind of information is collected, how it used by the company and if the information is correct. When consumers thus have more control of their personal information and are informed about how this information is stored in databases they might be less inclined to use ad-blockers. Sheehan and Hoy (1999) add to this and advise that advertisers prior to information collection should build relationships with their potential customers. This ensures that trust and confidence on the part of the consumer increases. In this study there was not enough evidence to state that privacy concerns had a direct negative effect on satisfaction and click –through. Sheehan and Hoy (1999) argue that when consumers perceive privacy concerns they are most likely to solve this problem with avoidance behaviors (e.g. avoiding the advertisement and not click on it). Furthermore the authors find that by and large online consumers tend to be fairly unresponsive to dissatisfaction. It can be concluded that based on this study there is no reason to believe that advertising in search engines is not effective anymore due to ad-blocking software, although researchers and practitioners should not overlook the influence of ad avoidance and privacy concerns of consumers.

There was not enough evidence to state that ad relevance influences ad-block usage. This was an unexpected finding since Edwards, Li and Lee (2002) find that when advertisements are irrelevant to the consumer they will evoke feelings of intrusiveness and irritation leading to ad avoidance.

However ad relevance did show to influence satisfaction and click-through regardless of whether people use ad-blocking software or not. An implication of this is that advertisers (in SEA) should focus not only on securing top positions in bidding. Instead they should make their advertisements more relevant as they do in SEO. Animesh, Viswanathan and Agarwal (2011) also find that advertisers that focus only on rank as predictor of click-through may engage in sub-optimal bidding strategies. The authors argue that the rank in the sponsored results listing is moderated by the advertiser’s ability to differentiate itself competition. And with differentiating from competition in this context is meant competing on the level of title and text (referred to as the creative) of the advertisement, which conveys the positioning strategy.

Unexpectedly this study doesn’t confirm that prior negative experiences directly influence satisfaction or click –through. This finding is contradicting the stream of research on the

(39)

31

A possible explanation comes from Yang and Ghose (2010) who argue that search engine advertising is (directly) related to the search queries of the user and therefore might be considered as far less unwelcome than for instance display and banner advertising. Hence consumers might have less prior negative experiences with advertisements in a search engine context.

The level of education did show to positively influence the decision to use ad-blockers in this research. This is in line with previous research from several authors (Rojas-Méndez, Davies, and Madran 2009; Zufryden, Pedrick and Sankaralingam 1993) since in this study ad-blocking is regarded as a mechanical means to avoid advertising. Furthermore the level of education negatively influences satisfaction. This is probably due to the fact that higher educated people evaluate websites more critically. Finally, in this study that the level of education doesn’t influence click-through.

6.2 Limitations and future research

This study has several limitations that should be taken into consideration. These limitations originate primarily from a lack of data. First, there was only found initial proof (marginal significance) for the influence of ad-block usage on satisfaction when advertisements where shown to respondents. Therefore conclusions drawn from this research should be interpreted carefully. Second, this study used a Sobel test which has low predictive power because the sample size (N= 167) is smaller than 200. Therefore future research should incorporate a much larger data set where different companies measure the influence of ad-blocking. This sample would be also more representative than the sample of the current research as the respondents in this research are largely originating from the author’s network. Typically research in the area of search engine marketing incorporates large data sets by cooperating with large retailers or search engines using a huge amount of different keywords (e.g. Jerath, Liye and Young-Hoon 2014; Yang and Ghose 2010). In order to gain access to these larger data sets researchers are recommended to seek such cooperation.

(40)

32

ad-blocking usage by using a within -subject design where participants are subjected to more than one treatment condition.

While it was not a goal of this research to determine the best metric for measuring advertising effectiveness, it does not mean the author is unaware that there could be better ways for measuring the construct. For instance Abou Nabout, Lilienthal and Skiera (2014), who are also measuring advertising effectiveness with the number of clicks and cost per click, acknowledge that advertiser profit may be a better measure of advertiser effectiveness. A final direction for further research can account for the fact that in some scenarios consumers leave a website, return on a later moment in time and then decide which link has their preference (e.g. Agarwal, Hosanagar and Smith 2011).

References

Abernethy, Avery M. (1990), “Television Exposure: Programs vs. Advertising,” Journal of

Current Issues and Research in Advertising, 13 (January), 61–77.

Abou Nabout, Nadia, Markus Lilienthal, and Bernd Skiera (2014), "Empirical Generalizations in Search Engine Advertising," Journal of Retailing, 90 (June), 206-216.

Agarwal, Ashish, Katrik Hosanagar, and Michael D. Smith (2011), “Location, Location,

Location: An Analysis of Profitability of Position in Online Advertising Markets,” Journal of Marketing

Research, 48 (December), 1057-1073.

Anderson, Simon P., and Joshua S. Gans (2011), “Platform Siphoning: Ad-Avoidance and Media Content,” American Economic Journal: Microeconomics, 3 (4), 1–34.

Animesh, Animesh, Siva Viswanathan, and Ritu Agarwal (2011), "Competing "Creatively" in Sponsored Search Markets: The Effect of Rank, Differentiation Strategy, and Competition on Performance," Information Systems Research, 22 (March), 153-169.

Assmus, Gert, John U. Farley, and Donald R. Lehmann (1984), “How Advertising Affects Sales: Meta-Analysis of Econometric Results,” Journal of Marketing Research, 21 (February), 650-74.