«Technische Universität München As marketing expenditures in the Internet rush, effectiveness and efficiency become increasingly relevant. ...»
rewarded by the website, the reciprocity mechanism aims to create a desire to reward the website. Thus:
H4: Informing customers about the benefits of advertising relevance and appealing to reciprocity interact such that the total effect on a) acceptance of targeting and b) perceived intrusiveness of employing both mechanisms simultaneously is weaker than the sum of the effects if each of these mechanisms were employed individually.
3.3. Mechanism Related to Procedural Justice Consumers’ assessment of procedural justice is informed by the norms of openness, honesty, permission, and information access (Ashworth; Free, 2006). Overall, consumers consider organizations’ procedures fair when they are vested with control over the procedures (Son; Kim, 2008). Allowing consumers to control how their information is collected and subsequently used reduces their risk of privacy intrusions and communicates a company’s respect and value for them (Ashworth; Free, 2006).
Thus, acting according to the standards of procedural fairness increases the perceived trustworthiness of a company and increases consumers’ willingness to disclose information for targeted marketing (Culnan; Armstrong, 1999). Specifically granting consumers access to their information might reduce their uncertainty about
a website’s data collection and, thus, the psychological cost of targeting. Thus:
H5: Providing consumers with a high level of control by allowing them to view and edit their information a) increases their acceptance of behavioral targeting and b) reduces the perceived intrusiveness of targeted advertisements, compared with not allowing them to view and edit their information.
4. Empirical Studies
4.1. Study 1 Design We employed a 2 × 2 × 2 between-subjects experimental design by administering an online survey with a scenario technique. Respondents were recruited with the help of a professional market research firm and represented the audience of German newspaper websites. The final sample consisted of 469 responses (i.e., 51 completed surveys per manipulation and 61 completed surveys of a control group).
First, we presented the respondents a screenshot of a popular German news website and asked them to imagine they were surfing on this website. Then, respondents saw a flash layer overlapping parts of the news website. This flash layer contained a text message with a short greeting (“Dear visitor”) and three paragraphs that represent the experimental conditions (reciprocity, relevance, control). Each paragraph consisted of either a text aiming to increase the acceptance of targeting or a neutral text of similar length and complexity. Respondents then accessed the survey in which they rated their responses to this scenario. We adapted the scales in the questionnaire from prior studies and extensively pretested them to ensure validity and reliability.
The paragraphs, the scales, and their respective quality criteria are available upon request from the authors.
Respondents were randomly assigned to one of the eight experimental treatments or to a control group. Respondents in the control group saw the news website but no message related to the behavioral targeting of advertisements on the website. They rated the perceived intrusiveness of advertisements not denoted as behaviorally targeted and answered the respective questions measuring the control variables in our model.
4.2. Study 1 Results
4.2.1. Impact of targeting knowledge on perceived intrusiveness.
In a first step, we compared the perceived intrusiveness of advertisements shown to respondents in the control group with the perceived intrusiveness of alleged behaviorally targeted advertisements displayed to surfers in scenarios 1 to 8. A t-test revealed that respondents perceived advertisements identified as behaviorally targeted as significantly more intrusive than regular advertisements (T = 2.328, p =.020); regular advertisements received an average intrusiveness rating of 3.602, targeted advertisements received 4.121. Therefore, H1 was supported.
4.2.2. Impact of mechanisms on acceptance and perceived intrusiveness of ads.
In a multivariate main analysis, we investigated whether the mechanisms we developed affected the acceptance of targeting and the perceived intrusiveness of targeted advertisements using multivariate analysis of covariance procedures. We included the experimental manipulations reciprocity (present/not present), relevance (present/not present), and the level of control over personal information (high/medium) as independent variables in our model as well as the following covariates: general attitude toward advertising, privacy sensitivity, and perceived utility of the website. Because there were no interactions between the independent variables, our main analyses included overall multivariate results and effect sizes, univariate effects and etasquared values, and post hoc analyses through the Brown–Forsythe test.
The effect of informing consumers about advertising relevance was not significant after adjusting for the effect of privacy concern, utility of website, and general attitude toward advertising (F =.077, p =.926, η2 =.000). Accordingly, there were no significant mean differences regarding the acceptance of targeting (AcceptanceRelevance =
2.884 vs. AcceptanceNo-Relevance = 3.036; Brown–Forsythe: F =.649, p =.421) or the perceived intrusiveness of the advertisements on the website (IntrusivenessRelevance =
4.141 vs. IntrusivenessNo-Relevance = 4.101; F =.064, p =.801) between respondents who had been told that targeting would make their advertisements more interesting to them and those who were told that advertisers would like to reach their target group more efficiently. Therefore, H2a and H2b were not supported.
Regarding the mechanisms of appealing to reciprocity, the omnibus test revealed a significant effect after adjusting for the effect of the control variables (F = 9.919, p.0001, η2 =.048). Because reciprocity was a significant factor, we conducted further analyses to examine its effects on the two dependent variables. Follow-up analyses of covariance indicated that reciprocity had a significant effect on the acceptance of targeting operationalized as a provision of an opt-in (F = 12.068, p =.001, η2 =.030) and the perceived intrusiveness of the advertisements shown on the website (F = 8.327, p =.004, η2 =.021). A post hoc comparison of mean differences between groups suggested that respondents exposed to the reciprocity mechanism had a higher acceptance of targeting (AcceptanceReciprocity = 3.302 vs. AcceptanceNo-Reciprocity = 2.619; F = 13.548, p =.000) and perceived the advertisements on the website significantly less intrusive (IntrusivenessReciprocity = 3.870 vs. IntrusivenessNo-Reciprocity = 4.371; F = 10.035, p =.002) than those not exposed to the reciprocity mechanism.
Therefore, H3a and H3b were supported. H4 was not supported because there was no significant interaction between the reciprocity and relevance mechanisms (F =.424, p =.655, η2 =.002).
Main effects of the level of control offered to respondents on the two independent variables were significant after controlling for the covariates (F = 3.268, p =.039, η2 =.016). Univariate analyses indicated that the level of control had a significant effect on the acceptance of targeting (F = 5.515, p =.019, η2 =.014), which is reflected in the respective mean differences (AcceptanceHigh-Control = 3.157 vs. AcceptanceModerateControl = 2.764; Brown–Forsythe: F = 4.386, p =.037). Therefore, H5a was supported.
However, the amount of control had no significant effect on the perceived intrusiveness of the advertisements (F = 1.164, p =.281, η2 =.003), so we did not find any significant mean differences (IntrusivenessHigh-Control = 4.038 vs. IntrusivenessModerateControl = 4.203; Brown–Forsythe: F = 1.070, p =.302). Thus, H5b was not supported.
The core result of Study 1 is that the norm of reciprocity can guide people’s behavioral intentions in the context of information privacy and targeted advertising. If a website offering free content makes that norm salient, consumers are significantly more willing to accept targeting. To validate this finding, we conducted a second study in a real-world setting, which enabled us to collect behavioral data.
4.3. Study 2 Design We conducted a between-subjects field experiment in cooperation with a large advertising network. On two German websites, a renowned news website and a query community, we ran a survey that appeared similar to a typical predictive targeting survey. In May 2010, approximately 120,000 visitors of the two websites were invited to participate in the alleged predictive targeting surveys through a small flash layer, with a teaser text appearing when they entered the website. Our manipulation involved showing surfers two teasers, one focusing on relevance and one focusing on reciprocity. Websites running behavioral targeting surveys typically employ a teaser that emphasizes advertising relevance. Thus, our experimental conditions involved the current industry practice (relevance) and an innovative teaser we developed for the study (reciprocity). The teaser texts are available upon request. The manipulated flash layers were displayed to a predefined number of distinct surfers (no repeat visits), with group size being relatively equal within each website (news website: nScenario1 = 19,566, nScenario2 = 19,721; query community: nScenario1 = 40,114, nScenario2 = 39,900). As is common in predictive targeting surveys, we asked respondents for information regarding their interests in specific products, shopping habits, media usage, and demographic information (e.g., gender, age, profession, and household size), but no personally identifiable information.
The dependent variable was acceptance of targeting, which we operationalized as provision of information for targeting purposes. The final number of profiles received is a funnel consisting of (1) the number of surfers who saw the flash layer and clicked on it and (2) the number of surfers who then completed the survey. Therefore, the target variables of Study 2 are the click rate on the flash layer and the response rate, defined as percentage of surfers who completed the full survey after clicking on the flash layer.
4.4. Study 2 Results 4.4.1. Click rates.
On both websites, click rates in Scenario 2 (reciprocity) were substantially higher than those in Scenario 1 (relevance). On the news website, the average click rate moved from.88% using the traditional relevance teaser to 2.1% when we employed the reciprocity mechanism. In the query community, the click rate increased from an average of.46% to.83%. To test whether the differences in the click rates were significant, we performed a chi-square test. The chi-square test yielded a significant association between the text on the flash layer and whether a surfer clicked on the flash layer (news website: χ2(1) = 99.524, p =.000; query community: χ2(1) = 43.526, p =.000). The odds ratio shows that surfers exposed to the reciprocity mechanism were 2.4 (news website) and 1.8 (query community) times more likely to participate in the predictive targeting survey than those who had seen the traditional teaser. Thus, these findings provide strong empirical evidence that an appeal to reciprocity positively influences acceptance of targeting, in support of H3a.
4.4.2. Response rate.
For a higher click rate to result in more profiles, it is important that this effect is not offset by a potential decline in response rates. A comparison of the response rates of the respective scenarios shows that surfers who were exposed to the reciprocity teaser instead of the relevance teaser completed the survey significantly more often after clicking on the flash layer (news website: χ2(1) = 21.393, p =.000; query community: χ2(1) = 16.489, p =.000). On the news website, the response rate rose from 19.8% to 39.8%. The odds ratio shows that surfers who clicked on a flash layer containing a reciprocity primer were 2.7 times more likely to complete the survey than those who clicked on a traditionally worded flash layer. For the query community, the response rate rose from 4.3% to 16.6%, implying an odds ratio of 4.4.
5.1. Theoretical Implications This article examines privacy concerns in the context of targeted online advertising from a managerial perspective by building on findings from different academic disciplines, in particular, marketing, social psychology, and information systems research.
Our research confirms that, in general, consumer privacy concerns have a negative effect not only on the acceptance of targeting but also on consumers’ attitude toward targeted advertisements because they perceive them as more intrusive than regular advertisements. In their study on targeting and obtrusiveness Goldfarb and Tucker (2011a) suggest that privacy concerns negatively affect advertising effectiveness.
Whereas their data do not enable validation by measuring the related latent constructs, our research systematically analyzes the underlying cognitive processes and develops tangible mechanisms to alleviate the challenges entailed by privacy concerns.
Specifically, we contribute to marketing research in the following ways. First, we find that increasing procedural justice by allowing consumers to view, edit, and delete their information stored on websites results in greater targeting acceptance. As such, our findings confirm Culnan and Armstrong’s (1999) work on the role of procedural justice in addressing information privacy concerns. However, we were unable to significantly decrease the perceived intrusiveness of targeted advertisements through this mechanism. This might be because allowing consumers to view and edit their information leads to a more intense elaboration of potential risks and creates a high level of cognitive disruption, a proven source of intrusiveness (McCoy et al., 2008).