Consumer attitudes towards Internet advertising and motives for Internet use were gathered from participants. Participants were asked to evaluate an automotive website text and answer questions about their attitude toward the brand, attitude toward the website, and behavioral intention based on the website text. Chapter three discusses the study design, selection of subjects, instruments used to collect data, procedures used to collected data, as well as the data analysis.
The researcher has a post positivist worldview according to Creswell (2009). The researcher wants to evaluate variables to determine outcomes, using numerical, statistical analysis. The goal of the study was to numerically evaluate attitudes and opinions of a study population by studying a sample of the population using a survey instrument. Creswell (2009) describes a research design based on the quantitative approach with a post positivist worldview. This was the structure for this proposed study.
In this scenario, the researcher tests a theory by specifying narrow hypotheses and the collection of data to support or refute the hypotheses. An experimental design is used in which attitudes are accessed both before and after an experimental treatment. The data are collected on an instrument that measures attitudes, and the information is analyzed using statistical procedures and hypothesis testing. (Creswell, 2009, p. 16)
A survey design best met the needs of this quantitative study. This allowed a numerical analysis of the relationship among variables and outcomes. Using an online, self-administered, cross-sectional survey, the data was collected quickly and efficiently from an online panel.
Hong (2006) conducted a somewhat similar study on the automotive industry using a student sample. Students were used in the sample because students are considered homogenous (Calder, Phillips, & Tybout, 1981 as cited in Hong, 2006) and early adopters of innovation (Gallagher, Parsons, & Foster, 2001 as cited in Hong, 2006). However, Hong (2006) also cites a study by Bracket and Carr (2001) who found that college students have more negative and less positive attitudes towards advertising than other people; this finding could have affected the results in Hong’s study. There might be a fundamental problem in using college students as a sample with the automotive industry, as college students might not be as likely to purchase cars because of lack of funds. A product more in line with what students might purchase might have been more appropriate with the student survey demographic.
This study used a non-student focused sample and combined research from multiple studies such as Ducoffe (1997) and Hong (2006). The research will build on previous automobile research centered on message design. For this study, the survey was based online instead of in a computer lab at a university as in Hong (2006) or an intercept survey as in Ducoffe (1997). This allowed a broader range of participants to complete the study with much more time flexibility than a rigid appointment time that might be required if using a lab or the geographical limitations of an intercept survey.
A sample was drawn from a research panel that several marketing and research websites offer. Online panels are an effective way of conducting research (Aaker, Kumar, & Day, 2007) by using existing groups of people who have already agreed to participate in surveys. Harris Interactive, Synovate, and many others offer these services. The online panel company sends an e-mail message to those panelists who qualify based on sample specifications given by the researcher (Malhotra, 2010) and the panelist is then linked to the online survey.
A benefit is that the data was collected quickly from the panel as compared to other non-Internet research formats. In addition, online panels can reach a broader audience over other research methods in a more cost-effective manner (Aaker et al., 2007). Online panels also provide “high-quality data acquired from willing, interested, and motivated participants” (Aaker et al., 2007, p. 172).
Selection of Subjects
The goal of the study was to test website message types on a non-student sample. This sample was different in terms of technology usage, income, and car purchase patterns than that of the college student sample in the Hong (2006) study. The sample contained high-earning households in the United States. Household income has shown a positive correlation with online purchase amounts (Hannah & Lybecker, 2010). If a consumer has a higher income, higher education, and owns a home, there was a higher intention to adopt an online retailer (Chen & He, 2003).
Census data was used to help operationalize high-earning households. The Census divides incomes into quintiles; for this study the top quintile was researched. To be in the fifth (top) quintile for income, household income would need to be above $100,065, that is the upper limit of the fourth quintile (U.S. Census, 2010). The Census reported that there are 118,682,000 households. Twenty percent of this population equated to 23,736,400 households with income reported above $100,065.
A sample size calculator from Creative Research Systems was used to calculate the sample size (Creative Research Systems, 2010). There was a financial limitation to this study. Each completed survey has a fixed cost with Survey Monkey. This cost forced the sample to be less than the ideal 384 participants that would provide a confidence interval of 5 at a 95% confidence level. Given a population of over 20 million people, a sample size of 160 participants at a 95% confidence level, the study would have a confidence interval of 7.75. The 160 participant threshold was the minimum study population desired for this study.
The online survey company Survey Monkey was used in the data collection procedure. With Survey Monkey, there are a variety of safeguards to provide for an accurate sample. Technology was used to ensure that each survey response was unique by tracking the computer that the survey was taken on as well as filtering technology that analyzed the responses for irregularities (MarketTools, 2008). Data validity was also tested by analyzing speed of responses to eliminate suspect participants that might be speeding through the survey or selecting the same answer on all questions (2008). The service also verifies the validity of panelists’ information to ensure proper demographics are selected by participants (2008).
The sample was comprised of members from the upper quintile of household income in the United States and excluded anyone less than 18 years old. Survey Monkey Audience provided the sample research panel. The Survey Monkey Audience contains over 3,000,000 participants (Survey Monkey, 2012a). The panel is composed of two groups ZoomPanel (a company Survey Monkey purchased) that has two million members in the United States, Australia, Canada, and the United Kingdom who opted in to take surveys for rewards and Survey Monkey Contribute that has one million members based in the United States who opted in to take surveys in exchange for donations to charity (Survey Monkey, 2012a). At the time of the survey, Survey Monkey Audience offers a $.50 donation to charity for a completed survey and a chance to win $100 for participants (Survey Monkey, 2012b). Twenty percent of participants in Survey Monkey Audience qualified for the study, based on income.
The sample was randomly selected from the survey provider’s research database of participants. The only criteria given to the provider is that all participants should live in the United States, be over 18 years old, and have a household income of $100,000 or more. Before participants began the survey, they were presented with a consent letter. Each participant needed to click to the next page and by doing so the participant agreed to provide consent to participate in the study before the survey appeared. Using the provider’s interface, the questions and website mockups were loaded into the system. The provider then solicited the sample to complete the surveys until the threshold of 176 completed surveys was reached. This equates to the 160 needed for the study plus an overage of 10%.
Survey Monkey has many validity tests, as identified in the previous paragraph, to evaluate the responses to ensure that the survey is completed accurately by participants. By having 176 completed surveys, any surveys that needed to be eliminated on top of the ones that Survey Monkey eliminated, there will still be at least 160 useable surveys. The sample was not stratified. The data was collected in October 2012 within a seven-day period using the provider’s sample. There was a technical error that allowed participants that had not met the income threshold to participate for the first few days of data collection. Those cases were not used in the study. Survey Monkey Audience granted additional participants to compensate for the cases that needed to be removed. Once data collection was completed and the cases that did not meet the income threshold were removed, 198 cases remained. Given a population of over 20 million people, a sample size of 198 participants at a 95% confidence level, the study has a confidence interval of 6.96. This is slightly higher than the originally estimated sample size of 160 with a confidence interval of 7.75.
There could be some sampling bias, based on the database the provider used to comprise the sample. Since the survey was meant to gather data from people who are online, there was no bias against those who do not have Internet access. Additionally, web surveys are particularly effective for researching consumers who shop on the web (Alreck & Settle, 2004) which was the target sample for this study. The data is not meant to be generalizable to other populations outside of the sample population.
The survey was divided into roughly three sections. The first section asked for Internet usage and demographic information. The second section evaluated participants’ general attitudes about Internet advertising and their motives for using the Internet. The final section asked participants to read a homepage text for two websites and respond to questions about their perceived attitudes of the website. Perceived attitudes of the website were measured from three variables including (a) attitude toward the site, (b) attitude toward the brand, and (c) behavioral intention.
Internet Advertising Attitude
The study first measured consumer attitudes towards Internet advertising in general. From the review of literature, four dimensions of beliefs towards Internet advertising were identified: perceived informativeness, perceived entertainment, perceived annoyance, and perceived credibility. Perceived informativeness, perceived entertainment, and perceived credibility relate to positive advertising attitudes (Ducoffe, 1996). Perceived annoyance relates to negative advertising attitudes (Ducoffe, 1996). These dimensions were established in Ducoffe’s (1996) study. The same dimensions have been used on several other studies (Yuan, 2006; Hausman & Siekpe, 2009). Respondents were asked to indicate their level of agreement to each item on a 5-point Likert-type scale ranging from strongly disagree to strongly agree. There were six questions for perceived informativeness, four for perceived entertainment, four for perceived annoyance, and six for perceived credibility. Through collecting this data, two groups were created, one with positive attitudes towards Internet advertising and one with less positive attitudes.
There were a variety of uses and gratifications sought from the Internet. Participants have a variety of motives that drive their Internet activity. Several studies have established motives for Internet motives (Papacharissi & Rubin, 2000; Rodgers & Thorson, 2000). Yuan (2006) found that motives were significant in predicting beliefs. Respondents who reported higher levels of using the Internet to gather commercial information were more likely to make purchases (Yuan, 2006). Two nominal groups were formed to divide participants into high and low groups, based on whether they gather information related to commercial products and services and shopping/e-commerce. A scale ranging from most of the time, to never, was used in Yuan’s (2006) study to measure the Internet motives of each user. The possible motives from the Yuan (2006) study were also used in this study and include:
• Gathering information related to commercial products and services.
• Gathering non-commercial information.
• Communication with others.
• Entertainment/killing time.
Attitude toward the Site
Attitude toward a website can predict a consumer’s predisposition to respond favorably or unfavorably (Chen & Wells, 1999). Attitude toward the website was measured using a similar scale developed by Chen and Wells (1999). Studies such as Cho (2003) and Hausman and Siekpe (2009) have also used similar scales. The 5-point Likert-type scale questions ranged from strongly disagree (1) to strongly agree (5). There are five questions from the studies mentioned above to evaluate attitude toward the site. The questions were tallied to create a cumulative score of attitude toward the site.
Attitude toward the Brand
Brand attitude in Chuang, Tsai, Cheng, and Sun (2009) was measured by using six 5-point Likert-type scales with larger numbers indicating more positive attitudes towards the brand. The scores were added to create a cumulative (interval) score for attitude toward the brand. The usage in the 2009 study was for an e-play cell phone. Previously, similar question sets have been used in other studies such as Hong (1999) and MacKenzie and Lutz (1989). The question is simply, “My attitude toward the brand based on the presented text was:” with response options being 1 = bad to good, 2 = unfavorable to favorable, 3 = negative to positive, 4 = unpleasant to pleasant, 5 = not entertaining to entertaining, and 6 = not useful to useful.
Behavioral intention was measured by a scale similar to Chen and He (2003) and Raney, Arpan, Pashupati, and Brill (2003). Participants should be more likely to return to a site that is found to be useful (Raney, Arpan, Pashupati, & Brill, 2003). In the Raney et al., (2003) study, four automotive websites were evaluated for interactivity by participants and the behavioral intention of returning to the website was gathered. The first four questions are Likert-type questions from the Raney et al. study (2003) aimed at measuring the likelihood of someone to return to the website. The likelihood to return to the site is operationalized as the likelihood of participants’ or participants’ friends to visit the site. The last two interval questions are based on questions from Chen and He (2003) and Hausman and Siekpe (2009) and were designed to measure future website behavior for future information searches. The questions were added for cumulative score in analysis.
Demographic information was collected including the participant’s age, ethnicity, and education. Further, two questions aimed at evaluating participants’ skill sets and usage of the Internet are included. One interval question solicited the amount of time the respondent has used the Internet. The other question is a self-reported skill level for Internet use. An additional nominal question asked for the primary search engine used when searching for information. Two nominal questions were aimed at evaluating where Internet access used for shopping takes place and the type of device used.
The questions in the survey were all based on scales that have previously been used in research. This research built upon an existing study and body of literature. This survey has combined several instruments to compose a new study. Validity and reliably was further examined during analysis for irregularities. Means, frequencies, and standard deviations were calculated and examined for missing cases and possible data entry errors.
Reliability and Validity
The key question areas in the survey were based from previous studies where the questions were used and it is assumed that the questions measure the content the questions were originally designed. Many of the questions were combined into constructs that are also assumed to be valid as used in previous research. The key questions in this study have been used in a variety of other studies that show similar results, so the predictive validity is assumed to be high. Internal validity was also tested for scale questions with Cronbach’s alpha. Cronbach’s alpha is a measure of internal coefficients resulting from different ways of splitting the scale items (Malhotra, 2010). The coefficient ranges from 0 to 1, and a value of 0.6 or less generally indicates unsatisfactory internal validity (Malhotra, 2010).
The texts follow the guidelines set by Ennis and Zanna (2000) in the literature review. A website message was created for both instrumental and symbolic message strategies for an automobile manufacturer’s website. Instrumental message strategies provide messages about the utilitarian functions, features, and benefits of the product and symbol messages appeal to self-image, self-expression, and social status (Ennis & Zanna, 2000). The websites of Ford, Chrysler, GM, and Toyota were visited and it was noted that the use of text content is minimal on the home page of each company’s website. In the Hong (1996) study the stimulus texts on the homepages were three long sentences and participants were limited to only viewing the homepage. In keeping with the brevity of the actual car manufacturer websites and the Hong (1996) study, the two stimulus messages were designed.
The stimulus texts are inserted into a mocked-up website. Toyota.com was used as the model for developing the stimulus website design. All of Toyota’s content, images, and other elements were removed leaving primarily the upper and lower buttons and links. The symbolic and instrumental texts were inserted with identical fonts, sizes, and colors into the website page template.
Participants were recruited into a survey panel by a market research company. Participants were asked questions about their attitudes towards Internet advertising using the scales for perceived informativeness, perceived entertainment, perceived annoyance, and perceived credibility. They were also asked questions about their motives for Internet use. As in Ennis and Zanna (1993), instructions let participants know they were going to access written texts that might be used in future advertisements for an automobile. Half of the participants were exposed to an instrumental text and half to a symbolic text on an automotive website. The design and layout was identical for both websites. The website was displayed as an image and participants were not able to access any of the links. After reviewing the website, participants were asked for their attitude toward the website, attitude toward the brand, and behavioral intention, based on the website text that was presented.
Data Processing and Analysis
Survey Monkey was used to collect the data. The survey tool generated an export feed of data that is SPSS compatible. The data was examined to look for potential errors, missing cases, and prepared for analysis in SPSS. Means, frequencies, and standard deviations were calculated for all questions and reported in the results chapter. The attitude toward Internet advertising was calculated by adding the composite scores for perceived informativeness, perceived irritation, perceived entertainment, and perceived credibility. This created two nominal groups including one with positive Internet advertising attitudes and one with less positive Internet advertising attitudes. In analysis, symbolic and instrumental website text groups were analyzed to see what group had more positive attitudes toward the site, positive attitude toward the brand, and the strongest behavioral intention. By using a two-way multivariate analysis of variance, positive and less positive Internet advertising attitude groups were examined for what website text led to a more positive response, as measured by attitudes toward the site, attitude toward the brand, and behavioral intention. Analysis of variance is a way to examine the differences between groups of responses (Malhotra, 2010). A two-way multivariate analysis of variance comparing high and low commercial Internet usage examined what website text led to a more positive response, as measured by attitudes toward the site, attitude toward the brand, and behavioral intention.
The key analysis was a three-way multivariate analysis of variance with positive and negative views of Internet advertising, and instrumental and symbolic attitude toward the site, attitude toward the brand, and high and low commercial Internet activity. By using a three-way multivariate analysis of variance, all of the variables were simultaneously compared for in-depth analysis and comparison. Hong’s (2006) study that looked at instrumental and symbolic texts by commercial Internet activity and attention-to-social-comparison-information scale score used a similar analysis.