In the light of the theoretical framework and the background information, which are introduced previously, this chapter of the study focuses on explaining how the empirical study was conducted. In addition, the following sections of this chapter include detailed information about, the choice of research method, the research design, the data collection, and the manipulation of the data and analysis.
Research Philosophy and Approach
As mentioned earlier, the aim of this study is to come up with an extended TAM model that explains the factors affecting the adoption of streaming music services, and the change in consumer behavior.Saunders, Lewis, and Thornhill (2007, s. 101) states that the research philosophy and approach adopted have a great influence on research strategy choice and on research method. This study espouses a positivist position to the development of knowledge within epistemological thinking of research philosophy. The positivist positioning focuses on working with an observable social reality to come up with a product that has law-like generalizations (Saunders, Lewis, & Thornhill, 2007, s. 103).In addition, this study is based on deductive research approach that is more applicable to positivism, than inductive approach (Saunders, Lewis, & Thornhill, 2007, s. 117). A deductive research approach is consisting of deduction of the hypothesis from the theory base, expression of the hypotheses to propose a relationship between two specific concepts or variables, test of the hypotheses with collected data, examination the specific outcome of the inquiry, and modification of the theory in the light of the findings (Saunders, Lewis, & Thornhill, 2007, s. 117). The TAM theory formulation studies of Davis et al. (1989) together with the other extension studies by Amoroso & Guo (2006), Malhotra & Galletta (1999), and Hiramatsu et al. (2009) concerning the TAM, provided the preliminary information to construct the theoretical framework of the study. Based on the findings of these previous studies an extended TAM was modeled, and the hypotheses, which are previously given in the research model and research hypotheses part, were formulated in order to test the functionality of the new model.
Research Strategy and Method Choice
Saunders et al. (2007, s. 134) mention that research problems of a study may have more than one purpose and can be both explanatory and descriptive. In terms of the explanatory purpose, this study aims to examine streaming music service adoption via explaining the relationships between different components of proposed TAM by subjecting the data to statistical tests. In addition, the study also aims to present an accurate profile of the streaming music service users, concerning the descriptive purpose.Survey strategy is highly associated with positivism research philosophy and deductive research approach. Moreover, when compared to the other research strategies, surveys allow researchers to collect a large amount of data form a sizable population in a highly economical way (Saunders, Lewis, & Thornhill, 2007, s. 138). Therefore, this study adopts the survey research strategy and uses survey to analyze the proposed theoretical framework. On the other hand, the survey strategy can use both quantitative and qualitative data as input. From these two techniques, the quantitative method was chosen for this study. Saunders et al (2007, s. 145) states, quantitative method includes data collection techniques and analyses (such as questionnaires, statistics), which generates and uses numerical data. Moreover, the data collected can be used to explain relationships between variables while producing the models of these relationships, since it is standardized and allowing the easy comparison of variables.This study, which adopted survey strategy, uses a single method. The choice of single method is suitable with the positivism philosophy, and the deductive approach of the study. As mentioned previously, deductive research approach and positivism philosophy aims to work with an existing theory to develop test hypothesis, and to come up law-like generalizations. At this point, the quantitative data collection techniques enable easy comparison of variables, and standardization. Furthermore, analyzation of the collected data by using statistics is also possible with quantitative method. Therefore, the choice of single method and using quantitative methods for data collection and the analysis is essential for testing the proposed TAM for streaming music services.
Data Collection Technique
According to Saunders et al (2007, s. 355), the questionnaire is a widely used data collection technique in survey research strategy, which best suits with descriptive and explanatory studies. In addition, it was stated that questionnaires are effective in collection of responses forms a large sample, since each respondent is asked to the same set of, standardized, close-ended questions. Therefore, in the light of the existing knowledge, questionnaire is chosen as the data collection method of this study, and a questionnaire is designed based on the theoretical framework.The questionnaire designed for this study consists of close-ended, standardized questions that target a specific target group, and is designed as a self-administrated questionnaire that completed by the respondents. After the formulization of the questions, an electronic internet-mediated questionnaire form was created by the help of the form sheets offered by Google documents. In addition, since the web link of the form file generated in Google documents was long and complex, the web link was directed to the http://spotifysurvey.blogspot.com/ domain address. This made the questionnaire link that was sent to respondents simpler and more attractive. Besides, the use of electronic forms of Google enabled the automation of capture and input of the answers of the respondents. After the completion of the data collection, the final dataset was exported as a Microsoft Excel file, which is a ready input file format for the statistical analysis programs. The automated capture and input of the data catered the extra time needed for the data collection period by eliminating the time required for manual entry of the data.
Sampling is essential when it is impracticable to collect data from the entire population, due to time, money and access restrictions (Saunders, Lewis, & Thornhill, 2007, s. 204). The survey strategy of this study aims to examine the adoption of streaming music services via studying Spotify, and Spotify has an approximate user population of seven million. Therefore, sampling was also important in this study because it was not possible to census the entire population due to time constraint.There are two main types of sampling technique, which are probability and non-probability sampling. According to Saunders et al (2007, s. 207), probability sampling provides equal chance of each case selected from the population is known. That, in result, makes it possible to answer each research question statistically through the characteristics of the population defined by the sample. On the other hand, non-probability sampling allows identifying the members of a group, when it is not possible to acquire the details of the all cases to store on the computer.The probability sampling technique that was used for this study requires the identification of the sampling frame as a first step. The research questions and theoretical model extension objective of this study was concerned with members of Spotify. Therefore, the sampling frame should have been the complete membership list of Spotify. In the light of this information, the members of Spotify Facebook group, and the Sharemyplaylist.com, which hosts Spotify users and gives access to complete membership list, was defined as the sampling frame. All of the cases listed in this sampling frame are Spotify users and can provide data to answer the research questions and hypotheses.In order to lower the likely error in generalizing the population, the sampling size should be as big as possible, which also in result increase the accuracy of the findings while lowering the money required for the collection, and the analyzation of the data. The choice of sample size is determined by the total size population, the types of analyses to be undertaken, the margin of error to be tolerated, and the confidence level needed (Saunders, Lewis, & Thornhill, 2007, s. 210). As of April 2010, the total size of Spotify users was around seven million (IFPI, 2010) and the population size of sampling frame was 340.000. The sampling size for the population was calculated as 384 at a level of certainty of 95 percent, and the margin of error of 5 percent. However, as Saunders et al (2007, s. 212) stated the 100 percent response rate was unlikely so actual sampling size required to suffice the margin of error was calculated as 1280 according to the expected 30 percent response rate.Saunders et al (2007, s. 215) emphasizes, once the suitable sampling frame is chosen and the actual sample size required is established, the most appropriate sampling technique needs to be selected. In this study, a combination of simple random and self-selection sampling methods was used to collect the data. The questionnaire link was sent to respondents, who were chosen by using random numbers, from the member lists of the previously mentioned web-community pages. Although, the complete member list was accessible, it was not possible to create a case database in order to assign random numbers to each case. Thus, the random number technique was used by assigning the random numbers manually each time before selecting the respondents from member list page. In addition to the simple random sampling, as a part of the non-probability sampling the self-selection sampling technique was also used. The questionnaire was advertised by posting the questionnaire website link as a comment to the wall posts of Spotify in Facebook. Since wall posts of Spotify only targets the members of the fan group in Facebook, only the cases within the sampling frame were successfully covered. Moreover, in Sharemyplaylist.com community webpage, the need for cases was advertised and members were asked to take part in filling the questionnaire.
Concerning the questionnaire design, Saunders et al (2007, s. 356) emphasizes that the design of a questionnaire directly effects the response rates, the validity and the reliability of the data collected. Furthermore, in order to maximize the efficiency, the questions should be carefully encoded, and the questionnaire must be clearly introduced, designed, pilot tested, and administrated.The validity of the questionnaire can be assessed under three groups, which are content validity, criterion-related validity and construct validity. According to Saunders et al (2007, s. 366), content validity refers to the adequate coverage of the research questions, hypothesis by the questionnaire, and criterion-related validity. The content validity of this study was made through intense identification of the research through the literature reviewed, and a test-group of potential respondents was interviewed to discuss whether the questions included are essential or not. Besides, the questions are designed to account TAM components; Table 3-1 presents the link between components, question numbers, and the research questions. The complete questionnaire can be seen in appendix 1.When designing the questionnaire, the measurement scales and the questions regard to base TAM were adapted from other questionnaires designed for the previous TAM studies. In addition, new questions were developed for the new attachments of the proposed model. The questionnaire designed includes a combination of open and closed questions. The only open-ended question included is the comment box added to question 13, which aims to get detailed answer about users‟ intention to buying digital music copies from Spotify, and to understand what is uppermost in their minds towards digital music purchase. Saunders et al (2007, s. 369) states that, although the open-ended questions are useful to capture deeper knowledge, they are extremely time consuming to code as a part of a largescale questionnaire and it is advised to keep their use to a minimum. Therefore, rests of the questions in the questionnaire were worded as closed-ended. The closed-ended questions were used because these questions require minimal writing, as they are quicker and easier to answer (Saunders, Lewis, & Thornhill, 2007). Saunders et al (2007, s. 368) lists six types of closed-ended, questions, list, category, ranking, rating, quantity, and grid. This study has used grid question type with Likert-style rating on a five point rating scale, for the questions connected to TAM components. In addition to the grid questions, list (questions 1, 7, 13), and category (questions 2, 17, 18) question types were used.The questionnaire that was designed contains eighteen questions. The first question in the questionnaire aims to exclude responses besides the target group, and the following questions from three to sixteen includes include factor questions that try to capture different aspects based on the theoretical framework model. The last two questions covers the demographics, in order to know the age and sex of the respondents. In order to be able to test each component of the proposed TAM for streaming music service, the part of the questionnaire related to these components divided into seven parts, which consist of factors concerning perceived usefulness, attitude toward use, behavioral intention, social influence, ad-charge, flow, and account type.The second question of the questionnaire was concerning the account type component of the proposed model, and respondents were asked to choose either free or premium to define their membership method. In the question 3, respondents were asked to rate different statements that aim to capture the degree to which users believe that using a streaming would enhance their overall music listening experience, or habits. These statements were asking whether if using the streaming music service enhances, the size of the music catalogue they can access, the sound quality of the music that they listen, their accessibility to music and their music sharing activities. The respondents‟ attitude toward use of streaming music services was asked in the question 4. The questions 8 and 9 concentrated on measuring the behavioral intention to use streaming music service. Behavioral intention to use was captured in two different questions in order to differentiate between intentions to use arising from different account types. The question 8 measures intention to use Spotify in general, whereas the question 9 focuses on intention to use a premium Spotify account. The factors that aim to measure the level of social influences were asked in the questions 5 and 10. The nine factors in these two questions were also divided into three categories to capture Kelman‟s three processes of social influence (1958). These factors targeted to different dimensions of compliance, identification and internalization. In the question 11, respondents were asked to rate the given statements in order to be able to measure their level of attachment. The attachment level, in the proposed model, was defined in flow component. The respondents were asked whether Spotify is their music player, and if they switch to another medium when the tunes they want to listen are not available in Spotify. The question 12 was asked to measure the ad-charge component, and covered the effect of advertisement and charges for premium account on use. In addition to the questions targeting the components of TAM, in questions 14, 15 and 16 the respondents were also asked about different factors that aim to measure the change in their consumption habits after Spotify. These questions were designed to answer the research questions of the study, which contains factors that concern the change of music consumption behavior, and the effect of streaming music service use on music piracy. The question 6 included to the questionnaire in order to measure respondents‟ opinion about piracy and legal music service use. Besides, questions 7 and 13 were also included to measure the streaming music services users‟ willingness to pay for music ownership in general.
Pilot Test of the Questionnaire
According to Saunders et al (2007, s. 386) the questionnaire should be pilot tested before using it to collect data. As mentioned before they also state that pilot testing is a part of the questionnaire design, which improves the efficiency. The pilot testing enables to find out, the time required to complete the questionnaire, the biased & wrongly encoded questions, and the problems related to design.The first pilot test of the questionnaire was conducted with two persons. According to the feedbacks of the test participants, a revised version of the questionnaire was formed that included the corrections on grammar, and encoding problems in order to increase clarity.The questions, which were related to the technical details about internet connection speed and type, the compression quality of digital music and media compression types, and the medium preference of the users‟ to listen to music were deducted from the questionnaire according to the supervisor comments. These questions were deducted since they were not directly related to the research hypotheses and questions. This also lowered the time required for completing the questionnaire.The second pilot test was conducted online with fifteen persons. The link of the web page that hosts the questionnaire was sent to the test participants through Facebook. The participants were informed that it was a pilot testing, and were asked to report the grammatical mistakes that they captured together with their comments about the questions that caused blur while answering. The final version of the questionnaire was created after the corrections done according to the feedbacks captured from the second pilot testing.
Saunders et al (2007, s. 388) underlines that during the administration stage of the questionnaire, the goal of the researchers is to gain access to sample and to maximize the response rate. They summarize different techniques that help to maximize the response rate, when conducting a postal questionnaire. Some of these techniques were also adopted to use in the questionnaire of this study. Two Facebook and one Sharemyplaylist account were created and a new e-mail address was registered for the research project, in order to give a more official appearance. The profiles included the genuine logo and brief information about the purpose of the questionnaire. In the information section, it has been noted that the questionnaire was a part of an academic research that has been conducted in the Jönköping University. Before the start of the administration, the web link of the questionnaire created in Google forms was directed to the spotifysurvey.blogspot.com domain, and tested with different web browsers in order to be sure that the image being displayed is the same.The administration of the questionnaire was started on 10th of March 2010, and finalized on 18th of April 2010. As a part of the probability sampling, the direct link of the questionnaire web page was sent to 600 randomly selected respondents, 200 through Facebook and 400 through Sharemyplaylist.com. Since Facebook has strict rules against spamming, each account has a temporary limit in sending web links via personal messages in order to prevent spamming. Therefore, it was critical to send out a maximum of twenty request messages every day. The accounts were blocked by Facebook two times, and during that block stages it was not possible to send out any request messages for ten days in total. Sharemyplaylist.com portal was less sensitive in this sense, and with the help of the site admin, the request message could be freely sent out to the randomly selected respondents. In addition, non-probability sampling was also used since it was not possible to reach to the targeted 1280 actual sampling size with probability sampling due to the time constraints. The web page link of the questionnaire was posted as a comment to the wall post Spotify in the company Facebook group. Fifteen comments were posted during the administration period. Although it is not possible to identify how many group members viewed the comments, when the timestamp and comment times compared it is possible to say that each comment provided five to ten responses within the next 20 minutes of the comment time.As of 18th of April 2010, 246 responses were achieved. The achieved response number resulted in an increase to in the margin of error to 6.25 percent, and decreased the confidence level to 88.3 percent. The limitations of using Facebook to gain access to sample were not taken in consideration during the planning stage. These limitations resulted in major time consumption, and hindered the achievement of the targeted sample size of the study. On the other hand, use of non-probability sampling in Facebook came up more efficient than it was expected.
Methodological Limitations of the Study
The adopted method for conducting the internet-based questionnaire resulted in some structural changes because of the practical limitations. The second question of the questionnaire aims to collect categorical data to explain the account type component included in the proposed TAM. The research hypotheses that are attached to account type component assert that SI, U, D, and F can differ for users, according to their account type. Since premium and free user are using the same service with different features, the questionnaire needed to address extra questions to the respondents according to their account type chooses in the question 2. However, the form sheets offered by Google documents do not have an option to filter and address different question depending on an answer given to a question. There fore the designed questionnaire is not capable of capturing different the dimensions that derive from account type preference. However, it is still capable of explaining the differences on beliefs of these two groups.Spotify Facebook group is a fan page that is administrated by Spotify, and gives full access to the list of members of the group.ShareMyPlaylists.com is a resource for Spotify users to share and explore Spotify Playlists, which is owned and operated by UK Offer Media Limited (Sharemyplaylist).
1.1 The Music Industry and its Consumers in a Post-Napster
1.2 Evolution of Streaming Audio
1.3 Streaming Music Services
1.4 What is Spotify?
1.5 Literature Review about Streaming Music
2 Theoretical Perspective
2.1 The Theory of Reasoned Action
2.2 Technology Acceptance Model (TAM)
2.3 Problem Discussion
2.4 Purpose of the Study and the Research Model
3.1 Research Philosophy and Approach
3.2 Research Strategy and Method Choice
3.3 Data Collection Technique
3.4 Data Sampling
3.5 Questionnaire design
3.6 Pilot Test of the Questionnaire
3.7 Data Collection
3.8 Methodological Limitations of the Study
4 Empirical Study
4.1 Analysis of the Questionnaire Results
GET THE COMPLETE PROJECT
Changing Consumption Behavior of Net Generation and the Adoption of Streaming Music Services