In this chapter, a research process is presented which synergizes well-acknowledged procedures for scale development, as well as a route to apply the new eWOM trust scale (eWT-S) to a current market problem. A review of relevant literature grants the insight that various scholars in the field of social sciences emphasize the issue of how to develop a quantitative, multi-item measure and how to validate it (e.g., Anderson & Gerbing, 1988; Bagozzi, 1998; DeVellis, 1991; Edwards, 2001; Netemeyer et al., 2003; Nunnally & Berstein, 1994; Straub et al., 2004). For instance, one of the most-cited and influential authors in the field of attitude measurement is Churchill (1979), who proposes a sequence of eight steps that shall be followed when developing measures for marketing constructs: (1) specify the domain of the construct; (2) generate a sample of items; (3) collect empirical data; (4) purify the measure; (5) collect new data; (6) verify reliability; (7) assess validity; and (8) develop norms. Other scholars have chosen to advocated slightly different steps and procedures (e.g., DeVellis, 2012; Hildebrandt & Temme, 2006; MacKenzie et al., 2011). However, the majority of these contributions typically share a common set of guidelines for the development process and analysis procedures. Figure 4 shows a diagram of the conducted research process, which begins with the conceptualization of the construct and culminates with the scale’s application as a segmentation tool of online information recipients. Due to this research’s various targets, a multi-stage/multi- study research design has been employed to test the scale’s psychometric properties. Specifically, the research has been executed in four sequential stages: (1) the identification stage; (2) the reliability stage; (3) the validity stage; and (4) the application stage (details are provided bellow). The methods and procedures used closely follow those described in the current and established psychometric literature (e.g., Bagozzi et al., 1991; Cortina, 1993; DeVellis, 2012; Robinson et al., 1991; Soh, 2007) and especially mirror the general recommendations put forward by Netemeyer et al. (2003), as well as the more recent work of MacKenzie et al. (2011). Both furnish the academic community with valuable frameworks and suggestions for constructing valid self-report measures of latent social-psychological
© Springer Fachmedien Wiesbaden 2017
W. Weitzl, Measuring Electronic Word-of-Mouth Effectiveness,
constructs. As suggested by these scholars, the developed eWOM trust scale was subject to rigorous reliability and validity testing using multiple consumer samples from various cultures.
The stages can be outlined as follows. The identification (or conceptualization) stage includes two fundamental steps. In the first step, the importance of theory in scale development was recognized by a thorough review of the interdisciplinary trust research literature. This enabled the proposal of a literature-based theory in which the construct of eWOM trust is embedded. Here, several key issues have been addressed: (i) the critical role of a clear construct definition and an explicit specification of the content domain; and (ii) based on these insights, an a-priori determination of the construct’s structure and dimensionality has been proposed. The findings of this step were presented in Chapters 2 and 3. The second step in the identification stage involved generating and judging a pool of observable items, from which scales of the subdimensions were derived. In accordance with Netemeyer et al. (2003), during this step (i) the theoretical assumptions about the items (e.g., domain sampling) have been considered; (ii) a large set of potential items, theorized to capture the domain of the construct as indicators of trust in eWOM, was created based on a review of prior literature, dictionaries and thesauri as well as interviews with consumers and experts; (iii) the response format was determined; (iv) marketing experts and members of the relevant population judged the items with a focus on both content and face validity; and (v) the items for the subsequent stage were finally identified - all possessing translation validity.
In the following reliability stage, the construct’s composition has been empirically tested with a reasonable number of respondents. For this purpose a study among university students was implemented and the received data were split in half for subsequent analysis. Here, exploratory factor analysis (EFA) was initially applied to the first half of the data in order to investigate the construct’s structure and to detect a smaller set of items which seem to be appropriate in psychometric terms to determine the higher-order construct (i.e., scale purification). A series of internal consistency measures was applied to evaluate the reliability of the identified trust elements. Afterwards, in order to validate the distinct components of eWOM trust, a confirmatory factor analysis (CFA) was applied to the same sample. The results were cross- validated with the data from the holdout sample.
In respect to Churchill’s (1979) and other scholars’ recommendations, new data from a representative sample of adult online consumers was collected during the validity stage. The application of CFA and other procedures enabled the confirmation of the construct’s order, dimensionality and the measure’s reliability. Additionally, various types of validity (e.g., concurrent) were assessed by using this study’s data. For instance, to test for convergent and discriminant validity, the relationships between eWOM trust and review credibility (RCred), as well as attitude towards reviews in general (RAtt), were investigated. During this stage, various other forms of validity (e.g. predictive) were evaluated by collecting additional data from new samples in a series of complementary studies. Here, for instance, in order to test for nomological validity, the earlier theorized relationships between the construct and its antecedents, correlates, and consequences were empirically investigated. Further, these studies also enabled this research to cross-validate the findings of the main study by using pooled data as well as the assessment of the scale’s generalizability to different contexts. The development of norms and the application of g-theory concluded the validity stage and the scale development process. As the aim of this research was also to demonstrate the usefulness of the new measure, a segmentation study was conducted, introducing a new typology of eW OM/Ad trusters by means of cluster analysis (application stage). In the remainder of this chapter, each of the major stages of the research process is discussed in greater detail. This discussion includes a presentation of the studies which were designed to answer the research questions and hypotheses proposed in Chapter 3.