Log in / Register
Home arrow Engineering arrow Measuring Electronic Word-of-Mouth Effectiveness: Developing and Applying the eWOM Trust Scale

Generating and Judging Items

Having conceptualized the dimensions of trust being best measured by reflective indicators, the next step was to generate a set of initial items that captures the domain of each dimension as specified. This research applied domain sampling and procedures of exploratory research that have been found productive in this respect (Churchill, 1979; MacKenzie et al., 2011; Netemeyer et al., 2003). Domain sampling suggests that a measure should be represented by a sample of items drawn from a large hypothetical domain of items (the sampling universe) which all mirror (in this case) the meaning of the diverse trust sub-dimensions (Netemeyer et al., 2003). Hence, in order to identify words and phrases that are associated with trust and associated constructs - in line with literature recommendations (e.g., Churchill, 1979; DeVellis, 2012; Haynes et al., 1995) - multiple source have been consulted. First, a systematic review of interdisciplinary literature in the areas of marketing, management, psychology, sociology, as well as information systems and communications research was conducted. Here, several items were derived from previous published sources on trust and trust-related attitudinal constructs (e.g., credibility, information usefulness/helpfulness, distrust) that were assumed to capture the domain of the construct, or at least parts of it. Also, this research paid attention to previous theoretical and empirical research on the focal construct (i.e., trust in online reviews and online reviewers), as well as other measures of the construct which already exist. Thus, items were generated from general and domain-specific literature. A review of various dictionaries and thesauri completed the research. Additionally, some items were freely constructed by the scale developer from the descriptions and definitions of the various dimensions outlined by academic research. Hence, this study used the often-recommended deductive approach to arrive at a comprehensive set of items.

This process resulted in an extensive list of 620 words and phrases that have been found to be frequently used to describe aspects of trust. Given the complex nature of the focal construct, this quantity of items appears adequate (Netemeyer, 2013). By means of a detailed review and rigorous item selection process performed by the author and a second marketing researcher, the extensive list of items was trimmed to a more manageable number, and the initial set of items for the individual sub-dimensions of eWOM trust was produced. Item elimination was based on a variety of a priori criteria regularly advocated in literature (e.g., DeVellis, 2012; Peterson, 2000; Podsakoff et al., 2003): Here, the researchers especially paid attention to issues of contextual fit and word redundancy, but they also looked to include various items with slightly different shades of meaning, which enabled this study to construct multi-item subscales. Additionally, the wording had to be simple, short, familiar to the respondent and clear, as well as precise. The researchers’ aim was also to eliminate items that were ambiguous, impossible in the context, had questionable relevance or contained obvious social desirability (Nederhof, 1985). There remained only words and phrases that are used in common language. In parallel, jargon and trendy slang, as well as double-barrel statements, were avoided. The reduced set of items should also only include items that exhibit meaningful semantic differences (Ablers & Hildebrandt, 2006).

By means of an iterative discussion process, a total set of 80 items was generated which was hypothesized to fully capture all the essential aspects of the domain of the focal construct. Note that none of these items exhibited negative expressions (was reverse coded). Due to the conceptualization of eWOM trust as a confident or positive orientation best mirrored by positive judgments, feelings, and motivations towards the eWOM information, this research refrained from using negatively worded items. Their enclosure in the new scale would automatically lead to the measurement of eWOM distrust, a related but conceptually different construct (see Chapter 3). Additionally, in scale development research, items worded in the opposite direction often perform poorly (DeVellis, 2012). However, the author is aware that some response bias in the final scale may be attributed to this circumstance.

Literature regularly demands that besides the deduction of items from reviews of literature, also suggestions by appropriate people - including (i) representatives of the population to which the construct of eWOM trust is expected to generalize (i.e., online consumers - individuals using the Internet for product research and/or shopping), as well as (ii) persons from the marketing environment - should be considered (e.g., DeVellis, 2012; MacKenzie et al., 2011). The research had to ensure that all content areas of the construct were systematically sampled. That is, they are representative for the various aspects of trust. Hence, all items were judged in terms of translation validity. Basically, the assessment of translation validity includes the evaluation of the items in terms of both (1) content and (2) face validity. By referring to Straub et al. (2004), the term content validity reflects “the degree to which items in an instrument reflect the content universe to which the instrument will be generalized” (p. 424). Other authors put forward similar definitions (Kerling, 1973). According to Haynes et al. (1995), content validity is concerned with “the degree to which elements of an assessment instrument are relevant to and representative of the targeted construct for a particular assessment purpose” (p. 238). Here, by “elements”, the authors refer to the diverse items of the scale, the response formats, as well as to the instructions to the respondents. In line with Netemeyer et al.’s (2003) argumentation, this thesis understands “representativeness” as the degree to which the items seem to be proportional to the different aspects of the focal construct and the degree to which the whole population of the aspects of trust has been sampled. Hence, it has to be clarified whether the individual items are representative for the distinct sub-dimensions of trust or not, but also the degree has to be assessed to which the items as a set collectively represent the entire content domain of the construct (MacKenzie et al., 2011). Literature often suggests that this can be achieved by a screening of the instrument by judges who have professional expertise in the area of research (DeVellis, 2012).

While some authors refrain from separating face validity from content validity, others do so. The same is true for this thesis. In agreement with the work of Netemeyer et al. (2003), this thesis understands face validity as the mere appearance that a measure, in addition to having pragmatic or statistical validity, is practical, pertinent and related to the purpose of the instrument as well (Kaplan & Saccuzzo, 1997; Nevo, 1985). This means that a measure should not only be valid, but also should “appear valid” in the eyes of the respondents. A face valid measure ensures that the items, instructions and the response format are clear and understandable. While face validity is more concerned of the evaluations of the items by the respondents of the target population, content validity on the other hand is focused on the scale’s validity from the viewpoint of experts. In order to cope with both forms of translation validity, as well as to trim and refine the pool of items, interviews with both marketing experts and online consumers were conducted.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science