Survey questionnaires are generally of two types: open-ended and closed-ended11, each with its own advantages. In open-ended questionnaires, respondents create their own answers whereas closed-ended ones are limited to yes or no answers, categories of responses and rank-ordered responses or scales. Closed-ended questionnaire responses are easier to record and analyse (Aaker, Day, Kumar, & Lawley, 2005).
The closed-ended questionnaire format was chosen to obtain the data for this study. Following Churchill (1979), the questionnaire was developed through a multi-stage process. A comprehensive review of the literature was performed to help obtain conceptual and measurement information about potential items that were further refined12 based on the findings from the focus group discussions. A list of measurement items was compiled, grouped in accordance with the subdimensions, and divided into each of the primary dimensions (see Tables 4.1 to 4.3) and then transformed into a prototype questionnaire.
The questionnaire had two parts: the measurement items and the evaluative part. The first part focussed on the items used to measure each construct; they were mostly adopted from previous foodservice studies but tailored to suit the multiracial restaurant patrons in Malaysia (Powpaka, 1996). The measured items included in the questionnaire were based on performance-only items following the suggestion in prior studies. Several researchers (Babakus & Boller, 1992; Bojanic & Rosen, 1994; Brown et al., 1993; Carman, 1990; McDougall & Levesque, 1994) used two separate measurements (expectations and perceptions measurement) designed by Parasuraman et al. (1988), but found theoretical and operational problems. Furthermore, some researchers (Fu & Parks, 2001; Teas, 1993) have indicated that they had difficulties in collecting expectations and perceptions data separately. Thus, based on the weakness highlighted by these researchers, this study used performance-only items in the questionnaire. In addition, performance-only items in several studies have demonstrated a higher correlation, higher adjusted R2 and superior construct validity and reliability compared with measurements of service quality based on
11
Closed-ended questionnaires are also known as structured questionnaires.
12 It should be noted that, although the existing scales were used as resources for the items, parts of those
the disconfirmation paradigm (Cronin & Taylor, 1994; Dabholkar et al., 1996; Zeithaml et al., 1996).
Several academics (Churchill, 1979; Hair, Black, Babin, & Anderson, 2010; Robert, 2002), emphasise that the measurement of constructs with single items has been criticised in the marketing literature, mostly because single items can cause measurement errors and often cannot capture the richness of a concept. A set of measurement items may better capture various facets of a construct than a single measurement item (Kline, 2009). Furthermore, in employing SEM, Hair et al. (2010) claim that for a single item measure, reliability and validity cannot be examined as with multiple-item measures. A single-item measure usually causes a problem with model identification and is more likely to be empirically under-identified than models that have at least three measured items per construct (Blunch, 2008; Byrne, 2009; Chinna, 2009; Kline, 2005; Schumacker & Lomax, 2004). Moreover, academics (Byrne, 2009; Chinna, 2009; Kline, 2005) suggest that at least three measured items per construct should remain after carrying out the CFA in order to counter problems (such as specification error and non-convergence of iterative estimation) which are more likely to occur in a model that had only two measured items per factor. In line with these claims, 12 subdimensions13, the primary dimensions and the related constructs of service quality, customer satisfaction, perceived value, restaurant image and behavioural intentions were measured with multiple items so that they could be measured more accurately.
The wording of the questionnaire is also an important consideration. Altinay and Paraskevas (2008) acknowledged questionnaire design as a communication exercise; words of questions often have different meanings for different people. It is important to choose the right words and minimise technical terminology and jargon (Altinay & Paraskevas, 2008; Schall, 2003). Thus, all questions in the questionnaire were simple, ordinary, positive and unambiguous so they could be understood by the respondents.
The sequence of statements is crucial in a questionnaire because it may influence the nature of the respondents’ answers and it may produce better data (Schall, 2003). Thus, in compliance with the suggestions of Schall (2003), in the questionnaire specific statements were asked first, followed by general questions14 (summative overall-
13
The subdimensions of each of the primary dimensions were expected to be multi-item measures.
satisfaction). All questions were arranged from one topic to another in a logical manner with questions focusing on completing the section before moving to the next section.
In the second part of the questionnaire, multiple-item likert scales were used in Sections A to D. Likert scales, according to Aaker et al. (2005), require a respondent to indicate a degree of agreement or disagreement with a variety of statements related to the attitude or object. A seven-point likert scale ([1] indicating “Strongly Disagree” to [7] indicating “Strongly Agree”) was used in line with Schall’s (2003) recommendation. Schall (2003) determined that a seven-point anchored scale was the optimum size for hospitality industry questionnaires when comparing four, five, and seven point scales. Based on Schall’s (2003) research, a seven-point scale was used with a balance between equal numbers of positive and negative responses and with a neutral point; to collect more accurate data and ensure the usage was also consistent with other studies in service quality (Dagger et al., 2007; Ha & Jang, 2010b; Kim & Moon, 2009; Ko & Pastore, 2005; Qin & Prybutok, 2008).
The demographic section of the questionnaire was designed to obtain personal information about the respondents such as gender and educational level, which were measured with a multiple-choice single response format using nominal and ordinal scales. A nominal scale (also known as a categorical scale) presents only categories or classes and usually measures demographic characteristics such as gender and marital status (Abu Samah & Suandi, 1999; Hair et al., 2010). With an ordinal scale, variables are ordered or ranked in relation to the amount of the attribute possessed; measured items usually can be compared with other measured items in terms of a “greater than” or “less than” relationships such as household income and average spending; they provide no measure of the actual amount, only the order of the values (Hair et al., 2010).