Chapter 4 – Research Methodology
4.2 Sampling Methods and Data Collection Procedures
4.2.1 Sample Size
The sample size determined in this study is based on two types of data analysis techniques; Exploratory Factor Analysis (EFA) and SEM. Sample size is an important factor in making generalisations about the constructs under investigation. It should provide reliable estimates and reflect the population parameters as closely as possible with a narrow margin of error (Sekaran, 2003). In general, the reliability of the factors emerging from a factor analysis depend on the size of the sample, although there is no consensus on what the sample size should be (Hair et al., 2010). Most researchers agree, however, that there should be more respondents than variables (Bryman & Cramer, 2005).
Academics (Hair et al., 2010; Kline, 2005; Pallant, 2007) recommend a minimum sample size of 100 or more for conducting EFA with at least five times as many observations as the number of measured items to be analysed and a more acceptable ratios of 10:1. Thus, considering these recommendations, a minimum sample size of at least 280 respondents would be appropriate for an EFA since there were 56 items. For SEM analysis
using Maximum Likelihood Estimation (MLE), in general, a sample size of at least 200 to 400 respondents is recommended (Hair et al., 2010; Tanaka, 1993). Thus, taking into account the claims of Hair et al. (2010) and Tanaka (1993), the ideal sample size for using SEM in this study should be between 200 and 400 observations. Accordingly, the minimum sample size for this study was set at 480 usable questionnaires to test the 16 hypotheses and satisfy the four research objectives. However, 550 questionnaires were distributed for the actual data collection to guarantee at least 480 usable questionnaires after taking account of the claim by Hair et al. (2010) that 100% completion of questionnaires was highly unlikely. Some questionnaires may be unusable or incomplete and invalid; incomplete questionnaires were therefore excluded from the analysis (Clemes et al., 2011; Kim, 2003; Ryu et al., 2012; Tabachnick & Fidell, 2007).
4.2.2 Sample Derivation
Before data collection could commence in Malaysia, approval from the HEC of Lincoln University was required. This was the second approval from the HEC after permission was granted for conducting the focus group sessions. Approval from the Economic Planning Unit of Malaysia was also needed. After receiving the approvals data collection in Malaysia commenced in January and ended in March, 2009.
In the interests of generalizability in the data collection and to improve the representativeness of the sample, following Brady et al.’s (2001) and Cronin et al.’s (2000) sampling method, four moderate upscale restaurants were investigated: the Hard Rock Café, T.G.I Friday’s Restaurant, DÔME Café and Victoria Station Restaurant. Most of the moderate upscale restaurants are located in major cities in Malaysia such as in the Klang Valley, Penang and Johor Bharu, where spending power and population concentration are higher (Euromonitor International, 2008). However, owing to monetary restrictions and time constraints, data collection was confined to restaurant patrons of moderate upscale restaurants in the Klang Valley.
4.2.3 The Data Collection Method and Procedure
In order to ensure effectiveness during data collection, several techniques such as intercept sampling, convenience sampling and a self-administered questionnaire were adopted.
4.2.3.1Intercept Sampling
An intercept (analogous to the mall intercept or street intercept) sampling was adopted since the researcher was not permitted to enter the restaurants. Using the intercept sampling method in this study was consistent with the work of other researchers (Ko & Pastore, 2005; Li, 2002; MacLaurin & MacLaurin, 2000; Shonk, 2006; Weiss et al., 2004). Moreover, in the last few years, the intercept sampling survey method has been widely used in market research and is a feasible alternative to traditional survey methods (Bush & Hair, 1985; Gates & Solomon, 1982; Keillor et al., 2007; Miller, Wilder, Stillman, & Becker, 1997).
Intercept sampling was deemed the most suitable method for this study after considering the advantages and disadvantages of various data collection methods such as door-to-door personal interview, telephone interview, mail survey and email survey (Bush & Hair, 1985; Gates & Solomon, 1982; Miller et al., 1997). The main advantages of intercept sampling are as follows: (1) it is less expensive; (2) there is greater control; (3) it has a high response rate; (4) accurate data is obtained in a face-to-face manner; (5) it represents the correct respondents; and (6) the respondents are able to supply a real experience (Bush & Hair, 1985; Gates & Solomon, 1982; Ko, Zhang, & Pastore, 2007; Miller et al., 1997).
4.2.3.2Convenience Sampling
Intercept sampling is a classic example of convenience sampling (Mallett, 2006), the sampling technique used in this study, following the work of several researchers (Kivelä, Reece, & Inbakaran, 1999; Ko et al., 2007; Ryu et al., 2008). The advantages of convenience sampling are: (1) a large number of people can complete the questionnaire; (2) the data collection can be conducted in a short time; (3) it is relatively inexpensive; (4) convenient; and (5) the respondents are often selected because they happen to be in the right place at the right time (Malhotra et al., 2002; Zikmund, Ward, Lowe, & Winzar, 2007).
In addition to the advantages of convenience sampling, this method is considered as an acceptable sampling technique for data collection if the purpose of the study is to: (1) test the theoretical; (2) test the hypotheses regarding how particular variables relate to behaviour as stated in the research objective; and (3) provide evidence in supporting or
rejecting the theory tested, regardless of the nature of the sample (Leary, 2004; Reynolds, Simintiras, & Diamantopoulos, 2003).
In respond to ethical reason in data collection, measuring respondents’ attitudes and perceptions through questionnaire surveys as used in this study requires respondent consent. Therefore, only restaurant patrons age 18 and above whom consent can be selected as respondents. Following Suhartanto (2011), to minimise the weakness of using convenience sampling, data were collected from several moderate upscale restaurants in Klang Valley area.
4.2.3.3Self-Administrated Questionnaire
Testing the 16 hypotheses formulated for this study required a large sample size (more than 400 sample size); therefore, a self-administered questionnaire was adopted for collecting the data due to cost and time considerations. Self-administered questionnaires are often used in shopping centres, or other central locations, where the researcher has access to captive respondents (Malhotra et al., 2002; Zikmund et al., 2007). In fact, the self-administered questionnaire field survey approach has been the dominant study design for service quality studies in the hospitality field (Aaker et al., 2005; Kivelä, Reece, et al., 1999).
Thus, this study used a self-administered questionnaire based on the reason that: (1) the respondents responsible for reading and responding to the questions; (2) it enables the researcher to distribute numerous questionnaires to many respondents in different places simultaneously; and (3) the data can be collected from various restaurants at different locations in a relatively short time period (Malhotra et al., 2002; Suhartanto, 2011; Zikmund & Babin, 2007; Zikmund et al., 2007).
4.2.3.4The Actual Data Collection Method
During the data collection sessions, the researcher intercepted every first of the five potential respondents (restaurant patrons) as they left the restaurant and explained the nature of the survey to them. They were informed that their participation in the study was voluntary and the information provided would be kept private and confidential. Restaurant patrons aged under 18 years were excluded from the sample because it was expected they might encounter difficulties in interpreting the questionnaire (Clemes et al., 2011; Miller et al., 1997; Weiss, 2003) and would probably not be making the decision on future
behavioural intentions (Weiss, 2003). The questionnaires were administered by the researcher in a public area near the entrance to moderate upscale restaurants during lunch (12.00 noon – 3.00 pm) and dinner time (6.00 pm – 9.00 pm) over a period of three months. Additionally, a non-response bias test (see Section 5.1.2) was conducted before analysing the data.
Restaurant patrons were asked to complete the 10 – 15 minute questionnaire and return the completed questionnaire to the researcher. By utilizing this method, a total of 480 questionnaires were expected to be completed. Respondents could ask the researcher for assistance if they had difficulty interpreting or understanding the questions which were based on performance-only items; there was no need for the respondent to be approached twice.
Response rates between 20% and 30% are common for hospitality studies (Hartline, Woolridge, & Jones, 2003). The problem of low response rates can be minimised by providing prepaid and non-monetary incentives (Aaker et al., 2005; Willimack, Schuman, Pennell, & Lepkowski, 1995; Yu & Cooper, 1983). Willimack et al. (1995) report no increases in measurement error due to using incentives. In line with this suggestion, incentives were given to the respondents in order to encourage them to participate and to ensure that the study achieved an acceptable response rate. Restaurant patrons were told that if they completed and returned the questionnaires to the researcher they would receive a gift, an option between a key-chain or a fridge magnet as a token of appreciation.