CHAPTER 4: RESEARCH METHODOLOGY
4.4. Sampling Process
The use of a sample to obtain precise information about a population is a very efficient technique, which has been extensively used in the literature (Hair et al., 2010). In spite of taking less time and providing cost savings when compared to a complete census, sampling enables researchers to draw general conclusions about the whole population. In this section, the sampling process is presented as a series of procedures, which includes establishing the research setting, identifying the sampling frame, selecting the sampling method, determining the sample size, and identifying target respondents (Churchill, 1995).
(1) Research Setting
Thailand is selected for the present study because its economy relies largely on exports.According to the World Development Indicators database, September 2009, Thailand is one of Asia‟s most export-oriented countries, where exports accounted for around 70 percent of GDP (www.worldbank.org/data/). More specifically, manufacturing exports have played an important role in Thailand‟s economic success. Thailand‟s industrial structure has undergone significant changes and has now become primarily export oriented (Phan, 2004). Thailand has enjoyed decades of robust, and sustainable, economic expansion by successful forays of Thai goods into the international markets. The share of manufacturing exports in GDP increased from just 7% in 1980 to around 40% in 2007 (Dhannani and Scholtes, 2002; Asian Development Bank, 2009). In addition, the growth of manufacturing exports is not only confined to two or three manufacturing products, but spread over a wide cross- section of the manufacturing sector (Phan, 2004). Thus, understanding the role of
export marketing resources in manufacturing export firms in Thailand is useful for export marketing and export performance research.
(2) Sampling Frame
After identifying the research setting, the next step is to select the sampling frame. The sampling frame of manufacturing export firms used for the study is based on the 2008 Thailand Exporter Directory, which contains a list of approximately 10,000 firms. This is a reliable and legitimate source because it was developed by the Department of Export Promotion, Ministry of Commerce, which oversees and supports exporters in Thailand. Thus, it should provide the most complete set of manufacturing export firms in Thailand.
(3) Sampling Method
A critical decision involving the sample is how the sample units are to be selected. This decision requires the selection of a sampling method (Hair et al., 2010). Sampling methods can be divided into two broad categories of probability and nonprobability sampling. The probability sample seems to be the preference of a majority of researchers, enabling them to capture the characteristics of the population in the sample through the element of chance (Tabachnick and Fidell, 2007; Hair et al., 2010). Through systematic random sampling, every fifth name was selected. Thus, the sample units are 2,000 manufacturing export firms.
(4) Sample Size
Within a quantitative survey design, determining the size is essential (Bartlett et al., 2001; Hair et al., 2010). In the study, the sample size is determined on the basis of the following considerations.
Sample size can be assessed by considering the trade off between significant level and statistical power (Churchill, 1995; Hair et al., 2010). Cohen (1988) provided a guideline for the size of sample to achieve alpha levels of at least .05 with power levels of 80 percent. To achieve this level, all three factors must be considered simultaneously (effect of size, alpha, and sample size). With an alpha level of .05, a power of 80%, research requires a sample size of 130 (Cohen, 1988). Other scholars recommend determining the sample size by analytical techniques, especially in multivariate data analysis. Bartlett et al. (2001) pointed out that if researchers plan to use factor analysis, the sample size should not be less than 100 to obtain valid results. In the case of structural equation modeling (SEM), a number of statisticians assert that a sample size from 100 to 200 is often recommended (Hair et al., 2010; Bagozzi and Yi, 2012). Thus, when structural equation modelling is used for data analysis in the study, the sample size of 200 would meet the criteria.
(5) Target Respondent
According to the conceptual framework from the previous chapter, information is required from respondents with knowledge of corporate philosophy, export marketing superiority, and export performance. Huber and Power (1985) indicated that the responses from the most knowledgeable informant can be more accurate than taking
average responses from multi-informants with a range of knowledge. Bowman and Ambrosini (1997) reviewed the empirical survey studies published in the Strategic Management Journal, and found that a majority of the studies used single respondents to make inferences of the organisation. Gatignon et al. (2002) argued that using a single knowledgeable informant is a valid approach to measuring strategy research questions and that bias introduced by such an informant is likely to be negligible compared to multiple informant responses. Kahn (2001) noted that respondents from different functions may provide conflicting results. Thus, the single key-informant approach appears more attractive, and the study relies on a single well-informed respondent from each firm (e.g., Piercy et al., 1998; Morgan et al., 2004; Lings and Greenley, 2005; Hooley and Greenley, 2005; Cadogan et al., 2006; Leonidou and Katsikeas, 2010).
The quality of response is usually obtained from informants whose roles are closely related to the phenomena under study (Kumar et al., 1993). It is generally accepted that the responses derived from a chief executive officer (CEO), a key member of the firm, can reflect the actual organisational characteristics most accurately. Cycyota and Harrison (2002) stated that CEOs may be the only reliable source of certain information about aspects of the firm as a whole. Zahra and Covin (1993) noted that the CEOs provide data as reliable and valid as multiple informants. The use of CEOs in organisational research is widespread, since firms are ultimately a reflection of their top management (Hambrick and Mason, 1984). Based on this view, CEOs are undoubtedly chosen for specific reasons. CEOs are supposed to be knowledgeable about the issues being researched. In relation to export performance, they tend to have the highest familiarity with this information, and in some non-large firms, much of
this information is limited to top management (Pelham, 1997). Moreover, not all firms have a formal marketing department. Therefore, the means of collecting data from CEOs rather than from marketing managers seems to be the most efficient and effective method for the present study. Nevertheless, it is likely that the CEOs may assign another key member of the firm such as managing director (MD), general manager (GM), or senior-level manager to take part in the survey instead. In some cases, MD or GM may be the highest management position in the company. A number of previous studies have considered this issue as an unavoidable limitation, thus response from senior-level managers was still acceptable (Venkatraman and Ramanujam, 1987; Kotabe and Czinkota, 1992; Leonidou and Katsikeas, 2010).