Sample Design

As has been mentioned above, there are two fundamental types of joint ventures: equity joint ventures (EJVs) and non-equity joint ventures (NEJVs). EJVs are the main form of IJVs in Thailand (BOI, 2005). EJVs comprise two or more legally distinct organisations (the parents), each participating in the decision-making activities of the jointly owned

entity (Geringer, 1991). In contrast, NEJVs are agreements between partners to collaborate in certain ways, but do not involve the establishment of a new corporate entity (Contractor and Lorange, 1988). A joint venture is regarded as an international joint venture if at least one partner has its headquarters outside the venture’s country of operation, or if it has an important level of operation in more than one country (Geringer and Hebert, 1989). Owing to the characteristics of the acquired dataset, this study concentrates only on EJVs. The database registers 2,251 IEJVs (international equity joint ventures).

Sampling methods are normally divided into two main types: the probability sampling method (representative or random sampling); and the non-probability sampling method, or judgemental sampling method (Fink, 2003). In the probability sampling method, the chance or probability of each case being selected from the population is known and is usually equal for all cases, whereas non-probability sampling does not employ the rules of probability theory and does not ensure representativeness. Babbie (1990) criticises the non-probability sampling method for the drawback that each case in the sample is selected from the total population by a method of subjective merit, and that the method is unable to answer research questions or address objectives which require statistical inferences in respect of the population in general. Selltiz et al. (1976) argue that a valid sampling method must allow the sample unit to be chosen in a manner which is systematic and objective, easily identifiable, clearly defined, each choice being independent of each other, and non-interchangeable. The sampling process must be rooted in standard criteria and should avoid errors, bias and any distortions.

Fink (2003) defines non-probability sampling as a method which cannot depend upon the rationale of probability and cannot be used to infer from the sample to the general population. Any generalistions obtained from this method must then be filtered through one’s knowledge of the topic being studied. Some studies have no choice but to use non- probability sampling either because, in social science research, there may be circumstances where it is not feasible, practical or theoretically sensible to do random sampling, or because of unintentional or unavoidable characteristics of the sampling method. Neuman (2003) divides the non-probability sampling method into seven types: haphazard,

accidental or convenience sampling; quota sampling; purposive sampling; snowball sampling; deviant case sampling; sequential sampling; and theoretical sampling. He divides the probability sampling method into four categories: simple random sampling; stratified sampling; systematic sampling; and cluster sampling.

For this study, a probability sampling method was chosen to select the final samples since, with this method, it is possible to answer research questions and to achieve objectives of this study which require the ability to statistically establish the characteristics of the population from the sample. This method is most commonly associated with survey-based research where the researcher needs to make inferences from a sample about a population (Neuman, 2003).

Previous studies (Beamish, 1988; Gomes-Casseres, 1987, 1989) have indicated that it is necessary to establish criteria for selecting the final samples (the qualifying IJV companies) mainly because of the heterogeneous characteristics of the population (in this case, all the IJV companies in the sampling frame), here particularly in terms of the percentage of each parent firm’s equity in the venture. That is, certain international joint ventures might cause distortion of a true picture of IJVs.

For the present study, then, three criteria have provided a framework for selecting the final samples:

1) IJVs in which neither partner holds more than ninety percent of the venture’s equity. This is because a number of researchers use at least 10% shareholding by parents as the minimum equity criterion for defining an international joint venture in a developing country (Beamish, 1988; Gomes-Casseres, 1987, 1989; Chowdhury, 1992). These researchers argue that a smaller percentage of equity ownership might not reflect a true picture of joint ventures and the nature of relations between the partners.

2) The companies in the database must still be operational (the present researcher checked the status of each IJV directly from the website of the Department of Business

Development, Ministry of Commerce of Thailand to determine their continued existence or termination).

3) The IJVs must have existed for at least three years, because the literature (Woodcock et

al., 1994; Pan and Chi, 1999; Pangarkar and Klein, 2004) shows that it takes

approximately three years for each partner to devise a comprehensive plan against which to assess the performance of the IJV.

After applying these criteria, the number of qualifying IJVs fell to 1,597 companies from the original 2,251 firms. Due to resource and time constraints, this study was unable to conduct a mail survey of all qualifying IJVs. Accordingly, an adequate sampling size needed to be determined using a statistical formula. For this study, a formula proposed by Krejcie and Morgan (1970) was adopted (see Appendix A). This method has been widely accepted by researchers, and is suited to studies where the exact size of the population is known (Cavana et al., 2001). Krejcie and Morgan (1970) have, moreover, converted this formula for use as a sampling table. In this table, the size of a population ranging from 10 to 100,000 is converted to provide the number of units required for an adequate sampling size. This provides a very convenient means for researchers to determine an adequate sampling size for a wide range of populations encompassed by the table. The table is popularly used by many researchers from a variety of fields of study (see, for example, Torres and Cano, 1995; Callan, 1995).

Ultimately, of the 1,597 qualifying companies 310 firms formed the sample size of this study. A stratified systematic sampling technique (Skalski, 1993) was adopted for sampling; it was used to classify the samples for each stratum (industrial sector), and then each stratum was sampled by a systematic sampling method. This method was used because it had a number of advantages compared to others. First, the stratified sampling method is suitable for use when differing characteristics of the population in each stratum are likely to appear. Second, this method is suitable for all sizes: there is no minimum size requirement for the sample. Third, it is relatively easy to explain to support workers, e.g., a research assistant (Saunders et al., 2003).