Revisiting Shareholder Value Creation via International Joint Ventures: Examining Interactions Among Firm- and Context-Specific Variables

Canadian Journal of Administrative Sciences, Jun 2004 by Merchant, Hemant

Turning to the contextual variables, this study measured cultural distance (CULDIST) in terms of Kogut and Singh's (1988) index which is extensively used to measure cultural distance between two countries. This index represents cumulated deviation between the variance-adjusted culture score(s) of partner countries. The culture scores were obtained from Hofstede (1980), which first published these numerical scores.

This study operationalized market size (MKTSIZE) in terms of private consumption as a percentage of joint venture host country's total gross domestic product. The data on consumption and gross domestic product were available from the Perm World Tables, a compilation of several macro-economic variables. A detailed description of how these tables are constructed can be found in Summers and Heston (1991) and is therefore not provided here.

The Penn World Tables also published numerical scores regarding openness of a country's trading regime (OPENNESS). The tables defined this variable as the total value of a country's imports and exports as a percentage of the country's real gross domestic product per capita. Thus, higher scores reflect greater trading openness, and vice versa.

This study operationalized investment climate in a host country (CLIMATE) in terms of aggregate political risk scores reported by Political Risk Services, a prominent international agency that regularly publishes country-specific political risk ratings. A higher rating suggests greater political risk, that is, less favourable investment climate, and vice versa.

This study utilized American parents' principal fourdigit Standard Industrial Classification (SIC) code to account for industry-level effects (INDUSTRY). The principal SIC code refers to the industry niche in which a firm does most of its business. The SIC codes identify various lines of business a firm operates in, and are fixed: each code refers to a specific product group. This study relied on the COMPUSTAT database to identify each firm's principal SIC code.

Cluster Analysis Methodology

Appropriateness of use. Cluster analysis refers to an array of multivariate techniques for reorganizing data into homogenous groups. The methodology is appropriate for this study principally because it accommodates interconnectedness among a large set of variables (e.g., Miller & Friesen, 1984). Cluster analysis can identify distinct groups of firm-specific and contextual variables: it enables a more meaningful evaluation of this study's research question about the collective impact, not individual impact, of above-mentioned variables on shareholder value creation. Thus, this study's emphasis differs sharply from that of previous work that has investigated the role of individual variables on value creation in joint venture parents. Although regression analysis would be appropriate for these latter studies, it is inappropriate given this study's research question.

The above position is valid even though regression analysis can specify the interaction effect of variables. However, modeling higher-order interactions raises the issue of how-even whether-the impact of three-way (and/or higher-order) interactions can be meaningfully interpreted (Miller & Friesen, 1984, p. 15n). Moreover, specification of interaction effects leads to multicollinearity and its derivative challenges. These issues justify the use of cluster analysis over regression for configuration studies such as the present one (Dess, Newport, & Rasheed, 1993, p. 786; Miller & Friesen, 1984, pp. 56-57). Indeed, a key merit of cluster analysis is that it accounts for multiple interactions among variables concurrently (Dess et al, 1993, p. 789; Miller & Friesen, 1984, Chapter 2).