Content Infrastructure Management

4 Empirical test of the content allocation model (p. 92)

In this chapter, the mid-range theoretical framework on content allocation, which integrates different theoretical lenses into a coherent whole, will be subjected to an empirical test. This requires to transform the theoretical language into an observable language (Kerlinger/ Lee, 2000, p. 54). In other words, the constructs have to be operationalized as measurable variables. In doing so, it is necessary to satisfy both the theoretical and the empirical requirements. One method that provides a formal structure which allows the matching of theory and data is that of structural equation modeling (SEM). This statistical modeling technique provides rules on how to specify a variance-based theoretical framework in a way that recognizes the requirements of the statistical procedures that are applied to rigorously estimate and evaluate the parameters of the model.

The theoretical groundwork that was laid by developing the theoretical framework facilitates the model building process. However, in order to specify and subsequently test the framework, the requirements of the modeling technique have to be considered a priori. Therefore, the fundamentals of the SEM method will first be introduced in the following chapter (see chapter 4.1). Subsequently, the constructs of the midrange theoretical framework on content allocation will be operationalized (see chapter 4.2). After the development of the measurement instrument, the conduct and analysis of the empirical study will be outlined. This includes the data collection (see chapter 4.3), the presentation of major descriptive characteristics of the sample data (see chapter 4.4), and the more extensive model estimation and evaluation process (see chapter 4.5).

4.1 Fundamentals of structural equation modeling

Within the last twenty years, structural equation modeling (SEM) techniques have become increasingly popular among social scientists (e.g., Chin, 1998a, Hildebrandt/ Homburg, 1998). They allow the rigorous statistical examination of theoretical relationships. Their popularity is essentially attributed to their ability to combine an econometric perspective, focusing on prediction, with psychometric modeling, which focuses on the measurement of not directly observable (i.e. latent) variables by multiple observables – also called indicators or manifest variables (Chin, 1998a, p. vii, Lee/ Barua/ Whinston, 1997, p. 120). The approach is primarily confirmatory in nature. It is generally used to determine whether a pre-specified model is valid, rather than to find a model by exploring the data – although it often includes some exploratory elements in the analysis (e.g., Chin, 1998a, Homburg/ Dobratz, 1991, 1992). SEM techniques allow the researcher to simultaneously test the strength of the relationships between multiple latent variables and the reliability of the measures of the latent variables (Chin, 1998a, p. vii). The relationships between latent theoretical variables form the structural model, which sometimes is also called inner model (see Figure 4.1-1). The structural model equals a variance-based theoretical framework.

Accordingly, the mid-range theoretical framework on content allocation, as illustrated in Figure 3.5-1, represents a structural model of content allocation. As an example, content specificity is hypothesized to have a direct negative impact on the comparative transaction cost advantages of centrally vs. decentrally deployed content (H1a-), and an indirect positive impact (two negative result into one positive relationship) on the allocation of content via comparative transaction cost advantages (H2a-).