The increasing use of structural equation modeling provides social scientists with a powerful analytic tool for describing the interrelations of both manifest and latent variables (Anderson and Gerbing, 1988). Although a number of writers recommend (Loehlin, 1998) simultaneous solutions of the measurement model and the structural model, the two-step approach (Anderson and Gerbing, 1988) presents some unique advantages by separating the two phases. The study presented here performs the first step of the two-step approach; that of developing a measurement model with confirmatory factor analysis (Hatcher, 1994).
The data are from a correlation matrix produced in a study (Hitt et al., 1996) of the relationship of the market for corporate control, management control systems, and firm innovation. This study included both archival and survey data from a sample of 250 industrial manufacturing firms. The use of archival data is common in the strategic management literature and, although based on self-reported data, is considered reliable. Past research has demonstrated the general reliability of self-reported performance data (Dess & Robinson, 1984), for the publishers of these data sources use in-house control procedures to verify the accuracy of their reporting. A correlational analysis of the data from the original company reports against the data reported by the publishers would likely yield a high correlation coefficient (Richard A. Johnson, personal communication, April 18, 2000). Hence, the secondary data used in organizational studies can be considered highly reliable. The survey data were collected from CEOs or their designees and exhibited a 36.9 percent response rate (Hitt et al., 1996). Thus, th ...