Proposal of a validation framework for a new measurement model and its application to the export performance construct.
| Autor | Carneiro, Jorge |
| Cargo | Report |
INTRODUCTION
Several phenomena in the social sciences exhibit a complex and abstract nature, which poses important challenges, both substantive and methodological, to their conceptualization and operational representation. The latent nature of a complex construct means that it cannot be directly observed, but needs to be inferred from its manifestations (Netemeyer, Bearden, & Sharma, 2003). Moreover, it has been recognized that "specifying the relationship between concepts and operational indicators is equally important to social research as the substantive theory linking concepts to one another" (Carmines & Zeller, 1979, p. 11).
A construct of critical importance to research on exporting is export performance. However, although the construct has received the attention of several scholars, none of the already proposed measurement models has reached consensual acceptance. This lack of agreement makes it difficult to compare research findings and develop a shared body of knowledge. As a result, the empirical literature has reached mutually inconsistent results about the effects of determinants of export performance (Zou, Taylor, & Osland, 1998). And although a multidimensional approach to represent the complex nature of the export performance construct has been advocated (Diamantopoulos, 1999; Katsikeas, Leonidou, & Morgan, 2000; Leonidou, Katsikeas, & Samiee, 2002; Madsen, 1987, 1998), most researchers have employed unidimensional models that do not adequately capture the multifaceted nature of the phenomenon.
In addition, even those scholars who have advanced quite elaborate measurement models of export performance (e.g., Cavusgil & Zou, 1994; Lages & Lages, 2004; Lages, Lages, & Lages, 2005; Shoham, 1998, 1999; Styles, 1998; Zou et al., 1998) have not provided a thorough validation screening of the proposed operational models. It is this methodological issue that we address in this paper. The study has the following objectives:
* to offer a rather comprehensive and integrated set of procedures based on structural equation models [SEM] for validating measurement models of complex and multifaceted constructs, which is rooted both in conceptual reasoning and empirical screening;
* to empirically apply these validation procedures to the development of a new measurement model of the export performance construct and the assessment of the degree of satisfactoriness of such a model; and
* to discuss the nature and structure of the construct based on the interplay between conceptual reasoning and empirical results.
In fact, two stages are involved in theory building: the first is the specification of "relationships between theoretical constructs", and the second is the description of "relationships between constructs and measures" (Edwards & Bagozzi, 2000, p. 155). These two stages are critical, since theory building requires "a high degree of correspondence between abstract constructs and the procedures used to operationalize them" (Peter, 1981, p. 133). In this paper, we address mainly the second aspect of theory building.
This paper is organized as follows. After this introduction, we describe the validation framework and present data collection and data treatment procedures. We then apply the validation framework, step by step, to the development of a new measurement model of export performance, and new highlights into the nature of the construct are addressed. Final remarks and some suggestions for future studies close the paper. Although the paper is rather methodological, we also discuss relevant theoretical implications that can be drawn from the analyses.
THE VALIDATION FRAMEWORK
We reviewed and contrasted several works, drawn from quite diverse areas of study, including psychology, education, organizational studies, statistics, strategic management, marketing, and international business. By putting together and operationalizing several perspectives and criteria by which to judge the adequacy of measurement models, we believe we have assembled a useful framework for construct validation.
Our validation framework covers the following eight steps:
1) Conceptualization of the construct and pursuit of content validity
2) Exploratory empirical verification of the dimensionality and content of the construct
3) Advancement of theoretically plausible competing models
4) Assessment of psychometric properties
5) Assessment of concurrent and predictive validity
6) Verification of overall adequacy of the measurement model
7) Verification of (measurement parameters) stability
8) Selection of the most likely model
Table 1 presents the steps and procedures involved in the validation process.
METHODS
Population and Sample
A survey was conducted of the largest Brazilian exporters of manufactured products selected from a list provided by FUNCEX, a private foundation supported by Brazilian exporters. Firms controlled by foreign capital were excluded because of potential different objectives and possible transfer pricing mechanisms. Service firms, exporters of commodities and trading companies were also removed in order to make the sample more homogenous, and thus avoid possible confounding effects, resulting in a population of 3,057 exporters of manufactured goods. The unit of analysis was the export venture, i.e., the exporting of a given product line to a given country (Matthyssens & Pauwels, 1996). A sample of 448 exporters was obtained resulting in a response rate of 15.5%, after correcting for non-eligibles. No systematic bias was observed between respondents vs. non-respondents or between early versus late respondents.
Data Collection and Data Treatment Methods
A four-page structured questionnaire covered not only indicators of export performance but also several variables related to determinants of export performance; only the export performance variables are reported here. Firms were mailed a questionnaire with a pre-paid return envelope.
Semantic-differential scales of perceptual measures were employed instead of asking firms to provide objective information (Matthyssens & Pauwels, 1996; Shoham, 1998). This was deemed necessary to improve the response rate and minimize missing values since most firms do not keep objective public data for each export venture, segregated from the firm's other ventures. Moreover, it has been reported that subjective measures correlate highly with objective measures of performance as well as with overall assessments of performance (Dess & Robinson, 1984; Venkatraman & Ramanujam, 1987) and that managers' assessments are as reliable as data from objective sources (Wong & Saunders, 1993). Also, managerial decisions tend to be driven by perceptions rather than solely by "cold" data (Bourgeois, 1980; Matthyssens & Pauwels, 1996).
Variables and cases with more than 15% missing values were removed (Hair, Black, Babin, Anderson, & Tatham, 2006), which led to the exclusion of one indicator of export venture performance (past export venture's volume vs. other Brazilian firms exporting to the same country) and 34 cases. Since missing data exhibited an missing completely at random [MCAR] pattern at the 10% significance level, it was possible to estimate the missing values. Given that three estimation methods (mean substitution pairwise, regression imputation and EM approach) provided very similar estimates, a simple average of these three methods was used (cf Hair et al, 2006). The resulting sample (414 cases) showed no indication of the presence of multivariate outliers as far as the 10 remaining operational indicators of export venture performance were considered. Parameters were estimated by an asymptotic distribution-free method [ADF] because variables did not follow a normal distributional pattern. SPSS 15 and AMOS 7.0 were employed to run the statistical analyses.
APPLICATION OF THE VALIDATION FRAMEWORK TO A NEW MEASUREMENT MODEL OF EXPORT PERFORMANCE
The validation framework consisted of eight steps.
Step 1: Conceptualization of the Construct and Pursuit of Content Validity
First of all, it is necessary to (1-a) map the conceptual domain of the construct (DeVellis, 2003; Spector, 1992). This task involves the identification of "what is and what is not included in the domain" (Churchill, 1979, p. 67). Therefore, we had to decide on the appropriate conceptual domain of the export performance construct for which our measurement model would be developed and strive for content validity, or at least provide evidence of content adequacy (Schriesheim, Powers, Scandura, Gardiner, & Lankau, 1993).
Content validity was sought by means of an extensive review of the literature on the focal construct as well as theoretical reflection and consultation with academic experts. We searched the most prominent journals on International Business (Dubois & Reeb, 2000) for a 30-year period (1976-2005) in order to identify studies (conceptual, empirical, meta-analytical, and consolidation works) which seemed to represent the best efforts to characterize the multifaceted nature of the export performance phenomenon. We also reviewed the proceedings of two leading conferences in the field: the Academy of International Business [AIB] and European International Business Academy [EIBA]. Building on these studies, export performance is conceptualized as a multi-dimensional construct that includes several classes of measures (economic, market, behavioral, strategic and overall), two alternative frames of reference (absolute and relative), and two perspectives of temporal orientation (static and dynamic), as shown in Figure 1.
After mapping the domain of the construct, it is necessary to (1-b) define the breadth of coverage of the phenomenon. From the various classes of measures in the literature, we chose to concentrate on only one, the economic aspect of export venture performance. This was deemed necessary in order to avoid having too many indicators, which might lead to fatigue bias when eliciting...
Para continuar a ler
PEÇA SUA AVALIAÇÃOCOPYRIGHT GALE, Cengage Learning. All rights reserved.
