Measuring the Degree of Internationalization of a firm: A Reply
- PDF / 5,337,607 Bytes
- 14 Pages / 450 x 771 pts Page_size
- 99 Downloads / 200 Views
Palgrave Macmillan Journals is collaborating with JSTOR to digitize, preserve, and extend access to Journal of International Business Studies ® www.jstor.org
180
JOURNAL OF INTERNATIONALBUSINESSSTUDIES,FIRST QUARTER1996
closed my work by discussing the conceptual and statistical properties of the composite measure, labeled DOIINTS, and their implications for construct and content validity. In an act of singular scholarship, RKR assess the theoretical rigor of that work, judge the logic of the methodological outcomes, and test the predictive usefulness of the measure through a step-by-step replication. In the process of their replication, RKR evidently were disturbed by my report. Their concern was not limited to proceduralissues but included conceptual and interpretative points. As the reader of their commentary will recall, two concerns underlie their objections: "First, it is debatablewhether or not the unidimensionality of the construct is theoretically justified .... Second, it is questionable whether the statistical methodology that resulted in a single factor is appropriate" (RKR, pp. 168-69). From my admittedly sensitive point of view, these two charges assert that the reader should seriously discount, if not dismiss outright, my original work. I disagree. I would like to begin my reply by addressing material elements of RKR's second qualm: the methodology was wrong. In sequence, I will address the issues of (1) The Appropriate Method, (2) Differential versus Unit Weighting, (3) Item Design and Selection, (4) Content versus Construct Validity, and (5) The Practical Utility of DOIINTS. Once done, I will turn to the more fundamental charge that "an overall index might be theoretically suspect" (RKR, p. 169). More precisely, this discussion will consider two matters: the theoretical integrity of the unidimensionality of DOIINTs and the matter of 'prematureversus past due.' THE APPROPRIATE METHOD My original paper used the "item-total analysis for constructing homogeneous measures"method to estimate the internal consistency of a linear combination of single-item proxies of the degree of internationalization [Nunnally 1978]. This algorithm correlates each potential item with the item-corrected scale score, the latter defined as the sum of the scores of all other items. So-called good items are ones that correlate significantly with the scale score. There is absolute clarity about the preferredoutcomes of this procedure: "The ideal is to obtain a collection of items which has a high averagecorrelation with total scores and is dominated by one factor only" [Nunnally 1978: 274]. RKR take particular exception with the particulars of the item-total method. At points, they declare that "it is questionable whether the statistical methodology that results in a single factor is appropriate" (p. 169), "the process of retaining only those variableswith high intercorrelationsand subsequently subjecting the chosen set to factor analysis to demonstrate unidimensionality is unreliable"(p. 169), and "there seems to be a fund
Data Loading...
