An Analysis of Construct Validity

This study presents a methodology to evaluate the construct validity of measures of collaborative venture (CV) performance and applies it to a sample of 69 CVs. After discussing content validity of the different measures, empirical results show that subjective measures of CV performance are reliable and present convergent validity. Results also support their discriminant validity with respect to other measures such as contractual changes, but not to longevity; no conclusive evidence is found with respect to survival. I would like to thank Peter Bentler, José de la Torre, Dick Goodman, Elaine Mosakowski, and Bill Ouchi for their helpful comments on earlier drafts of this paper. Funding for this research was provided by IESE, International Graduate School of Management, University of Navarra (Barcelona, Spain). MEASURES OF COLLABORATIVE VENTURE PERFORMANCE: AN ANALYSIS OF CONSTRUCT VALIDITY Introduction Construct validation research has been underemphasized in strategic management (Venkatraman and Grant, 1986). In particular, this is true of research on inter-firm collaborative ventures (1) (CVs) and, more specifically, of research on measures of CV performance (Parkhe, 1993 c). Whereas some work in the international management field has focused on the measurement of CV performance (Geringer and Hebert, 1991; Baird, Lyles and Reger, 1993), the validity of the underlying measures used in the literature –financial, objective, and subjective measures (Geringer and Hebert, 1991)– is still questionable. Content validity, reliability, and convergent, discriminant, and criterion-related validity, are the main aspects of construct validity. Baird, Lyles and Reger's (1993) review of methods for evaluating CV performance may be considered as a first step in assessing the content validity of existing measures. However, a clear description of the dimensions that define this concept is still missing. Furthermore, no attempt has been made to estimate the empirical validity of such measures (Parkhe, 1993 c). Therefore, assessing the construct validity of the measures of CV performance is important. Such a study is necessary for at least three reasons: first, to evaluate the soundness of inferences drawn from the relationships among CV performance and other theoretical constructs –inferences which are based upon the relationships among the measures of the constructs; second, to assist future researchers in the selection of their measures of CV performance; and, finally, to make valid recommendations to managers, based on the results from substantive research. This paper begins with a brief review of the measures of CV performance used in the literature, followed by a presentation of the concept of construct validity and its various aspects. The definition of CV performance proposed herein allows an assessment of the content validity of the various measures of CV performance. We suggest that the subjective measures capture the concept better than the “objective” ones. Next, we present a methodology to assess the empirical aspects of construct validity. We follow this methodology to assess the empirical validity of the CV performance measures using a sample of CVs involving companies based in Spain. After discussing the results, we conclude the paper by indicating its limitations and its contributions. (1) We define an inter-firm collaborative venture as an explicit agreement between two (or more) firms to collaborate in a limited aspect of their activity for a relatively long term, and which may or may not result in a separate organizational entity. Existing Measures of Collaborative Venture Performance Geringer and Hebert (1991) classify the measures of CV performance used in the literature into three main groups: financial, objective, and subjective measures. Financial measures include various measures of profitability (Good, 1972; Renforth, 1974; Dang, 1977), growth (Good, 1972; Dang, 1977), and cost position (Renforth, 1974). Frequently used objective measures include stability measures such as longevity of the venture (Harrigan, 1988 b; Kogut, 1989), ownership or contract stability (Franko, 1971; Gomes-Casseres, 1987; Blodgett, 1992), and survival (Franko, 1971; Raveed, 1976; Harrigan, 1988 b). The most commonly used subjective measure is an overall assessment of the firm's satisfaction with CV performance (Killing, 1983; Schaan, 1983; Beamish, 1984; Geringer and Hebert, 1991; Parkhe, 1993 a, b). Other subjective measures used include the degree of fulfillment of strategic goals that the venture was aimed at covering (Parkhe, 1993 a, b), and the net spillover effects of the venture on other activities of the firm (Parkhe, 1993 b). Construct Validity Construct validity represents “the correspondence between a construct (conceptual definition of a variable) and the operational procedure to measure [...] that construct” (Schwab, 1980: 5-6). Relationships among constructs are inferred from the observed relationships among their measures. Thus, the conclusiveness of these inferences depends upon the correspondence between constructs and measures (Schwab, 1980). Construct validity includes content validity, reliability, and convergent, discriminant, and criterionrelated validity of the measures of the construct. Content validity is “a qualitative type of validity where the domain of a concept is made clear and the analyst judges whether the measures fully represent that domain” (Bollen, 1989: 185). The domain of a concept is bounded by its theoretical definition, which should reflect the meanings associated with the concept in prior research (Bollen, 1989) and make its dimensions clear (Schwab, 1980; Bollen, 1989). Reliability is defined traditionally as the consistency of measurement. An alternative definition, appropriate from a structural perspective, is that the reliability of a measure is the magnitude of the direct relations that all variables (except the error terms) have on that measure (Bollen, 1989). Convergent validity is “the degree to which two or more attempts to measure the same concept [...] are in agreement” (Bagozzi and Phillips, 1982: 468). Measures of the same construct should present a large common variance. Convergence is especially important in areas in which there is a proliferation of measures presumably assessing the same construct (Schwab, 1980). Discriminant validity is “the degree to which measures of distinct concepts differ” (Bagozzi and Phillips, 1982: 469). Measures of different constructs should share little variance. The analysis of discriminant validity may help to corroborate issues of content validity when it is suspected that some measures actually correspond to another concept. Criterion-related validity is “the degree of correspondence between a measure and a criterion variable” (Bollen, 1989: 186). To evaluate it, a variable that is a standard against which to compare the measure in question is needed. Criterion validity may be concurrent –when the criterion exists at the same time as the measure– or predictive –when the criterion occurs in the future (Bollen, 1989). 2