مقایسه مفهومی و تجربی از سه مقیاس بازارگرایی

Although the topic of a market orientation has attracted considerable research attention, there still is no clear consensus on its definition and on how to measure it. The authors attempt to improve market orientation conceptualization and measurement by conceptually and empirically comparing three different scales of market orientation, the scales of Kohli and Jaworski, Narver and Slater and a newly developed extended market orientation (EMO) scale. Implications of the results and a future research agenda are also offered.

Although market orientation has attracted considerable research interest (e.g., Jaworski and Kohli, 1993, Kohli et al., 1993, Narver and Slater, 1990 and Slater and Narver, 1994), confusion still exists as to its definition, how to measure it (the construct of market orientation) and how it is developed (i.e., the antecedents to a market orientation). In this paper, our particular interest lies in the conceptual and measurement domains of market orientation. Although several different market orientation scales exist (e.g., Cadogan and Diamantopoulos, 1995, Kohli and Jaworski, 1990, Lichtenthal and Wilson, 1992, Narver and Slater, 1990 and Ruekert, 1992), there is no consensus on which is the better measure. As market orientation is considered a core concept of marketing (Narver and Slater, 1990), further conceptual and empirical investigation of existing scales is warranted. Therefore, the purpose of this paper is to conceptually and empirically (on psychometric grounds) compare an extended market orientation (EMO for notational purposes) scale with two existing market orientation constructs and scales. This is in the spirit of the comparative approach to theory testing (Sternthal et al., 1987) that argues a rigorous theory testing can be achieved by comparing the rival theoretical propositions. Because this approach demands a comparison be made on the grounds of a fit between the theoretical proposition and the empirical data, we first present how different market orientation scales are theoretically conceived and operationalized relative to other related constructs. We provide a conceptual review of market orientation, focusing on recent debate on construct definition and operationalization of market orientation scales. We then offer a model that organizes and reconciles two distinct market orientation conceptualizations, one behavioral and the other cultural. We introduce an EMO scale, which is built upon Kohli and Jaworski's (1990) conceptualization by specifically incorporating additional market factors into their scale. We compare the three measurement theories theoretically and empirically.

A written questionnaire, cover letter and a postage-paid return envelope were mailed, with two follow-up reminder mailings to nonrespondents. The three different versions experienced slightly different response rates. The response rates were 40.8%, 42.2% and 48.3% for Version A (EMO scale), Version B (KJMO) and Version C (NSMO), respectively. For each version, presence of nonresponse bias was examined by applying MANOVA and univariate F-tests to the seven performance variables. Based on these tests for all the versions, there was no indication that the responses to the performance measures were significantly different based on the number of mailings received before the respondent's actual response. In addition, the equivalency of the three different respondent groups (i.e., EMO, KJMO and NSMO) was examined by applying MANOVA to the seven performance variables across the three questionnaire types. Multivariate tests of significance (i.e., Pillai's Trace, Wilks' Λ and Hottelling's Trace) indicated there was no difference between the groups on these performance variables. The variance extracted (VE) at the first-order level was not satisfactory for any of the scales according to the well-accepted criterion of being greater than 0.5 (Fornell and Lacker, 1981). The VE ranged 0.21–0.37 for the EMO, 0.15–0.26 for the KJMO and 0.26–0.49 for the NSMO, suggesting a high level of measurement errors relative to the standardized loadings. The second-order confirmatory factor analysis for each scale produced reasonable general fit indices, producing, for example, CFI of between 0.97 (EMO) (Table 1) and 1.00 (NSMO) (Table 1). All three scales had VE at the second-order factor level greater than 0.5 (i.e., 0.59, 0.77 and 0.69 for EMO, KJMO and NSMO, respectively).A closer look at KJMO revealed that quite a few λ coefficient estimates (not standardized) in the IG (1 item) and ID (7 items) dimensions were found to be nonsignificant at the α=.05 level (i.e., the t statistic<1.96) due to a high level of standard errors of the estimates. In fact, the nonsignificant λ coefficient estimates led to a nonsignificant LISREL estimate of the path between ID and KJMO, and caution should be exercised in interpreting the standardized estimates of the KJMO scale (Table 1). The data suggest that the unidimensionality of the KJMO scale as a whole (in the second-order factor structure) is not supported without some further purification, which involves such procedures as item modifications and/or deletion particularly in the ID dimension. The second-order factor structures appear reasonable for both EMO and NSMO. Notably, the NSMO scale produced excellent fit statistics (χ2=50.29 at df=51, GFI=0.96, AGFI=0.94, PGFI=0.63, NFI=0.90, CFI=1.00) (Table 1). The EMO scale, on the other hand, produced a slightly lower though very good fit. At the first-order level, EMO's CFI ranged from 0.94 to 1.00, exceeding the well-accepted level of 0.90 (Bearden et al., 1982). The EMO second-order fit statistics (Table 1) were also good (χ2=227.65 at df=206, GFI=0.88, AGFI=0.85, PGFI=0.72, NFI=0.75, CFI=0.97) but not as good as the NSMO's. Note that the EMO scale contains 22 items while the NSMO scale contains only 12, which could account for the slightly lower fit of the EMO scale. Although both EMO and NSMO scales demonstrate adequate fit in the second-order factor structure, the reliability of the scales, however, is not extremely high for either scale. This is particularly notable given the number of items for each scale is relatively large, especially for EMO. The NSMO scale especially suffered from low Cronbach's α on two first-order dimensions (.38 for customer orientation and .45 for competitor orientation), while the interfunctional coordination dimension produced a reasonable α of .78. It was quite surprising that these reliability coefficients are substantially lower than those reported in NS's studies in the past. For example, in one of their studies (Slater and Narver, 1994), the reliability coefficients were reported as customer orientation (α=.878), competitor orientation (α=.726) and interfunctional coordination (α=.774). Although still relatively low, the EMO's first-order level reliabilities (IG=0.65, ID=0.75 and RESP=0.81) were slightly greater than those of both NSMO and KJMO (IG=0.61, ID=0.69 and RESP=0.81). To replicate the past performance-related studies of market orientation, a structural equation model was fit for each of the seven performance measures (dependent variables) separately with EMO and NSMO as independent variables. Since the KJMO scale's proposed second-order factorial structure was found untenable in spite of the adequate fit, the KJMO scale is not examined for its predictive validity here. Thus, one model equates EMO and the performance measures (EMO model), while the second equates NSMO and the performance variables (NSMO model). Each structural parameter was estimated using LISREL 8, and the overall fit for each model was estimated by χ2 goodness-of-fit statistics. The LISREL estimates for the β coefficients with regard to the dependent variables (ROA, ROI, ROS, SOM, SGRO, PCNTNP and OVERALL) are given in Table 2.The results show that both scales are positively related to all seven of the performance measures at the α=.05 level: a market orientation (measured in two different ways) was found to be positively correlated to the performance measures. Therefore, both EMO and NSMO scales were found to have predictive validity with regard to the performance outcome measures. The magnitude of the β coefficients is greater for the NSMO scale than the EMO scale for five of the seven performance measures. The R2's suggest that the EMO seems to explain the variation of the efficiency measures (i.e., ROI, ROA and ROI) as well as or better than the NSMO, while the NSMO explains the variation of the effectiveness (i.e., SOM, SGRO and PCTNP) and the overall measures better than the EMO. However, given the fewer number of items than the EMO scale, the NSMO scale is more efficient in predicting the performance measures.