Full Text

Standardization versus Customization in International Marketing: An Investigation Using Bridging Conjoint Analysis

Imad B. Baalbaki American University of Beirut

Naresh K. Malhotra Georgia Institute of Technology

We present and illustrate a methodology by which researchers can assess the relative importance and test the significance of various marketing-related factors as they influence the degree of standardization~customization of international marketing strategy. The standardization decision is viewed as a continuum with complete standardization and complete customization as the two extremes. Specific hypotheses related to the impact of marketing mix variables on the degree of standardization are formulated. These hypotheses are empirically investigated through a survey of international marketing managers. This investigation is carried out using conjoint analysis. Bridging methodology is introduced in order to accommodate the large number of variables in the study. The analysis is conducted at the individual level, at the group level, and at the aggregate level. Finally, we discuss the methodological and managerial implications of the findings and potential areas of future research.

International marketing managers periodically face the decision as to how much of their marketing strategy in one market applies to another. Standardization of marketing practices across markets is tempting because of potential economies of scale in production, promotion, distribution,

Journal of the Academy of Marketing Science. Volume 23, No. 3, pages 182-194. Copyright �9 1995 by Academy of Marketing Science.

and research and development. Standardization can also contribute to a coherent and consistent global image of the firm and its products. However, there are many obstacles to the application of uniform marketing policies. Vari- ations across markets in consumer attitudes, competitive environments, and marketing management related variables must be adequately assessed to insure the success of the product in a particular market (Baalbaki and Malhotra 1993).

The purpose of this study is to recommend a methodology and conduct an empirical investigation to assess the relative importance and significance of a large number of marketing management variables as they influence the standardization/customization of international marketing strategy. The standardization decision is viewed as a continuum with complete standardization and complete customization as the two extremes (Boddewyn, Soehl, and Picard 1986; Jain 1989; Sorenson and Wiechmann 1975; Whitelock and Chung 1989). At one extreme, complete standardization of marketing strategy means the offering of identical product lines and features at identical prices through identical distribution systems supported by identical promotional programs. At the other extreme, complete customization of marketing strategy means the development of distinctive tailor-made products, pricing, promotion, and distribution policies that have no standardized elements.

The article presents conjoint analysis as a means to empirically test hypotheses related to international marketing practices. Bridging methodology is introduced in order to accommodate the large number of variables in the study.

Baalbaki, Malhotra / STANDARDIZATION VS. CUSTOMIZATION 183

For the purpose of this study, certain marketing-related variables are selected out of a set of variables proposed by Baalbaki and Malhotra (1993), and 18 hypotheses related to the impact of these variables on the degree of standardization are presented. These hypotheses are empirically investigated by conducting a survey of international marketing managers. We conduct statistical analysis at the individual level, at the group level, and at the aggregate level. Methodological and managerial implications of the findings and potential areas of future research are then discussed.

THEORETICAL FRAMEWORK

The theoretical framework adopted is based on the recent work of Baalbaki and Malhotra (1993). They clas- sifted variables that influence the degree of standardization of international marketing strategy as either environmental or marketing management. Environmental variables include geographic, political, economic, and cultural variables. Market ing management variables include product-related, promotion-related, price-related, and distribution-related variables. Baalbaki and Malhotra (1993) explain that "these marketing management variables have a direct impact on the selection of the appropriate marketing mix strategy, and hence should be considered by the international marketing manager together with the environmental variables in formulating international marketing strategies" (p. 22). They also proposed several hypotheses that related these variables to the formulation of international marketing strategies and the level of standardization that is possible across international markets. The empirical investigation of these hypotheses was proposed as an area for future research.

This article focuses on a selected number of hypotheses pertaining to the various marketing management variables: product related, promotion related, price related, and distribution related. The rationale behind the development and formulation of these hypotheses was discussed in detail by Baalbaki and Malhotra (1993).

Product-related Hypotheses

HI: A greater degree of standardization of marketing strategy is possible for products that are perceived to be essential.

H2: Greater customization of marketing strategy is required for products that are in different life-cycle stages in different markets.

H3: Across-market differences in laws related to product standards, features, performance, and safety necessitate a higher level of strategy adaptation or customization.

H4: Intermarket differences in support requirements affect the acceptability and adoption of some products, thus increasing the need to adapt or customize a product's marketing strategy.

Promotion-related Hypotheses

H5: Across-market variations in the interpretation of an ad's theme, slogan, idiomatic expres- sion, words, symbols, and colors necessitate adaptation or customization in the marketing strategy.

H6: Differences in the availability and coverage of promotional infrastructure between markets call for a higher level of strategy customization.

H7: More customization of marketing strategy is needed in markets that differ significantly in terms of the relative importance customers assign to various media channels.

H8: A high level of strategy standardization is warranted in markets that are similar in terms of their laws and regulations related to promotional practices.

Price-related Hypotheses

H9:

H10:

H l l :

H12:

Greater customization of marketing strategy is needed in the presence of wide variations across markets in government pricing rules and regulations. The degree of risk associated with the volatility of exchange rates impedes the uniform- ity of pricing practices across markets and thereby lowers the degree of standardization of marketing strategy. Differences in consumers' price perceptions across markets hinder the possibility of uniform pricing and consequently lower the level of strategy standardization. Intermarket differences in the consumers' price elasticity of demand require an altera- tion in the pricing strategy across markets and thus necessitate more customization of marketing strategy.

Distribution-related Hypotheses

H13:

H14:

H15:

H16:

Markets that are similar in terms of their distribution infrastructure (as reflected in the availability, accessibility, complexity, and ef- fectiveness of the markets' distribution systems) are candidates for a standardized marketing approach. Between-market variations in government regulations and laws affecting distribution, such as those related to foreign ownership, licensing, and franchising, force the adoption of more customized marketing strategies. Differences in the geographic structure and dispersion between markets reduce the level of strategy standardization. Intermarket variations in certain social and cultural norms and preferences, such as shop-

184 JOURNAL OF THE ACADEMY OF MARKETING SCIENCE SUMMER 1995

ping habits, location preferences, and inven- tory policies, increase the need for strategy adaptation.

General Hypotheses

In addition to the above product-related, promotion- related, price-related, and distribution-related hypotheses, we selected two more general hypotheses based on Baalbaki and Malhotra (1993). These hypotheses relate to the effect of consumers' attitudes toward foreign products and the effect of the competitive environment on the level of standardization of international marketing strategy. The reason for the selection of those factors is that they play a unique role in the research design by serving as bridging factors.

H17:

H18:

Because attitudes toward foreign products in general, and toward the product's country of origin in particular, influence the consumers' perceptions and evaluations, variations in such perceptions between markets call for a greater strategy adaptation or customization. Differences in the level and intensity of industry competition between markets dimin- ish the level of standardization of marketing strategy.

RESEARCH METHODOLOGY

Conjoint Analysis

Conjoint analysis was used to assess the impact of the marketing management variables on the degree of standardization of international marketing strategy. Despite its increased popularity among practitioners and researchers, conjoint methodology has yet to be used in research related to aspects of international marketing strategy. The current study used an additive main effects conjoint analysis model. It is worth noting that the assumption of an additive compensatory model underlying the respondents' (in this case, managers') decision rule is reasonable. Research has shown that the compensatory model of conjoint analysis is likely to capture most of the predictable variance even when the actual decision rule follows a more complex compensatory or noncompensatory heuristic (Green and Srinivasan 1978, 1990; Messier and Emery 1980). Thus it is a very good approximation of reality (Green and Rao 1971). Moreover, Green and Srinivasan (1978) stated that the advantage of improved part-worth estimates resulting from the main effects design more than outweigh the disadvantage of reduced believability of the profiles. In addition, they argued that if the relative importance of the attributes is of interest (as in this study), then an orthogonal design produces less ambiguous results than a design incorporating interactions. In another study, Green (1984) revealed that incorporating interaction effects often results in a lower predictive validity. The gained realism of the model obtained by incorporating interaction terms is offset

by the deterioration in predictive accuracy caused by the inclusion of additional parameters.

Given the large number of factors of interest, one of the more important methodological considerations is the appropriate data collection method. In this study we preferred the full-profile method over the tradeoff (or the two- factor-at-a-time) method because it increased the realism of the experimental task and reduced the number of judg- ments required. Also, it was more flexible in supporting various measurement scales of the dependent variable (a crucial consideration because the dependent variable in this task was intervally scaled) (Green and Srinivasan 1978). However, the traditional full-profile method cannot incorporate a large number of variables without causing information overload on the respondents, even if a fractional factorial design or an orthogonal array is employed (Addelman 1962; Green 1974). Overwhelmed by the load of information, respondents attempt to simplify the experimental task by resorting to various simplifying tactics, thereby affecting the predictive validity of the part-worth estimates (Albaum 1989; Malhotra 1982a; Wright 1975).

Green and Srinivasan (1978, 1990) recommended con- fining the traditional full-profile procedure to a maximum of six factors at any particular sort, and resorting to the use of"bridging" designs to accommodate a larger number of factors. Bridging involves dividing the multitude of factors into several sets and creating a separate design for each set under the condition that some common factors exist in the different designs to enable the integration or bridging of their individual results at a later stage. Of course, bridging results tend to gain reliability with an increase in the number of matching factors. Bretton-Clark's BRIDGER software program can accommodate such a bridging technique by creating integrated design and utility files from the separate design and utility files corresponding to each separate design. Bridging factors should have an equiva- lent number of levels, carry identical level labels, and be analyzed using the same type of preference model in all designs (Albaum 1989). Moreover, BRIDGER can handle an unlimited number of factors, each having up to 20 levels, and is fully integrated with Bretton-Clark's two other conjoint programs, CONJOINT DESIGNER (Carmone 1986) and CONJOINT ANALYZER (Green 1987). It should be noted here that other approaches for dealing with large numbers of factors in conjoint studies have been suggested (e.g., in a recent article, Oppewal, Louviere, and Timmermann 1994 proposed such an approach). However, these alternative approaches are more complex to administer and apply.

In this study, each hypothesis was operationalized in terms of one variable representing a factor in conjoint analysis. Each factor was defined at two levels. Except for product essentialness, the levels were no difference and major difference. The two levels for essentialness were essential and nonessential. Four separate designs (Designs 1-4) were developed, each consisting of six factors as shown in Table 1. Although two factors (namely, attitudes toward foreign products and competitive environment-factors that correspond to the general hypotheses) were common across all four designs, each of the


TABLE 1 Individual-Level Analysis: Factor Significance

Factor Label

Percentage of Respondents with Significant Coefficients

Hypothesis p < .01 p < .05 p < .10

Design 1 Essentialness H 1 19.74 31.58 42.10 Attitude toward

foreign products H17 30.26 44.74 59.21 Media importance H7 22,37 ,32.89 43.42 Exchange risk H10 30.26 39.47 51,32 Distribution laws H14 18.42 30,26 42.11 Competitive environment H18 36.84 56.58 68.42

Design 2 Attitude toward

foreign products H17 21.05 36.84 43.42 Product laws H3 21.05 47.37 53.95 Message interpretation H5 14,47 26.32 32.89 Price perceptions Hl l 23.68 38.16 47.37 Geographic dispersion H15 14.47 27.63 36.84 Competitive environment H18 36.84 52,63 71,05


foreign products H17 30.26 53.95 60.53 Product support requirements H4 34.21 52.63 59,21 Promotion infrastructure H6 31.58 51.32 59.21 Pricing laws H9 30.26 47.37 53.95 Shopping habits H16 32.89 52.63 69.74 Competitive environment H18 53.95 68.42 80.26


foreign products H17 27.63 42.11 53.95 Product life cycle H2 27.63 40.79 47.37 Promotion laws H8 23.68 36.84 47,37 Price elasticity H12 23.68 39.47 53.95 Distribution infrastructure H13 32.89 38.16 55.26 Competitive environment H18 44.74 56.58 69.74

remaining four factors addressed a particular aspect of the marketing mix. Thus a product-related variable, a promotion-related variable, a price-related variable, and a distribution-related variable was present in each of the designs. This enhanced the realism of the designs. Having all factors defined on the same number of levels, as in our study, is encouraged in conjoint literature. It is argued that the relative importance of a factor may be positively associated with an increase in the number of levels defining that factor. This association impairs the comparability of the relative importance of factors that are explicated along different numbers of levels (Currim, Weinberg, and Wittink 1981; Wittink, Krishnamurthi, and Nutter 1982).

Population Definition and Sample Selection

The population under study was defined as all U.S.- based firms with international operation and orientation. Thus any firm having a representative office in the United States and currently involved in international business (whether a truly multinational company with offices, plants, and operations in several countries, or a mere

exporter of domestic-made products) was included in the population. The following sources and directories were used to define the sampling frame: Directory of American Firms Operating in Foreign Countries, Principal Interna- tional Business, and Ward's Business Directory of U.S. Private Companies.

The Directory of American Firms Operating in Foreign Countries was the primary source of sample selection. In the sample, we included U.S. manufacturing companies from this directory that were cross-listed with the other two directories. This procedure was employed to ensure the availability of sufficient background information about any selected company. Background information gathered from these sources included number of employees, sales volume, year established, industry affiliation, and primary product category (industrial or consumer). This information was collected in order to assess nonresponse bias.

A total of 816 U.S. manufacturing firms were identified. Of the selected companies, 515 dealt mainly in industrial goods whereas 301 dealt mainly in consumer goods. We excluded U.S.-based international firms dealing in ser- vices and commodities because of the unique business and marketing environment in which they operate.

Questionnaire Design

The questionnaire consisted of two major sections. First, the purpose of the study and the definitions of complete standardization and complete customization were presented. The respondent indicated the product category (industrial/consumer) with which he or she was more familiar in marketing internationally. The respondent was asked to complete the entire questionnaire for that specific product category. In section 1, respondents examined all 18 factors involved in the study and rated each of the factors on a 7-point scale according to its importance in deciding on the degree of standardization/customization of marketing strategy.

Throughout section 2, the survey instructed the respondent to assume that his or her firm was considering entering and marketing a specific product in several foreign markets. The respondent then examined four sets of hypothetical market situations or profiles, each set of profiles corresponding to one of the four designs (Table 1). Each set contained 12 hypothetical situations or profiles (8 profiles for estimation of model parameters and 4 for assessing the predictive validity of the model). The 8 estimation profiles were selected according to a fractional factorial design. Each situation, described as a profile, provided a comparison of the markets along six marketing-related conditions or factors. As described in an earlier section, two conditions (namely, customers' attitudes toward foreign products and competitive environment) were common across all four sets. Based on the information presented, the respondent rated each situation using a 7-point scale (1 = complete customization, 7 = complete standardization) according to the degree of standardization of marketing strategy that he or she believed as being most appropriate across markets.


The selection of a 7-point scale conforms to the broad guidelines for the number of scale categories suggested in the literature (Malhotra 1993). In order to reduce ambigu- ity on the part of respondents and insure a consistent understanding of the rating scale among all respondents, the scale was supplemented with an explanation of the two anchor points: complete standardization of marketing strategy and complete customization of marketing strategy. Respondents were explicitly advised to regard those conditions that were not included in a particular situation as having exactly the same impact across the markets. To insure that the task in section 2 was clear, an illustrative situation was also presented. This was followed by the four sets of hypothetical situations (profiles). The order of the presentation of the four conjoint designs was randomized across respondents to control for any order effects.

In the pretest phase, local business executives with adequate international marketing experience were ap- proached. Their feedback and comments were instrumen- tal in determining the order of appearance of the questionnaire sections and supplying adequate instruc- tions to facilitate the survey, especially in clarifying the rating task in the conjoint section (i.e., section 2).

Survey Implementation

The questionnaire was addressed to a specific top-level company executive responsible for the company's international operations. In instances where such information was lacking, the questionnaire was addressed to a top-level executive (president or CEO), and the executive was explicitly asked to forward the questionnaire to a manager with adequate experience in international marketing.

We mailed a prenotification letter introducing the survey and highlighting its potential to provide a better understanding of international marketing. The questionnaires were mailed one week later. The 9 in. • 12 in. clasp envelope included the questionnaire along with a cover letter that stated the purpose of the study and its significant managerial and academic contributions and encouraged management's participation. Respondents were assured that their responses would be held in strict confidence, and were offered a summary of the results. A postage prepaid 6 in. • 9 in. return envelope was also enclosed. Three weeks later a follow-up postcard reminder was mailed to those firms whose responses had not been received. A final follow-up questionnaire was mailed after another 3 weeks to those failing to respond. In general, the survey approach followed Dillman's (1978) recommendations.

Response Rate and Nonresponse Bias

Of the 816 companies that were targeted in the study, only 724 companies actually received the questionnaires. A total of 92 questionnaires were returned because of "moved/not forwardable" or "forwarding order expired" reasons. An additional 42 questionnaires were returned marked "delete from mailing list/person no longer at company." This reduced the actual sample size to 682 compa-

nies. Of these companies, 128 (a response rate of 18.77%) responded to the survey. However, 48 companies declined to participate in the study for a variety of reasons (against company policy, survey not suitable for company's line of business, firm no longer operated in the international market, firm filed for bankruptcy). In effect, we received 80 questionnaires (46 from industrial firms and 34 from consumer firms) for an 11.73% response rate. One questionnaire came in late and was not included in the analysis, and 5 questionnaires were unusable. Thus 74 questionnaires were retained for data analysis. It is important to note that the response rate is realistic considering the length and complexity of the questionnaire. Moreover, and in order to statistically investigate the presence of nonresponse bias, further analyses were performed.

We conducted a series of t tests to assess the extent of nonresponse bias with respect to the following background variables gathered from the secondary data sources: number of employees, sales volume, and year established. No significant difference was detected between firms that responded and those that did not. The next section details the statistical analyses of the responses and discusses the findings in detail.

STATISTICAL ANALYSIS

To conduct a more rigorous examination of the hypotheses, we conducted the analyses at the individual, aggregate, and group levels (Moore 1980). Two different approaches were used to group the respondents. One was an a priori segmentation scheme in which the managers were classified into industrial goods or consumer goods segments based on self-reported data. The consumer segment was further split into consumer durable and nondurable segments. The other approach to grouping was based on clustering respondents based on their individual-level part-worths.

Individual Analysis

At the individual level, a manager's ratings on the hypothetical situations in each of the four designs were analyzed using CONJOINT ANALYZER. Each design was analyzed separately, thereby obtaining the part-worths for each of the six factors in the design, and consequently the relative importance of each factor. BRIDGER was then employed to bridge the four separate designs to arrive at an overall master design. Because BRIDGER can only handle two designs at a time, two levels of bridging were necessary. Design 1 and Design 2 were bridged yielding combined design and utility files. Design 3 and Design 4 were bridged in the same fashion. Then, another level of bridging was done to combine the bridged designs and utility files of Designs 1 and 2 with those of 3 and 4. The final utility and design files, which were labeled as master utility and design files, were analyzed using CONJOINT ANALYZER to reveal the part-worth of each factor level and the relative importance of all 18 factors. We conducted


TABLE 2 Pooled Analysis: Bridged/Master Design

Aggregate

Factor Label Hypothesis Part- Worth Importance

Industrial Consumer

Part-Worth Importance Part-Worth Importance

Essentialness H1 +0.318 3.98% Attitude toward foreign products H17 +0.511 6.40% Media importance H7 +0.382 4.78% Exchange risk H10 +0.463 5.80% Distribution laws H14 +0.389 4.87% Competitive environment HI 8 +0.707 8.85% Product laws H3 " +0.567 7.10% Message interpretation H5 +0.269 3.37% Price perceptions HI 1 +0.512 6.41% Geographic dispersion H15 +0.433 5.43% Product support H4 +0.548 6.87% Promotion infrastructure H6 +0.438 5.48% Pricing laws H9 +0.369 4.62% Shopping habits H16 +0.483 6.05% Product life cycle H2 +0.427 5.35% Promotion laws H8 +0.350 4.38% Price elasticity H 12 +0.403 5.05% Distribution infrastructure H13 +0.417 5.22%

+0.265 3.09% +0.391 5.45% +0.563 6.58% +0.438 6.10% +0.352 4.11% +0.423 5.90% +0.462 5.40% +0.464 6.46% +0.369 4.31% +0.415 5.79% +0.802 9.36% +0.575 8.01% +0.622 7.26% +0.491 6.83% +0.257 2.99% +0.286 3.99% +0.603 7.04% +0.384 5.35% +0.439 5.12% +0.426 5.94% +0.635 7.41% +0.429 5.98% +0.435 5.08% +0.441 6.15% +0.413 4.83% +0.307 4.27% +0.518 6.05% +0.435 6.06% +0.507 5.92% +0.317 4.41% +0.349 4.07% +0.351 4.90% +0.467 5.45% +0.315 4.39% +0.509 5.94% +0.290 4.03%

this analysis on each of the 74 usable questionnaires, producing a total of 74 individual master utility files.

Because CONJOINT ANALYZER falls short of supplying the researcher with the significance of each of the factors, it was necessary to perform dummy variable regression on each of the four designs for each respondent. Because each factor was varied at only two levels, a factor level's part-worth was equal to half that factor's regression coefficient. A t statistic was computed for each factor and the correspondingp value indicated the significance of that factor. Table 1 summarizes the findings of the tests of factor significance at the individual level. Taking each design at a time, the table lists the percentage of respondents having a significant coefficient for each of the factors at the .01, .05, and the. 10 significance levels. These results are encouraging when it is realized that (a) only eight observations were available for estimation of the parameters, and (b) not all of the six factors can be expected to be significant for any given respondent.

Aggregate-Level Analysis

Analysis at the overall aggregate level resulted in a single conjoint utility function for all 74 respondents. In order to arrive at such a function, respondents' ratings on each hypothetical situation for each design were pooled to create a single data file for that design. The resulting four data files corresponding to the four designs were then fed into CONJOINT ANALYZER one at a time to estimate the relative importance and part-worths of the factors and their levels. Then, a two-stage bridging procedure was employed, first bridging Design 1 with 2 and 3 with 4, and subsequently bridging those intermediate-level designs to arrive at the aggregate master design and utility files. Table 2 lists the relative importance and part-worths of all 18 factors in the master design. The part-worths shown in the

table are those of the no-difference level of each of the factors (the part-worths for the major-difference level are equal in magnitude but are of opposite sign).

We performed dummy variable regression to determine the significance of the factors in each design. Table 3 summarizes the findings. For each design, the model fit was assessed with respect to the significance of its R 2 value. All models were significant at the .001 level. Fur- thermore, tests of significance were computed for each of the six factors in each design. All factors were significant at the .001 level. The signs of the part-worths were also examined to assess the direction of the impact of a factor on the degree of standardization of international marketing strategy. These results were consistent for all the factors in the four designs: positive part-worths were associated with the no-difference level of a factor, whereas negative part- worths were associated with the major-difference level of a factor. This further confirmed the hypotheses because a higher degree of standardization of marketing strategy was hypothesized to be plausible if markets were more similar in terms of any of the depicted factors.

A Priori Segmentation

Analysis of Industrial Segment The industrial segment comprised 43 respondents. We

combined the managers' ratings for each design to create a single data file. A similar analysis to the one explained in the preceding section was followed. The relative importance and the utilities of the factors and the factor levels are tabulated in Table 2. The significance of each design and each factor within a design are tabulated in Table 4. As can be inferred from Table 4, all 18 factors were found to have a significant impact on the standardization/customization decision that managers of industrial products have to face, thus supporting all 18 hypotheses.


TABLE 3 Pooled Analysis: Aggregate Model Factor Significance

Part- Worth

Factor Label Importance No Difference Major Difference p Hypothesis Conclusion

Design 1 (R 2 = .606, R = .778) Essentialness 11.02% +0.318 -0.318 .001 Attitude toward foreign products 19.70% +0.568 -0.568 .001 Media importance 13.25% +0.382 -0.382 .001 Exchange risk 16.06% +0.463 -0.463 .001 Distribution laws 13.48% +0.389 -0.389 .001 Competitive environment 26.49% ' +0.764 -0.764 .001

Design 2 (R 2 = .632, R = .795) Attitude toward foreign products 17.26% +0.490 -0.490 .001 Product laws 18.21% +0.517 -0.517 .001 Message interpretation 9.17% +0.260 -0.260 .001 Price perception 17.02% +0.483 -0.483 .001 Geographic dispersion 12.98% +0.368 -0.368 .001 Competitive environment 25.36% +0.720 -0.720 .001

Design 3 (R 2 = .636, R = .797) Attitude toward foreign products 16.02% +0.471 -0.471 .001 Product support 16.82% +0.495 -0.495 .001 Promotion infrastructure 14.06% +0.414 -0.414 .001 Pricing laws 13.03% +0.383 -0.383 .001 Shopping habits 15.44% +0.454 -0.454 .001 Competitive environment 24.63% +0.725 -0.725 .001

Design 4 (R 2 = .653, R = .808) Attitude toward foreign products 17.12% +0.492 -0.492 .001 Product life cycle 13.94% +0.400 -0.400 .001 Promotion laws 12.65% +0.363 -0.363 .001 Price elasticity 14.41% - +0.414 -0.414 .001 Distribution infrastructure 15.82% +0.454 -0.454 .001 Competitive environment 26.06% +0.748 -0.748 .001

H1 Supported H17 Supported H7 Supported H10 Supported H 14 Supported H 18 Supported

H17 Supported H3 Supported H5 Supported H 11 Supported H15 Supported H 18 Supported

H 17 Supported H4 Supported H6 Supported H9 Supported H16 Supported H 18 Supported

H17 Supported H2 Supported H8 Supported H12 Supported H13 Supported H18 Supported

NOTE: All designs significant atp < .001.

Analysis of Consumer Segment

The consumer segment consisted of 31 managers. Analysis of the consumer segment using the conjoint packages and dummy variable regression supported the hypotheses. These results are recorded in Tables 2 and 4.

We compared the consumer segment to the industrial segment in terms of the segments ' part-worths using two- group t tests. The two segments were found to be signifi- can t ly d i f fe ren t in the impor t ance p laced on three variables. Product 's essentialness was more important to the consumer segment than to the industrial segment in deciding on the degree of standardization of international marketing strategy between markets (p = . 10). Competi t ive environment and price elasticity were more important to the industrial segment than to the consumer segment (p = .02 and p = .08, respectively). Note that the consumer segment was further divided into consumer nondurable (20 managers) and consumer durable (11 managers) segments to investigate any inherent differences in the internat ional or ientat ion o f managers be longing to these segments. For both segments, all of the factors were found to be significant when it came to decisions regarding the level of standardization of marketing strategy.

Clustered Segmentation

An attempt was made to develop clusters homogeneous in terms of the master design part-worth utilities of the 18 variables. Given the large number of variables relative to the size of the sample, principal component analysis was applied to reduce the number of variables. The Kaiser- Meyer-Olkin measure of sampling adequacy (KMO = .822) and the Bartlett test of sphericity (p = .000) indicated that factor analysis was appropriate. Four resulting factors were extracted. The factors cumulatively explained 66.4% of the variance. Respondents ' factor scores were then cluster analyzed using Ward's method of hierarchical clustering. Two clusters emerged with 36 and 37 respondents, respectively, and one respondent was atypical.

A series of t tests were then conducted to examine any significant differences between the two clusters on the 18 variables. We found the clusters to have significantly different part-worths on five var iab les - -namely essentialness (p = .000), media importance (p = .000), exchange rate risk (p = .000), product support (p = .064), and promotion laws (19 = .026). Product-related variables of product essentialness and the availabili ty of product sup-


TABLE 4 Pooled Analysis:

A Priori Segments Factor Significance

Factor Label Importance Part-Worth p < Hypothesis

Industrial Segment Factor Significance

Design 1 (R 2 = .620) Essentialness 9.27% +0.265 .001 HI Attitude toward

foreign products 20.47% +0.584 .001 H17 Media importance 12.32% +0.352 .001 H7 Exchange risk 16.19% +0.462 .001 H10 Distribution laws 12.93% +0.369., .001 H14 Competitive environment 28.82% +0.823 .001 H18

Design 2 (R 2 = .681) Attitude toward

foreign products 17.27% +0.500 .001 H17 Product laws 17.47% +0.506 .001 H3 Message interpretation 8.63% +0.250 .001 H5 Price perception 18.07% +0.523 .001 Hll Geographic dispersion 12 .25% +0.355 .001 H15 Competitive environment 26.31% +0.762 .001 H18


foreign products 15.48% +0.468 .001 H17 Product support 18.17% +0.549 .001 H4 Promotion infrastructure 13.37% +0,404 .001 H6 Pricing laws 12.98% +0.392 .001 H9 Shopping habits 15.67% +0.474 .001 H16 Competitive environment 24.33% +0.735 .001 H18


foreign products 18.05% +0.517 .001 HI7 Product life cycle 14.60% +0.419 .001 H2 Promotion laws 11.76% +0.337 .001 H8 Price elasticity 15.21% +0.436 .001 H12 Distribution infrastructure 14.20% +0.407 ,001 H13 Competitive environment 26.17% +0.750 .001 H18

Consumer Segment Factor Significance

Design 1 (R 2 = .600) Essentialness 13.40% +0.391 .001 H1 Attitude toward

foreign products 18.65% +0.544 .001 H17 Media importance 14.50% +0.423 .001 H7 Exchange risk 15.88% +0.464 .001 H10 Distribution laws 14.23% +0.415 .001 H14 Competitive environment 23.34% +0.681 .001 H18


foreign products 17.25% +0.476 .001 H17 Product laws 19.30% +0.532 .001 H3 Message interpretation 9.94% +0.274 .001 H5 Price perception 15.50% +0.427 .001 H 11 Geographic dispersion 14 .04% +0.387 .001 H15 Competitive environment 23.98% +0.661 .001 H18


foreign products 16.81% +0.476 .001 H17 Product support 14.81% +0.419 .001 H4 Promotion infrastructure 15.10% +0.427 .001 H6 Pricing laws 13.11% +0.371 .001 H9 Shopping habits 15A0% +0.427 .001 H16 Competitive environment 25.07% +0.710 .001 H18


foreign products 15.83% +0.456 .001 H17 Product life cycle 13.03% +0.375 .001 H2 Promotion laws 13.87% +0.399 .001 H8 Price elasticity 13.31% +0.383 .001 H12 Distribution infrastructure 18.07% +0.520 .001 H13 Competitive environment 25.91% +0.746 .001 H18

port requirements were more important to the second cluster, whereas the price-related variable concerning the volatility of exchange rates and the promotion-related variables related to the markets ' relative importance of media channels and promotion laws were more important for the first cluster. Such a finding signals the presence of two streams when managing internat ional market ing plans. Although both streams acknowledge the importance of all the variables in affecting the international marketing strategy, one stream is found to be relatively more driven by the product it offers and the product-related variables in the foreign market. On the other hand, the second stream keeps a relatively close eye on some price-related and promotion-related variables.

Conjoint analysis was run on the two segments. Al- though these results are not shown due to lack of space, all 18 hypotheses were supported, as in the aggregate and a priori segmentation analysis.

VALIDITY, RELIABILITY, AND CONSISTENCY

Validity A s s e s s m e n t

We used the following validity tests to assess the validity of the conjoint models developed in the preceding analyses.

Internal Validity,

Internal validity tests the goodness of fit o f the conjoint model based on the estimated part-worths. Internal validity between the respondents ' input ratings and the model ' s estimated ratings could be assessed through Pearson's product-moment correlation, or alternatively through R 2, the model ' s coefficient of determination. The coefficients of determination for the four designs in the aggregate model (.606, .632, .636, .653), in the industrial segment model (.620, .681, .661, .683), in the consd~e r segment model (.600, .571, .609, .626), in the consumer durable segment model (.627, .649, .740, .735), and in the consumer nondurable segment model (.599, .566, .567, .592), were high and all were significant at the .001 level.

Cross or Predictive Validity

We conducted a cross-validation test to assess the pre- dict ive capabil i t ies of the est imated model. For each model, ratings on a hold-out set o f profiles were calculated based on the model 's estimated part-worths. The estimated ratings' correlation to the respondents ' input ratings reflected the predictive power of each model. These correlations and their respective significance levels are arranged in Table 5. All models reflect a high degree of predictive validity as the correlation between the estimated and the input ratings for the hold-out situations in each design were significant at either the .01 or the .05 levels. 1

Convergent Validity

Convergent validity could be evaluated by assessing the degree to which the estimated factor importance conforms

190 JOURNAL OF THE ACADEMY OF MARKETING SCIENCE

TABLE 5 Predictive Validity of Conjoint Models

Design Correlation p <

Aggregate Model (N = 74) Design 1 0.989 ,05 Design 2 0.987 .05 Design 3 0.997 .01 Design 4 1.000 .01

Industrial Model (N = 43) Design 1 0.997 .01 Design 2 0.987 , .05 Design 3 0.999 .01 Design 4 0.998 .01

Consumer Model (N = 31) Design 1 0.968 .05 Design 2 0,981 .05 Design 3 0.994 .01 Design 4 0.996 .01

to the respondents' self-reported importance. As part of the survey (specifically in section 1 of the questionnaire), respondents were asked to supply their subjective importance ratings associated with each of the factors reflecting the factor's impact on the degree of standardization of international marketing strategy. We used Pearson's product-moment correlation between a respondent's self- reported importance and the relative importance estimated in that respondent's master or bridged design as an indica- tor of convergent validity. Almost half the respondents (49.3%) had significant correlations at the .05 level between their self-explicated importance ratings and the relative importance portrayed in the bridged model. Al- most 90% (88.7%) of the correlations were significant at the. 10 level. Such results are encouraging considering that a respondent's ratings in the conjoint task underwent two stages of bridging before arriving at the respondent's master design.

Reliability Assessment

Conjoint results are usually assessed in terms of their reliability over changes in the designs. Reliability over context (or reliability over attribute set) shows the extent to which the part-worths for a certain factor (or set of factors) are stable as other factors in the stimulus profile are varied. Thus assessing reliability over context could involve adding some new factors to the original design (embedding analysis), or modifying the original design by deleting a few factors and substituting an equal number (structural analysis), and then reestimating the part-worths to check for any instability. Bateson, Reibstein, and Bould- ing (1987) regard such a test of reliability as a test of independence of effects assumption that characterizes the additive model of conjoint methodology. Conjoint analysis is acclaimed for its robustness over different embedding and structural contexts (Malhotra 1982b; McCullough and Best 1979; Scott and Wright 1976).

We checked the part-worths of the shared or bridging factors for stability across the different designs in a manner

SUMMER 1995

TABLE 6 Structural Reliability of Bridging Factors

Design 1 Design 2 Design 3 Design 4

Bridging Factor: Attitudes toward Foreign Products

Design 1 p = .000 p = .000 p = .000 Design 2 0.556 p = .000 p = .000 Design 3 0.406 0.635 p = .000 Design 4 0.451 0.717 0.667

Bridging Factor: Competitive Environment

Design 1 p = .000 p = .000 p = .000 Design 2 0.405 p = .000 p = .000 Design 3 0.570 0.426 p = .000 Design 4 0.426 0.485 0.607

NOTE: Lower-half triangular matrix lists correlations, upper-half triangular matrix lists significance.

parallel to that of structural analysis (i.e., instead of regarding the four designs as different designs sharing few common factors, one can regard them as basically the same design but with four factors deleted and substituted for by an equal number). The correlations between the part- worths of the bridging factors were computed for all possible pairs across the four designs. As Table 6 reflects, the part-worths of both bridging factors (attitudes toward foreign products and competitive environment) proved to be stable across all designs, confirming the findings of the earlier studies.

Consistency Checks

In addition to conducting validity and reliability tests, Scott and Wright (1976) recommended the use of various consistency checks as means of gaining insight into whether the estimated part-worths make sense. Consis- tency checks include analyzing and comparing the values and signs of the estimated parameters and checking whether these parameters conform to a priori theory or belief. As reported earlier, the signs of the part-worths were observed throughout all levels of the analysis to determine the direction of the impact of a factor on the level of international marketing standardization. As expected, negative significant part-worths were found to be associated with the major-difference level of factors, whereas positive significant part-worths were found to be associated with the no-difference level of factors.

CONCLUSIONS AND RECOMMENDATIONS

Conclusions Regarding Hypotheses

A main objective of this study was to investigate a number of hypotheses related to the level of standardization/customization of marketing strategy possible across markets. A total of 18 factors were selected from a list of factors extracted through a comprehensive review of the international market segmentation and strategy literature.


At the individual level, not all factors were found to be of significance for any particular manager. However, each of the 18 factors was important to a large percentage of managers. This result was further reinforced through aggregate and segment-level analyses, and all 18 hypotheses were supported.

Thus whenever a marketing manager undertakes the responsibility of marketing a product in the international market, he or she should be aware of these factors and their implications on marketing strategy. Following a standardized marketing approach in all foreign markets might be tempting given the cost advantages it carries through the various economies of scale, and the consistent global image it grants the product. However, variations across markets in terms of the 18 factors studied in this project indicate that a fully standardized marketing approach may not be appropriate. Adjusting the marketing strategy such that it takes into account these variations enhances the product's chance of success. As Baalbaki and Malhotra (1993) proposed, such factors could also be considered in segmenting a firm's global market. The resulting segments would then reflect a high degree of internal homogeneity, prompting the use of a highly standardized marketing approach in markets that belong to a single segment.

In addition to the above conclusions, the comparisons between the industrial and consumer segments proved insightful. Managers of consumer goods were found to regard some factors differently than industrial managers. Thus the nature of a product to be marketed internationally has significant implications for the particular variables that a manager must consider in deciding on the extent of standardization.

Clustering respondents based on their part-worths re- suited in two clusters. Managers in these clusters were characterized by different managerial orientations. Al- though both managerial orientations acknowledge the importance of all variables under study in affecting the international marketing strategy, some differences emerged mainly with regard to the relative importance of promotion- and product-related variables.

Indeed, the 18 marketing-related variables under investigation proved to be significant in all segmentation schemes, signaling the need for a very careful multifaceted analysis of foreign markets before international marketing strategies are extended, developed, and/or implemented. Nonetheless, a careful look at how the variables rank in terms of their relative importance (Table 7) reveal some interesting findings. Competitive environment was consis- tently regarded as the most important variable affecting the extent of standardization of international marketing strategy. In all analyses, whether aggregate-level analysis, a priori segmentation analyses, or clustered segmentation analysis, the level and intensity of industry competition ranked first in terms of its impact on the degree of possible standardization of marketing strategy across world markets. Moreover, and upon comparison between the four groups of marketing variables (with each corresponding to an element of the marketing mix), we found that international marketing managers (whether taken in aggregate, or

whether taken as segments of industrial, consumer durable, and consumer nondurable managers) tended to consider product-related variables as having the largest impact on the degree of standardization of marketing strategy.

Also, it can be noted that the consumer nondurable segment revealed the smallest variation in terms of the relative importance of the variables (range = 2.92%) as compared to the industrial goods segment (range = 6.37 %) or the consumer durable segment (range = 8.36%). This finding is explained by the fact that nondurable goods are the most affected by a wide variety of market-specific forces influencing the marketing environment. Hence there are no few sets of variables that predominantly affect the marketing strategy for such goods. Rather, consumer nondurable marketing managers engage in a thorough analysis of a market's environment, carefully taking into account the effect of a multitude of marketing-related factors on international marketing strategy.

Conclusions Regarding Bridging Methodology

Another major contribution of this study is the use of the bridging procedure. Despite the recommendations by conjoint methodology experts to resort to a bridging procedure to accommodate large numbers of factors in conjoint studies (Albaum 1989; Green and Srinivasan 1978, 1990), we did not find any published research that used the bridging technique. This article offers the first application of such a technique (at least in a research environment). Thus we feel it necessary to devote some attention to the conclusions drawn from the application of the bridging technique.

The main advantage in using bridging methodology is that it draws on the strengths of conjoint analysis while, at the same time, handles the limitation on the maximum number of factors that characterizes the traditional full- profile conjoint study. Green and Srinivasan (1978, 1990) advocated restricting the number of factors used in the traditional full-profile conjoint procedure to at most six factors. Having more factors raises a concern about information overload on the respondents, even when fractional factorial designs are employed. The usefulness of responses becomes questionable in such situations. Bridging methodology was developed to handle situations involving a large number of factors. As was apparent in this study, the bridging technique enabled the testing of 18 factors on the degree of standardization.

The tests of reliability and validity that were conducted with satisfactory results shed light on the efficacy of the bridging methodology. As was portrayed in Table 6, the bridging procedure proved to be reliable to changes in the stimulus profile, reflecting high structural reliability. The part-worths of the bridging factors were stable across the four designs. Moreover, the bridging methodology also reflected high convergent validity, even after the respondents' ratings underwent two stages of bridging to arrive at the master design. The factor importance estimated

o e.,

0 e~

e,,

. 2

~ ~ . ~ �9 ~.~~.~ ~ ~.~ ~.~ ~~

o

e~=~ ~

:~ .. i ~_ ~ ~

~.~ ~'~

o

r..)

.a

~.~ ~.~ ~.~ . ' ~ ~ .~

Q

1 9 2


through bridging the four individual designs into a single master design correlated significantly with the respondents' self-reported importance.

Contributions and Recommendations

In late 1986, Boddewyn, Soehl, and Picard commented on the paucity of systematic research dealing with the standardization of international marketing practices. Kaynak (1984) expressed a similar conclusion. Our methodological review of studies in this area, which we conducted for the purpose of this project, s~upported these researchers' concerns. Despite the growing interest of aca- demia in the field of international marketing, methodologi- cally sound research studies are lacking. This study offers a methodology that attempted to test and empirically verify various hypotheses depicting the impact of a large number of marketing-related variables on the degree of standardization of international marketing strategy. The bulk of these hypotheses had never been tested, although some methodological concerns could be raised regarding the manner with which the other few were investigated. Once researchers provide evidence on the significance of marketing-related variables, they can then be used for segmentation of the international marketplace.

The research also introduced conjoint analysis as a means of investigating the impact of marketing management variables on the degree of standardization of international marketing strategy. None of the past studies had used conjoint methodology in the study of international marketing strategy. Another contribution that transcends the field of international marketing lies in the use of bridging tech- niques. Several researchers have recommended the use of bridging methods to overcome the information overload problem associated with having a large set of factors. However, the review of published literature on conjoint analysis failed to locate any study in any field that took advantage of such a methodology.

In addition to the aforementioned contributions and recommendations, this study opens various venues of future research opportunities. The article examined the effects of 18 marketing management factors on the level of standardization of international marketing strategy. Addi- tional research should be done to empirically test the importance of the remaining marketing management factors and the environmental factors in shaping international marketing strategy (Baalbaki and Malhotra 1993). Another potential area of research is the verification of the proposed hypotheses in a cross-cultural context. The study lays solid foundations for a sound methodology that could be applied under various scenarios. The results of the present study were based on a sample composed entirely of U.S. manufacturing companies with international operations. A similar research design and methodology could be applied on a sample of foreign companies. The results of such a study could then be compared with the present research findings to uncover any particular managerial orientations that dif- ferentiate companies of various nationalities. We hope that our article will spark more research in this area.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the provision of the bridging software by Mr. Steve Herman, president, Bretton-Clark.

NOTE

1. Note that the higher correlation coefficients arrived at when assessing predictive validity (Table 5) as compared to those calculated upon evaluating intemal validity (Tables 3 and 4) are attributed to the difference in the number of observations upon which the correlations were calculated.

REFERENCES

Addelman, S. 1962. "Orthogonal Main-Effect Plans for Asymmetrical Factorial Experiments." Technometrics 4 (February): 21-46.

Abaum, Gerald. 1989. "BRIDGER." Journal of Marketing Research 26 (November): 486-7.

Baalbaki, lmad B. and Naresh K�9 Malhotra. 1993. "Marketing Manage- ment Bases for International Market Segmentation: An Altemate Look at the Standardization/Customization Debate." International Market- ing Review 10 (l): 20-45.

Bateson, John E. G., David Reibstein, and William Boulding. 1987. "Conjoint Analysis Reliability and Validity: A Framework for Future Research." In Review of Marketing. Ed. M. Houston. Chicago: Ameri- can Marketing Association, 451-81.

Boddewyn, J. J., Robin Soehl, and Jacques Picard. 1986. "Standardiza- tion in International Marketing: Is Ted Levitt in Fact Right?" Business Horizons 29 (November-December): 69-75.

Carmone, Frank J. 1986. "CONJOINT DESIGNER." Journal of Market- ing Research 23 (August): 211-312�9

Currim, Imran S., Charles B. Weinberg, and Dick R. Wittink. 1981. "The Design of Subscription Programs for a Performing Arts Series: Issues in Applying Conjoint Analysis." Journal of Consumer Research 8 (June): 67-75.

Dillman, Don A. 1978. Mail and Telephone Surveys: The Total Design Method. New York: Wiley.

Directory of American Firms Operating in Foreign Countries. 1991. New York: Uniwodd Business Publication.

Green, Paul E. 1974. "On the Design of Choice Experiments Involving Multifactor Alternatives." Journal of Consumer Research 1 (September): 61-8.

. 1984. "Hybrid Models for Conjoint Analysis: An Expository Review." Journal of Marketing Research 21 (May): 155-9.

�9 1987. "CONJOINT ANALYZER." Journal of Marketing Re- search 24 (August): 327-9.

Green, Paul E. and Vithala R. Rao. 1971. "Conjoint Measurement for Quantifying Judgmental Data." Journal of Marketing Research 8 (August): 355-63.

Green, Paul E. and V. Srinivasan. 1978. "Conjoint Analysis in Consumer Research: Issues and Outlook." Journal of Consumer Research 5 (September): 103-23.

Green, Paul E. and V. Srinivasan. 1990. "Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice." Journal of Marketing 54 (October): 3-19.

Jain, Subhash C. 1989. "Standardization of International Marketing Strat- egy: Some Research Hypotheses." Journal of Marketing 53 (January): 70-9.

Kaynak, Erdener. 1984. "Future Directions of Research in International Marketing." In International Marketing Management. Ed. E. Kaynak. New York: Praeger, 337-44.


Malhotra, Naresh K. 1982a. "Information Load and Consumer Decision Making." Journal of Consumer Research 8 (March): 419-30.

�9 1982b. "Structural Reliability and Stability of Nonmetric Conjoint Analysis." Journal of Marketing Research 14 (May): 199- 207.

-. 1993. Marketing Research: An Applied Orientation. Englewood Cliffs, N J: Prentice-Hall.

McCullough, James and Roger Best. 1979. "Conjoint Measurement: Temporal Stability and Structural Reliability." Journal of Marketing Research 16 (February): 26-31.

Messier, W. and D. Emery. 1980. "Some Cautionary Notes on the Use of Conjoint Measurement for Human Judgment Modeling." Decision Sciences 11 (October): 678-90.

Moore, William L. 1980. "Levels of Aggregation in Conjoint Analysis: An Empirical Comparison." Journal of Marketing Research 17 (November): 516-23.

Oppewal, H., J. Louviere, and H. J. P. Timmermann. 1994. "Modeling Hierarchical Conjoint Processes with Integrated Choice Experiments." Journal of Marketing Research 31 (February): 15-27.

Principal International Business. 1992. New York: Dun & Bradstreet International.

Scott, Jerome E. and Peter Wright. 1976. "Modeling an Organizational Buyers' Product Evaluation Strategy: Validity and Procedural Consid- erations." Journal of Marketing Research 13 (August): 211-24.

Sorenson, Ralph Z. and Ulrich E. Wiechmann. 1975. "How Multination- als View Marketing Standardization." Harvard Business Review 53 (May-June): 38-167.

Ward's Business Directory of U.S. Private Companies. 1992. Belmont, CA: Information Access.

Whitelock, Jeryl and Djamila Chung. 1989. "Cross-Cultural Advertising: An Empirical Study." International Journal of Advertising 8 (3): 291-310.

Wittink, Dick R., Lakshman Krishnamurthi, and Julia B. Nutter. 1982. "Comparing Derived Importance Weights Across Attributes." Journal of Consumer Research 8 (March): 471-4.

Wright, Peter. 1975. "Consumer Choice Strategies: Simplifying vs. Op- timizing." Journal of Marketing Research 12 (February): 60-7.

ABOUTTHEAUTHORS

Imad B. Baalbaki holds a Ph.D. from the Georgia Institute of Technology and is currently assistant professor of marketing at the Graduate School of Business and Management at the Ameri- can University of Beirut, Lebanon. His previous research has appeared in the International Marketing Review and the Journal of Euromarketing.

Naresh K. Malhotra is regents' professor and coordinator of marketing and management science at the Georgia Institute of Technology, Atlanta. He is also the president of the Academy of Marketing Science, a distinguished fellow of the academy and fellow of the Decision Sciences Institute. He has published 60 articles in major refereed journals including the Journal of the Academy o f Marketing Science, Journal o f Marketing Research, Journal of Consumer Research, Marketing Science, Journal of Marketing, Journal of Retailing, Journal of Health Care Market- ing, as well as leading journals in statistics, management science, and psychology. In addition, he has also published numerous refereed articles in the proceedings of major national and international conferences. He is the winner of numerous awards and honors for research, teaching, and service to the profession.

Full Text

Documents

distributionrelated

level of standardization

academy of marketing

environmental variables

degree of standardization

cultural variables

set of variables

standardization decision