ship varies in response to a variety of company-wide and subsidiary level conditions. ... School of Business Administration, New York University. His current ...
CENTRALIZATIONIN HEADQUARTERSSUBSIDIARY RELATIONSHIPS Stephen R. Gates* New York University and WilliamG. Egelhoff** New York University Abstract. Research to date on centralization in MNCs has produced many inconsistent findings. The present study attempts to add clarity to the situation by re-testing many of the existing hypotheses with data from a recent study of centralization in 50 large US, UK and European MNCs. It examines how the degree of centralization inherent in the headquarters-foreignsubsidiary relationship varies in response to a variety of company-wide and subsidiary level conditions. The influence of company-wide conditions on centralization appears to be much clearer than the influence of subsidiary-level conditions. The study also identifies several new contingency patterns that appear to be at work in MNCs today. The extent to which decisionmaking in multinational corporations (MNCs) is centralized or decentralized has recently emerged as an area of considerable research interest (Picard, 1977; Goehle, 1980; Hedlund, 1981; Doz and Prahalad, 1981; Garnier, 1982; Van Den Bulcke, 1984). This MNC research addresses issues of importance to both headquarters managers and host government policymakers alike. Doz and Prahalad(1981) observe that as subsidiaries reduce their dependence on the parent for strategic resources, it is more difficult for headquartersmanagersto enforce top-down decisionmaking. To assure headquarter's role in directing its subsidiaries, they recommend implementing more sophisticated control mechanisms such as career planning or measurement and reward procedures. Singh * Stephen R. Gates is an Instructorin InternationalBusinessManagementat the College of BusinessAdministration,New York University.He holds an MIA from Johns Hopkins' School of Advanced InternationalStudies, and is completinga Ph.D. at New York University. His dissertation research investigates the link between corporate strategyand technologicalcooperationin the global semiconductorindustry. ** William G. Egelhoff is an Assistant Professor of Managementat the Graduate School of Business Administration,New York University. His current researchdeals with the implementationof strategy in multinationalcorporations,and he has written severalarticleson this subject. Date Received: October 1984; Revised: May and November 1985; Accepted: January 1986.
Palgrave Macmillan Journals is collaborating with JSTOR to digitize, preserve, and extend access to Journal of International Business Studies ® www.jstor.org
JOURNAL OF INTERNATIONALBUSINESSSTUDIES, SUMMER1986
(1981) notes that tension may intensify between the MNCs' tendency to manage both production and non-production activities globally and the host government policymakers' desire to integrate MNC subsidiaries activities more closely into local economies. Several areas of divergence between host government objectives and the operations of MNCs are most likely to result in conflict. Among others, these include the ownership of subsidiaries, the amount of manufacturing value-added in the local economy, and local participation in decisionmaking. One of the most salient conflicts over centralization between headquarters and subsidiaries is the European Community's Vredeling Directive (Van Den Bulcke, 1984). Introduced by the EC Commission to the EC Council on October 24, 1980, the proposal establishes procedures for informing and consulting MNC employees in several important decision areas. After three years of dispute between trade union organizations and employers' representatives, and many amendments passed by the European Parliament, a modified draft was presented on June 15, 1983. A particularly nettlesome point of debate was the bypass provision whereby trade union representatives would communicate directly with the management of the parent company if the management of the subsidiary did not provide required information. The trade union spokesmen assumed that MNCs have a centralized decisionmaking structure. This provision has been weakened from its original form to a one-way written request for information by the employees' representatives to headquarters management. An answer would be provided by the local subsidiary management. Given the salience of the issue to policymakers, empirical research on centralization should seek to inform the debate. To date, however, MNC researchers have generated contradictory hypotheses regarding centralization and its determinants, and reported inconsistent results. In the context of an MNC, centralization has been defined as "the division of decision making authority between the headquarters and the various operational units" (Garnier, 1982). This definition is consistent with the notion of vertical as opposed to horizontal centralization in an organization (Mintzberg, 1979). Picard (1977) and Hedlund (1981) have used similar definitions of centralization, while Goehle (1980) focused more on the relative levels of influence exerted by subsidiary and headquarters management. The definition used in the present paper is closer to Garnier's. It is the level at which a decision would need to be approved before being implemented. Furthermore, to highlight the major points of agreement among the above-mentioned researchers, this paper uses a conceptual framework drawn from organization theory. Moreover, by emphasizing the distinction between company-wide and subsidiary-level conditions within this framework, it is possible to observe additional patterns in the results. Organizational theorists have considered centralization to be an important determinant of organizational effectiveness (Boseman and Jones, 1974) and adaptiveness (Blau and Schoenherr, 1971). In turn, centralization
has been found to have a negative relationship with size (Pugh, et al., 1969; Blau and Schoenherr, 1971; Child, 1973), complexity (Hage and Aiken, 1967; Thompson, 1967), and environmental change (Burns and Stalker, 1961; Lawrence and Lorsch, 1967). Many of the contingency variables investigated in the MNC context can be identified as measures of size, complexity or environmental change. Size has been measured as the size of foreign operations (Picard, 1977; Goehle, 1980; Garnier, 1982), and total company size (Picard, 1977; Hedlund, 1981; Garnier, 1982). Complexity has been measured as foreign product diversity (Picard, 1971; Goehle, 1980; Garnier, 1982), product modification differences between subsidiaries (Picard, 1977), extent of outside ownership in foreign subsidiaries (Picard, 1977; Hedlund, 1981; Garnier, 1982), and extent of foreign acquisitions (Hedlund, 1981; Garnier, 1982). Environmental change has been measured as product change (Picard, 1977; Hedlund, 1981), and competitive climate change (Hedlund, 1981). This ability to subsume specific MNC contingency variables under these three broad contingency concepts of organizational theory provides a basis for bringing new clarity to the issue of centralization in MNCs. The present study seeks to reexamine many of the conflicting relationships that previous research has found between centralization and the MNC contingency variables. It is based on a new data collection that involves a relatively large sample size and, in some cases, more explicit and consistent measurement of the variables than previous studies of centralization in MNCs. First, the sample and measures will be described. In the ensuing section, hypotheses drawn from previous MNC researcherswill be presented, and the present study's findings will be contrasted with their results. METHODOLOGY
The study examined centralization in fifty large MNCs. Twenty-four companies are headquartered in the US, six in the UK, and twenty in Europe. The companies tend to be the largest companies in the following industry groups: auto/truck, electrical/telecommunications equipment, industrial equipment, chemicals, pharmaceuticals, consumer packaged goods, and tires. Companies with less than 15 percent foreign sales or with only minimal foreign manufacturing were excluded from the sample for not being sufficiently multinational. Data was collected through published company documents and three structured interviews conducted at each company's headquarters. The first interview dealt with characteristics of the company's international strategy and environment and was usually conducted with a manager having general responsibility for international operations. The other two interviews focused on two specific foreign subsidiaries (Brazil and a major European subsidiary) and their relationships with the parent. Here the interviewee tended to be the managerto whom the foreign subsidiarymanagerreported.
JOURNALOF INTERNATIONALBUSINESSSTUDIES, SUMMER1986
In total, data was collected on 94 specific parent-foreign subsidiary relationships, and the unit of analysis is the degree of centralization existent in a specific parent-foreign subsidiary relationship. Measures
Centralization was measured as the level in the organization which would need to approve a decision before it could be implemented. The levels were ranked: 1) within the foreign subsidiary, 2) within the international division, worldwide product division, or geographical region headquarters, and 3) above the divisional level and within the corporate headquarters. The degree of centralization was determined for 22 important decisions, which are described in Table 1. The individual decision scores were averaged together to produce three separate centralization scales: one for marketing decisions, a second for manufacturing decisions, and a third for financial decisions. TABLE1 Decisions Used to MeasureDegreeof Centralization Marketing Decisions 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Small changes in product prices Large change in product prices (sufficient to break or destabilize the market) Small changes in product design Large change in product design (substantially altering the cost, image and possibly end-use) Small changes in sales commissions or the basis upon which selling agents are paid Decisions regarding selling methods (whom to call on, approach to be used, etc.) Basic changes in the method of distribution (e.g., selling direct instead of through wholesalers) Advertising and promotion decisions Decisions implementing product guarantee Decisions regarding how to service products
Manufacturing Decisions 11. The decision to subcontract out large portions of the manufacturing instead of expanding the subsidiary's own facilities 12. Approval of quarterly production schedules and plans 13. The decision to switch to a new manufacturing process, employing different methods and equipment, when expanding new plant capacity 14. Decisions regarding routine purchasing activities 15. Decisions regarding quality control Financial Decisions 16. 17. 18. 19. 20. 21. 22.
The decision to significantly increase inventories in order to keep the plant full Change credit terms to customers Decisions involving credit checking and the extension of trade credit Borrow short term from local banks Decisions involving how to finance a major facilities expansion Write off $100,000 of questionable inventory Decisions involving what insurance to carry and whom to carry it with
Levels of centralization were ranked: 1 = within the subsidiary; 2 = within the international divi= sion, product division or geographic region headquarters; 3 above the division level and within the corporate headquarters.
75 TABLE 2
Means, Standard Deviations, and Correlation Matrix of Centralization Scales (N = 89-94)
1. Marketing centralization 2. Manufacturing centralization 3. Financial centralization
1.45 1.52 1.55
.28 .32 .36
**p < .01 (two-tailed test); ***p < .001
Table 2 shows the means, standard deviations, and intercorrelation among the three centralization scales. The significant positive correlation among the three scales indicates some overall tendency toward centralization or decentralization in a parent-subsidiaryrelationship, even across functional areas. Yet, there is still sufficient difference to consider each type of centralization separately. Consequently, a composite scale, such as Garnier (1982) employed, was not constructed. Later analysis, in fact, will show that the three scales share differing relationships with a number of the contingency variables. The study also measured fifteen contingency variables. Some are similar to contingency variables used by previous MNC researchers, while others represent new attempts to operationalize the concepts of organizational size, complexity, and environmental change. These variables exist at two different levels of analysis. Nine of the variables were measured at the company level and apply to all parent-subsidiary relationships within a company: size of foreign operations, size of MNC, foreign product diversity, product modification differences between subsidiaries, extent of outside ownership in foreign subsidiaries, extent of foreign acquisitions, industry, nationality, and age of company abroad. Six variables were measured at the subsidiary level and apply to a specific parent-subsidiaryrelationship: relative size of subsidiary, size of subsidiary, degree of product change in subsidiary, competitive climate change faced by subsidiary, level of intracompany purchases by subsidiary, and age of subsidiary. Measurement of the contingency variables is described in the Appendix. Table 3 shows the correlations among the contingency variables. While there is significant correlation among some of the variables, the levels are still sufficiently low that each variable can be viewed as representing a different contingency condition. HYPOTHESESAND RESULTS Company-Level
The hypotheses involving centralization have been divided into two groups: 1) those dealing with company-level contingency variables and
TABLE 3 Correlation Matrix of Contingency Variables, 3
Company-Level Variables (N = 44-50) 1. Size of foreign operations 2. Size of MNC 3. Foreign product diversity 4. Product modification differences 5. Extent of outside ownership 6. Extent of foreign acquisitions 7. Age of company abroad
.27* -.08 .17 .31* .28*
.07 33** .04 -.07 .36**
Subsidiary-Level Variables (N = 75-94) 8. Relative size of subsidiary 9. Size of subsidiary 10. Product change 11. Competitive climate change 12. Intracompany imports by sub. 13. Age of subsidiary
-.03 -.10 -.12 -.15 .17 .23*
-.23* .17 .63*** -.14 .03 -.15 .20 -.19 -.10 .12 .26** .39***
* p < .05 (two-tailed test), ** p < .01, *** p < .001
.10 .41 ** .23 .44***
-.05 .24* -.02
.10 .28** .12 .04 -.46*** 0
-.02 -.03 -.03 -.14 .12 .07
-.11 -.08 -.05 -.16 -.06 -.04
-.09 .1 -.10 -.01 .0 .7
2) those dealing with subsidiary-level contingency variables. Within each group, hypotheses have been further subdivided by the broad contingency concepts of organization theory to which they relate (i.e., size, complexity, change) and another subgroup consisting of all other contingency variables. Size
Among organization theorists the general consensus is that increasing size leads to decentralization. Their arguments range from the notion that size leads to more structuring of activities which then facilitates decentralization (Pugh, et al., 1969) to the notion that size forces managersto decentralize by threatening to overload them with too many decisions (Blau and Schoenherr, 1971). Organization theorists have generally looked within the organization, at its structure and processes, to generate hypotheses concerning the impact of size on centralization. On the other hand, MNC researchers have usually turned to the multinational environment itself, and its attendant riskiness, to develop hypotheses about the impact of size-related contingency variables on centralization. Two of these contingency variables and their related hypotheses are presented and discussed below. Size of foreign operations. Garnier, Picard and Goehle hypothesized a positive correlation between the size of foreign operations and centralization. They maintained that as the percentage of the firm's revenues from abroad rises, dependence risk increases. In response, headquarters exercises greater direct control over its subsidiaries. H. 1. Centralization is positively correlated with the size of foreign operations in an MNC. Reported in Table 4, this study's results contradicted the hypothesis. All three correlations are in the opposite direction, and the negative correlation with marketing centralization is significant. In Table 5, it is shown that although Garnier's results supported this hypothesis, neither Picard's nor Goehle's was significant. The present study's findings confirm the organization theorists' hypothesized negative correlation. Size of MNC. Garnier argued that the size of the total company is negatively correlated with centralization because a larger MNC has more difficulty exercising direct control over each of its many foreign units. To lessen the demands of the control task, it institutes indirect controls and decentralization simultaneously. H.2. Centralization is negatively correlated with size of the MNC. The findings in the current study were mixed (Table 4). Although there is a significant negative correlation for financial centralization, there is also a positive correlation for marketing centralization. It is interesting that Garnier also found a significant positive correlation contrary to his hypothesis (Table 5). Furthermore, while Picard's negative result supported the hypothesis, Hedlund's was inconclusive. The present study's finding for
Testingthe Hypotheses-CorrelationBetween Centralizationand the Con Hypothesis
Company Level SIZE H.1. Size of foreign operations H.2. Size of MNC
Hypothesized Correlation with Centralization
COMPLEXITY H.3. Foreign product diversity H.4. Product modification differences between subsidiaries H.5. Extent of outside ownership in foreign subsidiaries H.6. Extent of foreign acquisitions OTHER H.9. Age of company abroad
Subsidiary Level SIZE H.10. Relative size of subsidiary H.11. Size of subsidiary ENVIRONMENTAL CHANGE H.12. Product change H.13. Competitive climate change
OTHER + .14 .08 H.14. Intracompany imports by subsidiary .06 -.01 H.15. Age of subsidiary Note: When the results contradicted the hypothesis, a two-tailed test was used to determine the significance level. Ot *p < .05
**p < .01
***p < .001
Summary of Research on Relationships Found to Exist between Centralization and C
Pugh, Child Blau &Schoenherr
Lawrence& Lorsch Burns& Stalker
Company Level SIZE H.1. H.2.
Size of foreign operations Size of MNC
COMPLEXITY H.3. Foreign product diversity Product modification differences H,4. between subsidiaries Extent of outside ownership in H.5. foreign subsidiaries Extent of foreign acquisitions H.6.
OTHER H.7. Industry H.8. Nationality H.9. Age of company abroad Subsidiary Level SIZE H.10. H. 1.
Relative size of subsidiary Size of subsidiary
ENVIRONMENTAL CHANGE Product change H. 12. H.1I3. Competitive climate change OTHER H.14. Intracompany purchases by subsidiary H.15. Age of subsidiary 1 2. 3.
Only for European and UK MNCs Only for U.S. MNCs Insignificant, mixed results at nationality subgroup level of analysis
+ m.r. * n.s. incl.
n n.s. one or more significant positive resu one or more significant negative resu mixed results (significant results in b significant differences across groups not significant results inconclusive due to data limit
JOURNAL OF INTERNATIONALBUSINESSSTUDIES, SUMMER1986
financial centralization confirmed the organization theorists' negative correlation. However, none of the arguments regardingsize and centralization explains the positive correlation with marketing centralization. Complexity
Organization theorists tend to agree that both internal and external complexity are associated with decentralization. Internal complexity derives from additional occupational specializations, and top decisionmakers are likely to draw on these sources of expertise (Hage and Aiken, 1967). External complexity requires an organization to establish new structural units to cope with uncertain elements in the environment, thereby decentralizing decisionmaking (Thompson, 1967). A number of the contingency variables studied by MNC researchers can be taken as measures of internal or external complexity: Foreign product diversity. This variable has elements of both types of complexity. Each new product line that is transferred to foreign subsidiaries requires additional specialization in the foreign subsidiaries and additional coordination between the foreign subsidiaries and headquarters. This results in higher internal complexity for the foreign subsidiaries and headquarters. External complexity also increases, since the marketing of different product lines is apt to force the foreign subsidiaries into different market segments, with different buyers, distribution systems, aftersales service, local competition, etc. Both Garnier and Picard predicted that foreign product diversity is negatively correlated with centralization. Garnier argued that the more product diversity there is in an MNC, the greater the probability that some products will have been designed by subsidiaries specifically for their local market. Consequently, more decisions would tend to be made at the subsidiary level. H.3. Centralization is negatively correlated with the level of foreign product diversity in an MNC. The hypothesis was supported for marketing and financial centralization (Table 4). Picard's result also supported the hypothesis, but Garnier's finding was mixed, and Goehle's was inconclusive (Table 5). The current study supported the direction hypothesized by both MNC researchers and organization theorists. Product modification differences between subsidiaries. As this factor increases in a company, it raises the level of internal complexity in the organization. The standardization of products and personnel across subsidiaries is hampered, and differentiation and specialization tend to occur between subsidiaries. Picard posited that the need to alter products to better reflect local environmental or cultural differences leads to decentralization. H.4. Centralization is negatively correlated with the degree of product modification differences between subsidiariesin an MNC.
The hypothesis was supported for manufacturing centralization (Table 4). Picard's finding confirmed the negative direction hypothesized by both MNC researchersand organization theorists (Table 5). Extent of outside ownership in foreign subsidiaries. The extent of outside ownership in an MNC's foreign subsidiariesis a common source of external complexity. Outside equity-holders in a foreign subsidiary will be primarily interested in the goals and performance of the individual subsidiary. They may prefer short-term profit maximization or non-economic goals, especially if the equity-holder is the local government. Such goal multiplicity increases external complexity both for the individual subsidiary and for the parent headquarters. Garnier hypothesized that the higher the extent of outside ownership in foreign subsidiaries, the lower is the subsidiaries' legal dependence on the parent. Hence, there is a tendency to decentralize. H.5. Centralization is negatively correlated with the extent of outside ownership in foreign subsidiariesin an MNC. The hypothesis was supported for marketing and financial centralization. The negative correlation was especially strong for financial centralization (p < .01) (Table 4). Both Garnier's and Picard's results supported the hypothesis, although Hedlund's findings were not significant (Table 5). Most of these results, then, support the negative direction hypothesized by both MNC researchersand organization theorists. Extent of foreign acquisitions. Foreign acquisitions generate both internal and external complexity. An MNC inherits the acquired subsidiary's occupational roles and organizational coordination practices, which are likely to differ from the pattern established by the parent firm. In addition, the acquired subsidiary may be operating in market segments or local markets that are unfamiliar to the MNC. Hedlund offered two reasons why an MNC decentralizes as the extent of foreign acquisitions increases: 1) the parent encounters difficulties in gaining control over a subsidiary with an independent history and structure, and 2) acquisitions are often undertaken to obtain expertise that the parent does not have, and therefore does not feel it should interfere with. H.6. Centralization is negatively correlated with the extent of foreign acquisitions in an MNC. The hypothesis was supported for marketing and manufacturing centralization (Table 4). Both Garnier'sand Hedlund's results also supported the negative direction hypothesized by both MNC researchers and organization theorists (Table 5). Other
The following company-level variables are not related to any of the broad contingency concepts identified by organization theory. They are included here because they have appeared as important contingency variables in previous MNC research studies dealing with centralization. For the nominal
JOURNAL OF INTERNATIONAL BUSINESS STUDIES, SUMMER 1986
variables, industry and nationality, ANOVA test results are presented in Tables 6 and 7. TABLE6 Centralizationand Industry
Elect & Auto Tel equip
Consumer Pharm pack goods
***p < .001
TABLE7 Centralizationand ParentCountryNationality
Marketing centralization Manufacturing centralization Financial centralization *p < .05
US 1.51 1.54 1.63
UK 1.28 1.31 1.68
European 1.43 1.56 1.41
F 3.1* 3.0 5.0**
**p < .01
Industry. Goehle found differences in the degree of centralization between her industry groups, especially when she looked at functional area decisions between industries. She concluded that industry should be considered an important contingency variable. She suggested that the product lines in each industry could be responsible for differences in centralization. Porter (1980) set out four global strategies (and therefore implicitly the structural aspects of centralization) based on an analysis of how six critical industry-specific conditions affect a firm's product lines. Within each industry, the global competitive factors include scale economies in both production and non-production activities, transportation costs, distribution channels, comparative factor costs, heterogeneity of market demand, and governmental barriers to trade. There are four possible strategies to compete under these conditions. A global strategy requiring strong centralization can be pursued in all product lines in a perfectly competitive market. Where only some product lines face low barriers to trade, the global strategy works for only part of the firm's operations. A more decentralized, national strategy is optimal where product lines are subject to economic impediments such as transportation costs or heterogeneity of demand. Lastly, a firm may choose a highly decentralized protected niche strategy where governments grant preferential access in a given product
line. Given that for different industries, these six critical conditions and potential strategies vary, it is likely that centralization differs across industry groups. H.7. Centralization differs significantly according to the primary industry group in which the MNC operates. The hypothesis was supported for marketing centralization (Table 6). The automobile and industrial equipment industries are above the mean level of marketing centralization whereas the chemical, electrical and telecommunications equipment, consumer packaged goods, and pharmaceutical industries tend to be less centralized than the mean. There is greater variance between industries concerning local marketing decisions than concerning local manufacturing or financial decisions. Goehle (1980) found that the tire and rubber industry was the most centralized, and that the food processing industry was the least centralized. Our results show that the tire industry had only an average level of centralization, and we did not consider the food processing industry. In both studies, the least centralized industries (food processing and consumer package goods) are ones which must respond closely to differences in local tastes. This observation that some industries must consider local market receptivity more than others is reinforced by our finding that there is greater variance between industries concerning local marketing decisions than concerning local manufacturing or financial decisions. It also confirms Porter's statement that economic impediments to global competition, especially heterogeneity of market demand, lead to more decentralized national product line focus. Nationality. Hedlund observed that centralization varied according to the parent company's country of origin. He found that Swedish and Japanese firms are more decentralized that US firms. For Swedish firms, decentralization resulted from firms' expansion out of a small market at an early, entrepreneurial phase of their development. Coordination between national subsidiaries and headquarters was done on an informal, personal basis. This developed into the mother-daughter structure that Stopford and Wells (1972) and Franko (1976) found more among continental European firms than among the US or UK firms. Recent studies of US and Japanese management styles suggest historical and cultural reasons for greater centralization in US firms (Ouchi and Jaeger, 1978; Ouchi, 1980). Organizational practices adapt to the culture in which they are generated. In the US, rugged individualism, nonconformity and racial and ethnic diversity necessitate more explicit, formalized control. Explicit rules are needed to govern competition among organizational members. In Japan, collectivism, conformity and racial and ethnic homogeneity allow more social, implicit control among organizational members. It is likely that formal structural centralization will vary by nationality of the firm. H.8. Centralization differs significantly according to the nationality of the MNC parent company.
JOURNAL OF INTERNATIONAL BUSINESS STUDIES, SUMMER 1986
The hypothesis was supported for both marketing and financial centralization (Table 7). Relative to US and European firms, UK firms have more decentralized marketing and manufacturingdecisionmaking. Compared to US and UK firms, European firms have more decentralized financial decisionmaking. In general, the data suggest that US firms exhibit greater centralization. Future research may establish historical or cultural reasons for the functional differences between these parent country nationalities. Age of company abroad. Both Hedlund and Goehle tested the hypothesis that firms with greater international experience tend to centralize more. Neither, however, presents any rationale as to why this relationship should exist. H.9. Centralization is positively correlated with the age of the MNCabroad. The hypothesis was contradicted for marketing and financial centralization (Table 4). Neither Hedlund's nor Goehle's results supported the hypothesis (Table 5). Among the sample companies, it appears that the longer a firm has been managing significant foreign operations, the more decentralized decisionmaking for its foreign subsidiarieswill be. SUBSIDIARY-LEVEL HYPOTHESES
Size Organization theorists have not considered the impact of subunit (subsidiary) size on the degree of centralization in an organization. Therefore, the only hypothesized relationships for subsidiary size are those of the MNC researchers. Relative size of subsidiary. Garnier maintained that the larger the relative size of a subsidiary to its parent firm, the better bargainingposition it has. Its bargaining strength can lead to a more decentralized pattern of decisionmaking.
H.10. Centralization is negatively correlated with the relative size of a subsidiary. The findings of the current study were mixed. The hypothesis was supported for manufacturing centralization, but contradicted for marketing centralization (Table 3). Although Hedlund's results supported the hypothesis, Garnier, Picard and Goehle all failed to find a significant relationship (Table 5). Size of subsidiary. Picard claimed that the parent grants more autonomy to a subsidiary that is large in an absolute sense because the latter has the ability to support a full management staff. H. 11. Centralization is negatively correlated with the absolute size of a subsidiary. The hypothesis was contradicted for marketing centralization (Table 4). Although Picard's and Hedlund's results supported the hypothesis, Garnier and Goehle found no significant relationship (Table 5).
Organization theorists seem to agree that an organization facing a dynamic environment tends to decentralize. Both Burns and Stalker (1961) and Lawrence and Lorsch (1967) have argued that environmental change presents fresh problems and unforeseen requirements for action. In the latter group's study, they found that subunits facing more dynamic sub-environments tend to decentralize decisionmaking more than other subunits of the same organization facing stable sub-environments. MNC researchers have also looked at the relationships between centralization and sub-environments of an organization, but their approach differs from that of Lawrence and Lorsch. The latter examined differences between the functional subunits of organizations, and consequently they focused on the marketing, manufacturing, and research sub-environments of firms. The MNC researchers have examined centralization in different foreign subsidiaries, and their focus has been on geographical sub-environments (which also implies cultural and political differentiation at the subenvironment level). The following paragraphs will look at two different measures of sub-environment change which have been used by MNC researchers. Product change. Product change within a foreign subsidiary measures the degree to which new products are being introduced by the subsidiary to its local markets. It implies change in the marketing and possibly also the manufacturing and research environments of a subsidiary. Picard asserted that a greater degree of autonomy is given to subsidiaries experiencing a high degree of product change. This is due to their need to react quickly to rapid style changes and a heavy flow of new products. H. 12. Centralization is negatively correlated with product change. The hypothesis was contradicted for marketing centralization. The correlations were in the positive direction for manufacturing and financial centralization as well (Table 4). Although Picard found a negative relationship, Goehle's result was not significant (Table 5). The current study's findings were in the opposite direction from that predicted by both MNC researchersand organization theorists. Competitive climate change. Competitive climate change measures the extent to which a subsidiary experiences an increase in its local market competition due to new entrants (other MNCs or local competitors), lower tariffs, etc. Hedlund posited that change in the competitive climate of a subsidiary would induce an MNC to grant it more autonomy. H. 13. Centralization is negatively correlated with competitive climate change. The hypothesis was contradicted for marketing centralization. Similar to the findings for product change, the results were in the positive direction for manufacturing and financial centralization (Table 4). Hedlund's results were not significant (Table 5). The current study's findings were in the
JOURNAL OF INTERNATIONAL BUSINESS STUDIES, SUMMER 1986
opposite direction from that predicted by both MNC researchers and organization theorists. Other
The remaining subsidiary-level variables are not directly related to any of the broad contingency concepts which have emanated from organization theory. Once again, they are included because previous MNC researchers have posited relationships between them and centralization. Intracompany purchases by subsidiary. Garnier theorized that the greater the intracompany purchases by a subsidiary, the more decisionmaking for the subsidiary would be centralized. He argued that intracompany purchases are a critical type of dependence which leads to reduced autonomy. H.14. Centralization for a subsidiary is positively correlated with intracompany purchases by the subsidiary. The hypothesis was contradicted for financial centralization (Table 4). Though not significant, the results were in the predicted direction for marketing and manufacturing centralization. Garnier and Hedlund found a positive relationship, but Picard's results indicated a significant negative relationship (Table 5). Age of subsidiary. Garnier reasoned that older subsidiaries, having more experience, represent less risk to an MNC. Therefore, they should enjoy greater autonomy. H. 15. Centralization is negatively correlated with the age of a subsidiary. The hypothesis was supported for financial centralization. However, there was little indication of any support for marketing or manufacturing centralization (Table 4). None of the MNC researchersfound significant results 'for this hypothesis (Table 5). Since there are a number of significant correlations between company-level and subsidiary-level contingency variables, the correlations involving company-level variables and centralization were rerun controlling for any significantly related subsidiary-level variable. Similarly, the correlations involving subsidiary-level variables were rerun controlling for any significantly related company-level variable. The only relationships between centralization and the contingency variables that changed significantly were those involving size of MNC and subsidiary, and age of company and subsidiary. Table 3 shows there is substantial interaction between company and subsidiary size, and company and subsidiary age (e.g., larger MNCs have larger subsidiaries and older international companies have older subsidiaries). Even here, what changed was the strength of the relationships and not the direction. The only exception to this involved the relationship between centralization and age of subsidiary. When age of company abroad was controlled for, there was no meaningful correlation between financial centralization and age of subsidiary, but both marketing and manufacturing centralization revealed significant negative correlations
with age of subsidiary. Thus, the interaction of company and subsidiarylevel variables has relatively little impact on the findings reported in Table 4. Since there are significant relationships between nationality and centralization (Table 7) and between nationality and the contingency variables (European MNCs, for example, tend to have a higher percentage of foreign sales than do U.S. MNCs), the correlations shown in Table 4 were also rerun separately for each of the three nationality subgroups. While significance levels obviously changed due to the smaller sample sizes, only three of the relationships showed significant differences among the nationality subgroups. The negative relationship found between size of foreign operations and centralization in H. 1 is entirely due to strong negative relationships in European and UK MNCs. U.S. MNCs actually showed a significant positive relationship between size of foreign operations and manufacturing centralization and weak positive relationships with market ing and financial centralization. The positive relationship between subsidiary size and marketing centralization in H. 11 occurs only in U.S. MNCs. It is weakly negative for both European and UK MNCs. Finally, the negative relationship between intracompany imports and financial centralization in H. 14 disappearsinto insignificant and mixed findings at the nationality subgroup-level of analysis. The results in Table 5 have been footnoted to indicate these exceptions. Otherwise, findings at the nationality subgroup-level of analysis tend to support findings at the total sample-level of analysis, indicating the robustness of most relationships across parent company nationality. Industry is another factor that might potentially influence the relationship between centralization and the other contingency variables. Unfortunately, the present study was unable to test this because there are too few companies in each industry group. DISCUSSION
It is apparent from Table 5 that the relationship between centralization and various measures of company-level complexity is quite consistent across both measures and studies. While a number of MNC studies have produced inconclusive or nonsignificant results, none has contradicted the negative relationship posited by organization theorists and supported by the present study. The specific measures of complexity employed by the present study suggest that if MNCs introduce more product lines in foreign markets or modify products to fit differing local environments, such strategic changes are likely to require more decentralization of decisionmaking. Similarly, MNCs that allow substantial outside ownership in foreign subsidiaries or attempt to grow through the acquisition of foreign firms are also likely to decentralize decisionmaking. The negative relationship posited by organization theorists between organizational size and centralization is less apparent in the MNC studies. The
JOURNAL OF INTERNATIONAL BUSINESS STUDIES, SUMMER 1986
present study found that European and UK MNCs tend to decentralize as the relative size of their foreign operations increase, but this tendency was not apparent in U.S. MNCs. The absolute size of a company shared mixed relationships with the different centralization scales used in the present study and results across different MNC studies were also mixed. This seems to suggest that organizational size might have a different influence on different types of centralization (e.g., marketing centralization, manufacturing centralization) within an organization. Other contingency factors at the company level also share significant relationships with the degree of centralization existant in parent-subsidiary relationships. The longer MNCs have had significant foreign operations, the more decentralized parent-subsidiary relationships seem to. be. Age is a correlate of size of foreign operations, but continues to exert a significant influence on centralization even when the latter is controlled for. The present study also found a significant relationship between marketing centralization and type of industry. Automobile and industrial equipment companies were found to have the highest levels of marketing centralization. Centralization also varies by nationality of the parent company, with U.S. MNCs tending toward more centralization of marketing and financial decisions than European MNCs. UK MNCs have the most decentralized marketing and manufacturing decisionmaking and the most centralized financial decisionmaking. Of the nine hypothesized relationships between centralization and company-level variables, six were supported by the present study, two were contradicted, and one was characterized by mixed results. It should be noted that the hypotheses involving industry and nationality only called for some significant relationship without specifying its direction. While the results for one of the contradicted hypotheses (H.1) failed to confirm the hypothesis of MNC researchers, they supported the relationship specified by organization theorists. This rather consistent support means that many of the relationships between centralization and company-level contingency variables appear to be quite robust. The relationship between centralization and the six subsidiary-level variables produces a somewhat different picture. The relative size of a foreign subsidiary was hypothesized in earlier studies to be negatively related to centralization, although only one of the four previous MNC studies actually found a significant negative relationship. The present study found a significant negative relationship with manufacturing centralization, but a significant positive relationship with marketing centralization. The absolute size of a subsidiary was also hypothesized to be negatively related to centralization, but U.S. MNCs show a strong positive relationship between the size of a subsidiary and marketing centralization. In European MNCs the relationships between subsidiary size and centralization tend to be negative, but not statistically significant. The generally weak and nonsignificant relationships between size and centralization found in previous studies and the mixed results of the present study suggest that the relationship between organizational size
and centralization is not a simple one. As was suggested above when the relationship between centralization and the size of an MNC produced mixed results, one possible explanation is that organizational size exerts a different influence on different types of centralization. The present study uncovered strong positive relationships in U.S. MNCs between marketing centralization and: 1) size of MNC, 2) size of subsidiary, 3) relative size of subsidiary. Relationships involving manufacturing and financial centralization in U.S. MNCs and relationships between all types of centralization and size in non-U.S. MNCstend to be insignificant but negative. This pattern of results suggests that large U.S. MNCs with large foreign subsidiaries might be pursuing a more global marketing strategy than other MNCs. Largerfirms and larger subsidiariesmay be able to achieve scale economies in certain marketing activities (Comanor and Wilson, 1979), but only if they pursue globally integrated marketing strategies. Since such strategies require subsidiary managers to relinquish some control over marketing decisions, this possibility is consistent with the data observed in the study. The two hypotheses involving environmental change at the subsidiary level were not supported by the present study. Previous studies hypothesized that subsidiaries facing more change in their local product lines and competitive climates would be allowed more decisionmaking autonomy. Only one study, however, found any empirical support for this relationship. The present study results suggest a new hypothesis. When a subsidiary faces more environmental change, there are more critical decisions to be made, and this is accompanied by a tendency to centralize decisionmaking for the subsidiary. This hypothesis obviously needs to be tested with further research across different types of subsidiary-level change. An examination of other factors at the subsidiary-level relealed that older subsidiaries are likely to have more autonomy than newer subsidiaries.
Overall, relationships between centralization and subsidiary-level conditions are less clear than those observed between centralization and company-level conditions. An examination of Table 5 reveals many more nonsignificant findings in previous studies. The present study supported only one of the earlier hypotheses, contradicted four and produced mixed results for another. It is interesting that Van Den Bulcke (1984), in a recent but somewhat different study of centralization conducted at the subsidiary level, reported insignificant results for each of his seven variables. (His results are not directly comparable, since he measured a different group of seven strategic decisions.) The relative weakness and lack of consistency which seems to characterize relationships between centralization and subsidiary-level variables suggests that future research should concentrate more on addressingthis level of analysis. The two new relationships suggested by the present study that need to be explored in future research studies are: 1) that some large MNCs seem to be developing global marketing strategies involving their largest subsidiaries, and that this leads to a greater centralization of marketing decisions, and
JOURNALOF INTERNATIONALBUSINESSSTUDIES, SUMMER1986
2) that MNCs tend to centralize decisionmaking for subsidiaries that face high environmental change (and this is most pronounced for marketing decisions). A methodological issue that needs to be considered when designing studies of centralization in MNCs is whether one measures where decisions are made by asking managers in foreign subsidiaries or managers in the parent headquarters. Ideally, one would like to take measures at both levels, but when field study is involved, the cost may be prohibitive. The present study only interviewed managers and took measures at parent headquarters. While this may well introduce a bias toward reporting greater centralization than actually exists, there is no reason to believe this bias is not consistent across all companies in the sample. Since the hypothesis testing involved correlations rather than absolute levels of centralization, the results would be unaffected by such a bias. Without measures at the subsidiary level, however, one should be careful not to draw conclusions about absolute levels of centralization in a parentsubsidiary relationship. Centralization is an important aspect of the structuringof parent-subsidiary relationships in MNCs. The present study has attempted to further our understanding of this subject by investigating and defining important contingency patterns surrounding the degree of centralization existant in a parent-subsidiary relationship. These patterns are important since many of these contingency variables are changing in the environments and strategies of MNCs. Company-level complexity is increasing for most firms, and this should exert continued pressure for more decentralization. At the same time, however, the development of global product and marketing strategies in more MNCs would be a countervailing force for increased centralization in the marketing area. Finally, there is some indication that parent headquarters may be devoting more attention to those subsidiaries facing rapid environmental change. Improvements in the global communications systems of MNCs would certainly facilitate this kind of selective attention or centralization of decisionmaking. Our study suggests these may be the most important forces shaping the pattern of centralization in MNCs today. REFERENCES Blau,P. & Schoenherr,R. TheStructureof Organizations.New York, BasicBooks, 1971. Boseman, F. & Jones, R. Market conditions, decentralization and organizational effectiveness.
HumanRelations, 1974, 27, 665-676. Brooke, M.Z. & Remmers, H.L. The Strategy of MultinationalEnterprise: Organizationand Finance. London, Pitman, 1978. Burns, T. & Stalker, G. The Management of Innovation. London, Tavistock Institute, 1961. Child, J. Predicting and understanding organization structure. Administrative Science Quarterly,
1973, 18, 168-185. Comanor, W.S. & Wilson, T.A. Advertising and competition: a survey. Journal of Economic Literature, June, 1979, 17, 453-476. Doz, Y. & Prahalad, C.K. Headquarter's influence and strategic control in MNCs. Sloan Management Review, Fall, 1981, 15-29.
Franko, L.G. The European Multinationals. London, Harper & Row, 1976. Garnier, G. Context and decision making autonomy in the foreign affiliates of US multinational corporations. Academy of Management Journal, 1982, 25, 893-908. Goehle, D.G. Decision making in Multinational Corporations. Ann Arbor, UMI Research Press, 1980. Hage, J. & Aiken, M. Relationship of centralization to other structural properties. Administrative Science Quarterly, 1967, 12, 72-92. Hedlund, Gunnar. Autonomy of subsidiaries and formalization of headquarters subsidiary relationships in Swedish MNCs. In Lars Otterbeck (ed.), The Management of Headquarters-Subsidiary Relationships in Multinational Corporations, New York, St. Martin's Press, 1981. Lawrence, P. & Lorsch. J. Differentiation and integration in complex organizations. Administrative Science Quarterly, 1967, 12, 1-47. Ouchi, W. Markets, bureaucracies and clans. Administrative Science Quarterly, 1980, 25(1): 129141. Ouchi, W. & Jaeger, A. Type Z organization: stability in the midst of mobility. Academy of Management Review, 1978, 3(2): 305-314. Picard, J. Factors of variance in multinational marketing control. In L.G. Mattson and F. Widersheim-Paul (eds.), Recent Research on the Internationalization of Business. Uppsala, Almqvist & Wiksel, 1977. Porter, M. Competitive Strategy: Techniques For Analyzing Industries and Competitors. New York, Free Press, 1980. Pugh, D., Hickson, D., Hinings, C., Turner, C. The context of organization structures. Administrative Science Quarterly, 1969, 14, 91-114. Singh, R.K.D.N. Policy issues and trends in parent-affiliate relationships in developing countries. In Lars Otterbeck (ed.), The Management of Headquarters-Subsidiary Relationships in Multinational Corporations. New York, St. Martin's Press, 1981. Stopford, J. & Wells, L. Managing the Multinational Enterprise. London, Longmans, 1972. Thompson, J. Organizations in Action. New York, McGraw-Hill, 1967. Van Den Bulcke, D. Decision making in multinational enterprises and the information and consultation of employees: The proposed Vredling Directive of the Ec. Commission. International Studies of Management and Organization, 1984, 14 (1): 36-60.
Company-Level Variables 1. Size of foreign operations was measured by the percentage of a company's sales occurring outside of the parent country. 2. Size of MNC was measured as total number of employees. 3. Foreign product diversity was measured by the number of broad product lines a company offered for sale in two designated foreign countries. One country was a major European country and the second was Brazil. Since both tended to be large, actively developed markets for most companies in the sample, the product offerings in these two markets were considered representative of the company's total foreign product offering. Correlation between the two measures was R=.87. The highest of these two measures was used to represent the company's foreign product diversity. 4. Product modification differences between subsidiaries was measured on a 5-point scale, which the interviewer used to rate the MNC
JOURNALOF INTERNATIONALBUSINESSSTUDIES, SUMMER1986
following a discussion with the respondent of product differences across subsidiaries. One, on the scale, indicates "identical products, parts are interchangeable worldwide;" 3 indicates "substantial cosmetic differences, but share same basic technology engineering;" and 5 indicates "basic technology differences requiring separate basic technology engineering." Extent of outside ownership in a company's foreign subsidiaries was measured by the percentage of a company's foreign sales accounted for by subsidiaries with greater than 30 percent outside ownership. This variable was measured by categories (e.g., 0-5 percent, 5-10 percent) with the final category open-ended (more than 25 percent). Extent of foreign acquisitions was measured by the percentage of a company's foreign sales accounted for by acquisitions made within the past 10 years. This variable was measured by categories (e.g., 0 percent, 0-5 percent) with the final category open-ended (more than 25 percent). Industry was divided into the following categories: auto/truck, electrical/telecommunications equipment, industrial equipment, chemicals, pharmaceuticals, consumer packaged goods, tires and others. Nationality of the parent company was either US, UK or European. Age of company abroad was measured as the number of years since the company began its first significant overseas expansion.
Subsidiary-Level Variables 10. Relative size of subsidiary was measured as the subsidiary's sales as a percent of total foreign sales. 11. Size of subsidiary was measured as the number of employees in the subsidiary. 12. Product change was measured as the percent change in the number of products offered for sale by the subsidiary over the last five years. 13. Competitive climate change was measured as the extent to which the competitive climate of the subsidiary has changed over the past three years, in terms of competitors entering, leaving, changing position, and changes to the basis of competition. Measuredby 5-point scale. 14. Intracompany purchases by subsidiary was measured as the percent of a subsidiary's sales covered by purchases from some other part of the company. 15. Age of subsidiary was measured as the number of years since the subsidiary was founded.