Abstract

This paper examines the importance of MNE distribution subsidiaries to overall MNE activity and tests an agency theory of distribution FDI, in which a firm seeking to sell its product in a new foreign market faces the choice of investing in its own sales operation or contracting with a local sales agent.  This choice is first estimated as a function of a number of macroeconomic variables.  Results suggest that US MNEs are more likely to opt for distribution FDI the larger the foreign market and the greater the degree of economic freedom in the host country. Next, a supply-side variable is introduced, based on the hypothesis that cross-industry differences in sales effort may explain some of the industry variation in distribution FDI.  Results of a negative binomial regression analysis suggest that US MNEs operating in sectors with more complex selling processes are more likely to establish foreign distribution affiliates. 

Introduction

Much of the theoretical work on multinational enterprises (MNEs) focuses on incentives for foreign direct investment (FDI) and resulting impacts on patterns of trade and host country economies.  Prevailing theories assume that FDI consists of migrating manufacturing facilities overseas.  However, anecdotal evidence suggests that many MNEs establish or acquire foreign distribution operations either before or in addition to establishing foreign manufacturing affiliates.[1]   This paper examines the importance of these distribution subsidiaries to overall MNE activity, and tests an agency theory of distribution FDI, in which a firm seeking to sell its product in a new foreign market faces the choice of investing in its own sales operation or contracting with a local sales agent. 

The paper is organized as follows.  Section II presents a brief review of the relevant FDI literature.  Section III presents empirical evidence demonstrating the significance and pattern of distribution FDI.  In Section IV, two empirical models of the distribution FDI decision are presented and estimated.  Section V presents conclusions.[2]

Literature Review

Foreign direct investment occurs when a firm establishes or acquires an interest in a business enterprise in a foreign country, and exerts some degree of control over that enterprise.  Traditional theories of FDI focus on two primary incentives for establishing such operations:  a reduction in trade costs and a reduction in factor costs.   The vertical theory of foreign direct investment, introduced by Helpman (1984), suggests that firms locate different stages of their production process in different geographical locations to take advantage of factor cost differentials.    The horizontal model, developed by Markusen (1984) and others, suggests that firms will duplicate production facilities in various countries, in part to reduce trade costs.  Each of these models assumes that the foreign affiliate performs a manufacturing operation, since only then are the assumed cost savings realized.    Foreign direct investment is clearly preferable to exporting from the home country in these models.

Markusen, Venables, Konan, Eby, and Zhang (1996) integrate the horizontal and vertical approaches into a model with both trade costs and differing factor intensities, and determine under what conditions either type of FDI might arise.  In this model the firm is assumed to own certain firm-specific intangible assets that give it a cost advantage over foreign producers.  The existence of this “knowledge capital” gives rise to multi-plant economies of scale, since it can be transferred to multiple foreign locations at little or no cost.  Markusen, Carr and Maskus (2001) later test the predictions of the knowledge capital model and determine that both horizontal and vertical are FDI are sensitive to country characteristics such as market size, home and host country size differentials, and relative factor endowments.   Here again, the focus is on the firm’s location of production facilities.    

In the case of distribution FDI, the foreign subsidiary must purchase the product for resale, often from an affiliate in another country. Consequently, this type of foreign investment does not generate trade or production cost savings.  Perhaps for this reason, distribution FDI has received scant attention in the theoretical literature. An exception is a paper by Horstmann and Markusen (1995), in which the authors examine the MNE’s choice of entry mode into a new market.  They construct an agency model, in which a firm seeking to sell its product in a new foreign market faces the choice of investing in its own sales operation or contracting with a local sales agent.  Because the firm is new to the market, revenues are uncertain.  The arm’s-length contract affords the multinational low-cost access to information about local market characteristics, including whether or not the market is large enough to support a subsequent direct investment.  However, the contract is costly, since the firm must cede a portion of its rents to the agent.  

The authors assume a firm (the MNE) producing in its home country that wishes to sell an established product in a foreign market, where it will not face any competition.  The MNE has no experience selling in this market and chooses between establishing a wholly-owned sales subsidiary and contracting with a local sales agent.  The market is populated by identical customers with demand characteristics that are known to both the MNE and to any potential agent.  However, while the local agent knows the size of the market, the MNE does not.  The demand for the MNE’s product is, in part, a function of sales effort.  The efficiency of the agent’s sales force (i.e. the number of customers produced per unit of sales effort) may be higher or lower than that of the MNE’s own sales force.  If the MNE establishes its own sales subsidiary, cost includes a one-time set-up cost and ongoing per period fixed costs.  If instead the MNE contracts with an established local sales agent, there are no set-up costs, but there are ongoing per period fixed costs.

The firm’s choice of entry mode depends upon the size of the returns to information gathering relative to the size of the rents extracted by the agent, and upon whether the agent is more or less efficient at selling the product than the MNE.  The authors conclude that multinationals are more likely to opt for immediate FDI the larger the expected market, and the less variable are expected profits.   The authors also suggest that a multinational that initially chooses a contract arrangement will use a shorter contract when, among other conditions, its sales force is more efficient than the agent’s.  The firm will again migrate to owned sales operations should the market prove to be large.  Thus the decision to enter via foreign direct investment or to export to a local agent rests on market size, variability in expected revenues, and relative sales force efficiency, among other factors.  

Hanson, Mataloni and Slaughter (2001) examine distribution FDI from an empirical perspective.  They explore the US multinational’s choice between what they term “production-oriented” and “distribution-oriented FDI” to determine if and how it is related to host country characteristics.  They hypothesize that MNEs prefer distribution FDI to manufacturing FDI in higher-tax countries because the benefits of deferring US income taxes on income earned would be smaller than on income earned in lower-tax countries.  They find that US multinationals tend to engage in distribution FDI in higher-tax countries and in host countries located further from the United States.  Since their focus is on the MNE’s choice between production-oriented and distribution-oriented FDI, Hanson et al implicitly assume that exports and distribution FDI represent similar strategies.  Therefore, their results cannot be used to confirm or contradict the theoretical predictions of Horstmann and Markusen.

Blonigen (2005) reviews the empirical literature on determinants of FDI.  He identifies host country institutional quality as an important factor in attracting FDI, particularly for developing countries.  He points out, however, that most measures of institutional quality are composites and have questionable cross-country comparability.  Moreover, since institutions tend to be persistent, researchers can expect minimal in-country variation over time, rendering the inclusion of institutional variables in empirical analyses problematic.   Busse and Hefeker (2005) study the impact of a variety of institutional factors on FDI.  They find that host country investment climate, democratic accountability, law and order, and quality of the bureaucracy significantly influence the investment decisions of MNEs. 

Chakrabarti (2001) reviews empirical studies of the determinants of FDI and examines the robustness of partial correlations between levels of FDI and a number of economic indicators.[3]  Using Extreme Bound Analysis (EBA), the author determines which of the explanatory variables in existing studies are robust to changes in the conditioning information set.  Among many other variables, he examines two measures of market size, per capita GDP and total GDP.  His analysis suggests that per capita GDP is a more robust measure of market size than GDP.   He concludes that the significance of other variables, such as tax and political stability, are highly sensitive to cross-country model specifications.

This paper contributes to the existing literature in several ways.  First, it focuses on the importance and motivation for distribution FDI, which has been largely ignored in traditional horizontal and vertical models.  In particular, it answers the question: “Why do firms choose to establish wholesale trade affiliates in foreign markets rather than exporting to those markets, when the establishment of such operations will not reduce trade or factor costs?”[4]  Second, the paper extends the knowledge capital model put forth by Markusen et al.  If sales expertise and efficiency represent a firm-specific intangible asset that can be easily transferred to multiple locations, then the advantages and FDI incentives conferred by knowledge capital may extend to foreign distribution operations.  The paper also contributes to the existing literature by testing the predictions of the agency model proposed by Horstmann and Markusen.  Unlike the work by Hanson et al, the empirical analysis explicitly models the MNE’s choice between exporting and distribution FDI.  Finally, the paper connects to the existing empirical literature, by testing the influence of some of the commonly accepted ad hoc determinants of manufacturing FDI on distribution FDI. 

Evidence of Distribution FDI

For purposes of this paper, distribution FDI refers to the establishment of a foreign affiliate that engages exclusively in distribution, or sales, activities.  Distribution affiliates purchase finished goods from the parent or other affiliates, and resell them in the host country or other foreign markets.  This section examines the significance of this type of foreign direct investment activity, and discusses national and regional patterns.  Table 1 presents data on the number of US MNE distribution and manufacturing affiliates worldwide.

Table 1  Number of Distribution and Manufacturing Affiliates5
 

                             Source:  Bureau of Economic Analysis

Table 2 presents data on sales by foreign distribution and manufacturing affiliates of US MNEs.

Table  2  Sales by Distribution and Manufacturing Affiliates
 

                                   Source: Bureau of Economic Analysis

From 1984 to 1999, both the number and the sales of US MNE distribution affiliates increased as a percentage of total affiliates and total sales.  These data suggest that distribution affiliates represent a continuous and important part of the FDI picture. 

Empirical Analysis

This paper borrows from the Horstmann and Markusen approach, and focuses on several supply and demand-side factors that may influence the MNE’s decision to establish wholesale distribution subsidiaries in foreign markets, rather than contracting with sales agents in those markets.  The empirical analysis will focus on distribution FDI by US multinational enterprises.  I first test Horstmann and Markusen’s proposition that market size influences the choice between distribution FDI and exporting through a local sales agent.  Consistent with Chakrabarti’s findings, the model will include per capita GDP as the measure of market size.  Host country corporate income tax rates are included in the analysis in order to determine how their influence on the choice between distribution FDI and exports compares to the findings in Hanson et al.[6] 

In an effort to capture sources of expected revenue variability in the foreign market, I include a composite measure of the host country’s investment climate.  After examining several cross-country measures of economic and political risk, the Heritage Index of Economic Freedom was selected.  This index includes country-specific institutional factors such as the level of corruption, trade barriers, property right protections, and the level of regulation.  The index also includes a score for FDI; this component was removed to avoid biasing the results of the empirical analysis.  The rationale for this variable is that the lower the level of economic freedom in the destination market, the greater the uncertainty associated with future revenues, and the less likely the multinational will be to opt for distribution FDI over arm’s-length sales contracts with local agents.[7]   Note that a low index score indicates a high level of economic freedom.  Consequently, the coefficient of this variable is expected to be negative. 

Basic Model Specification

In order to examine the role of host country characteristics in determining the location of distribution FDI, the analysis relies on the basic empirical framework used in many related studies in the existing literature.[8]  Specifically, I use a panel comprised of BEA and country-level data from 1994 and 1999 to estimate the following model:[9]

ln((DFDI)_it /(EXPORTS)_it) = β₀ + β_1*ln(PCGDP_it) + β_2*ln(TAX_it) + β_3*ln(FREEDOM)_it + μ_it (1)

In equation (1), DFDI represents sales by US parent companies to distribution affiliates for resale without further manufacture; EXPORTS reflects US exports to non-affiliates[10] and is a proxy for sales to unrelated local agents; PCGDP refers to real per capita GDP (measured in 1995 dollars); TAX refers to the host country corporate income tax rate; and FREEDOM represents the country-specific Heritage Index score of economic freedom.  The subscripts i and t refer to country and year, respectively.  The Data Appendix describes these data and their sources in detail.

The dependent variable represents a firm’s decision to sell to a foreign market through its distribution affiliate rather than exporting to a local sales agent; its specification captures the differential effect of the independent variables on distribution FDI relative to exports and controls for the impact of any unobserved variables that affect both distribution FDI and exports the same way.

I estimate Equation (1) using OLS.[11] Due to the small sample size, the model does not include a set of country fixed effects.  However it does include a dichotomous variable to reflect year fixed effects.  The results are presented in Table 3.
 

Table 3.  Regression Results, Ratio of DFDI to Exports
 

Notes:  Constant and year fixed effects are unreported.  Robust standard errors are in parentheses.  *** and ** indicate significance at the .01, and .05 levels, respectively.

Consistent with the prediction of the Horstmann and Markusen theoretical model, the results suggest that MNEs are more likely to opt for distribution FDI the larger is the host country market, as measured by average income.  Specifically, the model predicts that a 10 percent increase in per capita income will increase the ratio of distribution FDI to exports by approximately 3 percent.   The tax variable coefficient is negative, but not significant, suggesting that US MNEs are not heavily influenced by corporate income tax rates in deciding between exports and distribution FDI.   The coefficient of the economic freedom variable is significant and negative, indicating that MNEs are more likely to export to countries with less economic freedom than to establish distribution subsidiaries in those countries.  Specifically, a 10 percent drop in the index score can be expected to reduce the ratio of distribution FDI to exports by approximately 20 percent.

Alternative Model Specification

Horstmann and Markusen also consider sales force efficiency in their theoretical treatment of the MNE’s foreign market entry choice.  Sales force efficiency is defined as the units of sales effort required to generate customers.  They conclude that, ceteris paribus, the more efficient the MNE’s sales force is relative to that of the local agent, the more desirable is foreign direct investment.  

Since measuring sales force efficiency is problematic, the empirical model below approaches the issue from a slightly different perspective.  It assumes that the more complex the selling process, the more likely is the MNE to have an efficiency advantage over the local agent and, consequently, the more likely is distribution FDI.  Unlike the agent sales force, the MNE sales force will be dedicated exclusively to selling the product and will likely have experience with prior product versions.  Moreover, the costs of training the local sales agent are likely to be higher as the selling process becomes more complex.  To the extent that revenue generation requires a long and/or complex selling process, establishment of a wholesale distribution affiliate may represent a profit-maximizing alternative to exporting through a local sales agent.  In addition, sales expertise or efficiency can be considered a firm-specific intangible asset that can be transferred at low cost to multiple distribution locations.  The more complex the selling process, the more likely that the process is valuable to the firm and that sales expertise represents a valuable intangible asset.  Consistent with the knowledge capital model, as the value of this expertise to the firm rises, the firm may be more likely to engage in distribution FDI than to export the good.  This prediction has interesting empirical implications since it suggests that the choice of exporting through an agent or establishing a distribution affiliate depends not only on host country characteristics but on industry characteristics as well.  The alternative empirical model, then, incorporates selling effort as an industry-specific variable.

Ideally, selling effort would be captured by comparing ratios of selling expenses to revenues across industries.  However, public companies do not report these data. Ratios of selling, general and administrative expenses (SG&A) to sales represent the narrowest classification available by industry.  While intra-industry variations in this ratio are commonly understood to reflect cross-firm efficiency differences, cross-industry comparisons should more closely reflect differences in selling effort.[12]  Moreover, there is no reason to expect that firms that establish distribution subsidiaries will be burdened with higher SG&A to sales ratios than their exporting peers, since the exporting firms’ payments to local sales agents will be included in this ratio.  Consequently, if there is a correlation between the SG&A to sales ratio and distribution FDI, the direction of causality can be interpreted to flow from the former to the latter variable.  Finally, there is precedent for viewing SG&A ratios as a basis for sales expertise and/or value.  Fu and Lev (2003) refer to SG&A as an intangible value driver that can create economic benefits for the firm and compare it to the intangible value created by R&D expenditures.    Table 4 presents SG&A to sales ratios for 1994 and 1999.

Table  4:  Selling, General & Administrative Expense to Sales Ratios by Industry
 

                           Source:  Compustat

Note that the drug and professional and commercial equipment sectors have the highest SG&A to sales ratios, while the ratios for metals and paper products are relatively low.  The cross-industry differences in these ratios are consistent with a sales effort interpretation, since pharmaceutical and computer sales are likely to require more expertise and training than paper products and metals. 

Ideally, the sales variable would be added to the original model.  However, the BEA does not provide industry-specific wholesale affiliate sales data.  Consequently, the regression equation must be re-specified, using a different dependent variable.     The alternate model relies on the number of distribution affiliates in each industry (in each host country) as the dependent variable.

The alternate empirical model is specified as follows:

DISTAFF_it = β₀ + β₁*lnSGARatio + β_2*ln(PCGDP_it) + β_3*ln(TAX_it) + β_4*ln(FREEDOM)_it + μ_it (2)

where DISTAFF represents the number of distribution affiliates by sector in each country in each year (1994 and 1999),  and SGARatio represents the mean ratio of selling, general, & administrative expenses to sales for each industry.  The remaining variables are specified in the same way as in the basic model.  Note, however, that this model tests the hypothesis that the number of distribution affiliates will be positively influenced by the level of sales effort in addition to host country characteristics.  It does not explicitly test the MNE’s choice between exporting through an agent and distribution FDI.   Consequently, variable signs and coefficients are potentially subject to different interpretation.

Because the dependent variable is now in count form, OLS regression analysis is no longer appropriate.  The dependent variable does not exhibit the characteristics necessary to allow a Poisson distribution, since it violates the Poisson assumption of equality between mean and variance.  The empirical analysis relies instead on a negative binomial-based maximum likelihood model.[13]  Since the data set now includes a large number of observations, dummy variables can be included to reflect country fixed effects.

The negative binomial analysis is presented in Table 5. 
 

Table  5:  Negative Binomial Analysis
 

Notes:  Constant, country and year fixed effects are unreported.  Robust standard errors are in parentheses.  ***and ** indicate significance at the .01and .05 levels, respectively.

The results indicate that host country market size positively influences the establishment of distribution affiliates.  Specifically a 10 percent increase in per capita income is expected to increase the distribution affiliate population by approximately 14 percent, holding other variables constant. Again, the coefficient of the tax variable is not significant, suggesting that tax rates do not influence the MNE’s decision to engage in distribution FDI.  Interestingly, the coefficient on the economic freedom variable is no longer significant.[14]    This result may be due to the specification of the dependent variable.  In the earlier model, where the dependent variable represented the MNE’s choice between exporting and distribution FDI, economic freedom was highly influential.  Here, however, the dependent variable measures only the MNE’s decision to engage in distribution FDI.  The MNE’s alternatives could now theoretically include manufacturing FDI, exporting or simply not entering the market.  While greater levels of economic freedom in the host country would be expected to encourage distribution FDI over exporting or no market entry, it could be expected to discourage distribution FDI relative to manufacturing FDI.[15]  These two effects may be counteracting one another in the analysis.

The focus of this model, of course, is on the selling effort variable, which exerts a positive and highly significant influence on the number of distribution affiliates established in each country.  This result suggests that the more complex the sales effort, the more likely the MNE is to opt for distribution FDI.  Specifically, a 10 percent increase in the ratio of SG&A to sales is expected to increase the number of distribution affiliates by more than 7 percent, holding other variables constant. This result is consistent with Horstmann and Markusen’s theoretical prediction that the MNE is more likely to opt for an owned sales operation the more efficient is its sales force relative to that of a local agent.  It is also consistent with the proposition that knowledge of complex selling processes represents an intangible asset that can be transferred at low cost to multiple locations, thereby encouraging the establishment of distribution affiliates.

Conclusion

This paper contributes to the literature by identifying the importance of distribution FDI to overall MNE activity and exploring its determinants.  It first tests the theoretical proposition that distribution FDI will be greater, the larger the host country market and the higher the variability in expected revenue.   Results indicate that US MNEs are inclined to choose distribution FDI over local sales agents the larger the host country market and the greater the degree of economic freedom in the host country. 

The paper then tests a second proposition, namely that firms are more likely to engage in distribution FDI the higher the level of sales effort associated with the product.  Negative binomial regression analysis using count data on the number of distribution affiliates supports this prediction.  Results suggest that the number of distribution affiliates is higher in those sectors where selling effort is greater.  This latter result also connects distribution FDI to the knowledge capital model, where sales expertise represents a valuable intangible asset that can be used across distribution affiliates.

 

I am grateful to Raymond Mataloni at the Bureau of Economic Analysis for valuable clarification of BEA data, Brittany Savory for research assistance, and several anonymous referees for helpful comments on earlier drafts.

 

References
 

Beach, William W. and Marc A. Miles.  “Explaining the Factors of the Index of Economic Freedom, Chapter 5, www.heritage.org/research, accessed October 2006.

Blonigen, Bruce.  “A Review of the Empirical Literature on FDI Determinants,” presented at the International Atlantic Economic Society session of the Allied Social Sciences Association Annual Meetings, January 2006.

Busse, Mathias and Carsten Hefeker.  “Political Risk, Institutions and Foreign Direct Investment.” Hamburg Institute of International Economics Discussion Paper 315, April 2005.

Carr, David L.; Markusen, James R.; and Maskus, Keith E. “Estimating the Knowledge-Capital Model of the Multinational Enterprise.”  American Economic Review, June 2001, pp.693-708.

Chakrabarti, Avik.  “The Determinants of Foreign Direct Investment:  Sensitivity Analyses of Cross-Country Regressions.  KYKLOS, Vol. 54-2001, Fasc.1, 89-114.

Hanson, Gordon H.; Mataloni, Raymond J., Jr.; Slaughter, Mathhew J. “Expansion Strategies of U.S. Multinational Firms,” National Bureau of Economic Research (Cambridge, MA) Working Paper No. 8433, August, 2001.

Helpman, Elhanan.  “A Simple Theory of International Trade with Multinational Corporations.”  Journal of Political Economy, June 1984 92(3), pp. 451-471.

Horstmann, Ignatius J. and James R. Markusen. “Exploring New Markets:  Direct Investment, Contractual Relations and the Multinational Enterprise,” National Bureau of Economic Research (Cambridge, MA) Working Paper No. 5029, February 1995.

Gu, Fen and Baruch Lev  “Intangible Assets: Measurement, Drivers, Usefulness.”  Working Paper 2003-#5, Boston University School of Management Accounting, 2003.

Markusen, James R.; Venables, Anthony J.; Konan, Denise Eby.; and Zhang, Kevin H.  “A Unified Treatment of Horizontal Direct Investment, Vertical Direct Investment, and the Pattern of Trade in Goods and Services,” National Bureau of Economic Research (Cambridge, MA) Working Paper No. 5696, August 1996.

Mataloni, Raymond, Jr. “A Guide to BEA Statistics on U.S. Multinational Companies.” Survey of Current Business, 75, 1995, pp.38-55.

U.S. Bureau of Economic Analysis. U.S. Direct Investment Abroad:  Final Results from the 1999 Benchmark Survey.  Washington, DC:  U.S Government Printing Office, April 2004.

U.S. Department of Commerce, Bureau of Economic Analysis, International Investment Division.  U.S. Direct Investment Abroad:  1994 Benchmark Survey, Final Results.  Washington, DC:  US Government Printing Office, May 1998.

U.S. International Trade Commission.  USITC Interactive Tariff and Trade Dataweb. http://dataweb.usitc.gov, retrieved July 2004.

World Bank Group.  World Development Indicators Database, http://publications.worldbank.org/subscriptions/WDI, retrieved July 2004.

Appendix A

BEA Data

Data on operations of foreign affiliates of U.S. multinational corporations were obtained from the Bureau of Economic Analysis (BEA) surveys of U.S. direct investment abroad.[16]  The surveys collect detailed data on the operations of US companies and their foreign affiliates.  Benchmark surveys are conducted every five years, beginning in 1982 and collect more data items than annual surveys.  This paper relies on benchmark survey results for 1994 and 1999.  Data are collected for every enterprise that was a U.S. parent or foreign affiliate at the end of its 1994 and/or 1999 fiscal year.

This paper relies on data for non-bank majority-owned foreign affiliates of U.S. parents.  A majority-owned affiliate is one in which the US parent’s (or parents’) ownership interest exceeds 50 percent.  Each foreign affiliate is classified according to its country of operation and its industry.  Affiliates are classified according to industry using the BEA’s International Survey Industry (ISI) classification system.  For the 1999 Benchmark Survey, the ISI system is based upon the North American Industry Classification System (NAICS).  The 1994 Survey classifies affiliates according to two- and three-digit ISI codes based upon the U.S. Standard Industrial Classification (SIC) codes.

Survey data are valued in current year prices and exchange rates.  Therefore, changes in foreign affiliate data over time may reflect nominal as opposed to real differences in the underlying variables.  Moreover, cross-country comparisons may be subject to measurement error to the extent that market exchange rates do not reflect the relative purchasing power of the underlying currencies.

Heritage Index of Economic Freedom

The Index of Economic Freedom rates countries based on the quality of their political, legal and economic institutions.  The index, published by the Heritage Foundation and the Wall Street Journal, presents an annual composite rating for 161 countries, based on 50 variables in the following categories:

• Trade policy

• Fiscal burden of the government

• Degree of government intervention in the economy

• Monetary policy

• Foreign direct investment and other capital flows

• Banking and finance