Shabtai Donnenfeld 
York University 

 Working Paper No. 95-6 

Table of Contents

• Abstract 
• 1. Introduction 
• 2. The Model 
• Equation 1 
• Proposition 1 
• 2.1 Equilibrium Analysis 
• Equation 2 
• Equation 3 
• Equation 4 
• Equation 5 
• Proposition 2 
• Equation 6 
• Proposition 3 
• Proposition 4 
• Proposition 5 
• 2.2 Discussion 
• 3. Empirical Investigation 
• Equation 7 
• 3.1 Description of Data 
• 3.2 Empirical Results 
• 4. Concluding Remarks 
• Appendix A 
• Proof of Proposition 1 
• Equation A1 
• Equation A2 
• Proof of Proposition 2 
• Proof of Proposition 3 
• Equation A3 
• Proof of Proposition 4 
• Proof of Proposition 5 
• Properties of functions hatI(t), tidleI(t), I*(t) 
• References 
• Table 1 
• Table 2: Descriptive Statistics of the Variables in the Sample 
• Table 3: Correlation Matrix of the Variables in the Sample 
• Table 4: Tobit Results for the Period 1984 - 1991 (Investment by country -- full sample of countries) 
• Table 5: Tobit Results for the Period 1984 - 1991 (Investment by country -- four largest countries) 
• Table 6: Tobit Results for the Period 1984 - 1991 (Total industry investment by Japan 
• Footnotes 

During the last decade the extent of inward foreign direct investment (FDI) in the U.S. has increased dramatically. An important aspect of industries where most FDI takes place is their high level of concentration. In this paper we investigate how strategic interactions among domestic and foreign producers influence the structure of the industry. Especially, we focus on the foreign firms choice between exporting versus servicing these markets by investing in a plant in located in host country when is facing competition from a domestic oligopoly. The empirical investigation reveals that the relationship between FDI and tariffs is not as simple as previously thought. In highly concentrated industries, where strategic behaviour may play an important role, high tariffs rather than low tariffs may lead to less FDI and more imports. 

 JEL Classification - F21 

1. Introduction

 In this paper we model two concurrent aspects of competition: actual competition between domestic oligopolist and potential competition between the non-cooperative domestic oligopoly and the foreign firms. This model attempts to capture an important feature of many industries in the US, such as the automotive, electronics, steel, chemicals, aluminum sheet and foil, petroleum refining, communications, etc., where domestic oligopolist are exposed to foreign competition from Japanese, EC and Canadian MC's. The two dimensional type of competition has important implications for the mode of penetration of foreign MC's in the US market. Foreign firms may serve the US market either via exports from their source countries or by setting up plants near the markets they intend to penetrate. Domestic firms which face competition from foreign firms in their home markets, have a stake in the actions that foreign firms may take with regard to the mode of supplying these markets. The nature of competition between the domestic and the foreign firm will depend on the foreign firm's choice of location of its production facilities. In addition, trade restrictions such as tariffs, quotas and local content requirements also can affect the choice of the location of production facilities of international firms. Plant location may be a mean of avoiding tariffs, quotas or local content requirements imposed by foreign governments.[3] 

 This paper explores how the interplay between the size of tariffs and non-tariff barriers imposed on imports, the size of costs involved in setting up plants in foreign countries, the size of the market and the degree of industry concentration affect the strategies of domestic firms and foreign firms. When facing potential competition from foreign firms, the domestic firms may engage in strategic behaviour in order to prevent entry of foreign firms in their domestic markets. Two kinds of entry prevention can be pursued by the domestic firms: deterring exports and preempting foreign direct investment. Both types of entry prevention are costly to the domestic firms since both involve over-production, relative to the production levels that would have been chosen when entry is accommodated. The strategy chosen by the domestic firms with regard to entry of the foreign firm ultimately will determine how the foreign firm will service the domestic market. [4] 

 We show that strategic interactions between domestic and foreign firms may give rise to different equilibria depending on the level of tariffs. These equilibria include exports accommodation (EA), FDI accommodation (IA), FDI deterrence (ID), exports deterrence (ED) and blockaded entry, (BE). The likelihood that a particular outcome will occur depends on the industry concentration, the size of the set up costs incurred while setting up plants in the foreign country, the size of the market, the level of tariffs and non-tariff barriers imposed on imports and the exchange rate. Specifically, we show that under plausible industry conditions lower tariffs may be conducive to more extensive foreign investment that is contrary to the conventional wisdom. 

 We test the predictions from the basic model on foreign direct investment and imports in the U.S. We conduct this test using a sample of 219 US manufacturing industries for the period 1984-1991 for which we have data. We find that the response to an increase in protection depends on the structure of the industry. In general, in industries with high level of protection there is significant foreign direct investment activity relative to imports. However, in highly concentrated industries, this relationship is reversed, and foreign direct investment is less significant than imports. These findings are consistent with the prediction that strategic considerations play an important role regarding the mode used by foreign firms to penetrate concentrated U.S. markets. 

 The paper is organized as follows. In section 2 the basic model is presented, and we derive predictions on the determinants of FDI activity in concentrated industries. Section 3 is devoted to the empirical investigation and includes the description of the data and the empirical results. The final section provides a few concluding remarks. 

2. The Model

• Export: To export the good to the home country's market, where imports are subjected to a per unit tariff t. 
• Foreign Direct Investment (FDI): To produce the good in a plant located in country H after incurring a fixed cost I. By serving the market with goods produced in this plant the foreign firm can circumvent the tariff. 
• To stay out of the country H market. 

 ¼[d] = x[d] ( A - X[H] - x[f] ) (1) 

where 

        n
X[H] = … x[d]
        d=1

We now examine the behaviour of firm the foreign firm. In particular, we shall determine the foreign firm's best response, given the level of the combined output of the domestic firms X[H], both in the exporting regime and in the case of FDI. 

 Exporting regime: Exports are prevented when the domestic firm choose the combined level of output X[H] „ A - t. Thus, exports are viable only if the value of tariffs t is less than A - X[H]. It is easy to verify that for X[H] < A - t the optimal quantity of exports of the foreign firm is given by x[f]^E(X[H],t) = (A - X - t) / 2 and its profits are ¼[f]^E = (A - X[H] - t)^2 / 4. 

FDI regime: In this regime, for any value of the domestic firms combined output X[H], the foreign firm optimal output produced in the plant which is located in country H is x[f]^I(X[H],I) = (A-X[H]) / 2 and its gross profits are ¼[f]^I = (A-X[H])^2 / 4. It is straight forward to verify that the foreign firm's profits when it engages in FDI are positive if the domestic firms combined output is not too large relative to market size, i.e, X[H] < A - 2 sqrt{I}. However, if X >= A - 2 sqrt{I}, FDI is deterred. 

So far we characterized the optimal strategy of the foreign firm, for given tariff rates and setup entry cost (t, I), when at most one option only is viable, either exports or FDI. It remains to consider the situation where both exports and FDI are viable, which will occur when the combined domestic output X[H] < min[A - t, A - 2 sqrt{I}]. In order to determine the foreign firm's optimal strategy now we need to compare its profits in the Exports and FDI regime. It turns out that if the level of combined domestic output which deters exports is less than the output which prevents FDI, then the foreign firm will choose the FDI option for all X[H] < A - t. However, when A - t > A - 2 sqrt{I}, the foreign firm will choose the FDI option if the level of X[H] is small but will prefer to export if the combined domestic output, X[H] is large. [6] 

The proposition below, the proof of which is relegated to the Appendix, characterizes the optimal response of the foreign firm to various levels of the domestic firms combined output.[7] 

Proposition 1. Let the values of tariffs t and the value of fixed cost I be given. 
(i) If X[H] >= max [A - t, A -2 sqrt{I}], the foreign firm stays out of the home country's market. 
(ii) Let t >= 2 sqrt{I}. If X[H] <= A - 2 sqrt{I}, the foreign firm will exercise the FDI option. 
(iii) Let t < 2 sqrt{I}. Define by X^0 the value of the combined domestic output which makes the foreign firm indifferent between FDI and Exports option. Then if X^0 <= X[H] < A - t, the foreign firm will exercise the exporting option and if 0 <= X < X^0 it will prefer FDI.  

2.1. Equilibrium Analysis

In either, the exports or the FDI regime, the best response of a domestic firm d to other domestic firms output X[-d] while forseeing the foreign firm's best response is obtained from maximizing the profits stated in (1) and is x[d] = A - X[-d] / 2. However, the equilibrium choices in the two regimes differ. In the Exports regime, while taking into account the response of other (domestic) firms, each domestic firm will produce A+t / n+1 so that the total domestic output is 

X^E = n(A+t) / n+1 (2) 

and each firm earns profits ¼[d]^E = (A+t)^2 / 2(n+1)^2. 

In the FDI regime each domestic firm will produce A / n+1 units of output yielding a total level of domestic output 

X^I = nA / n+1 (3) 

and earn profits %#188;[d]^I = A^2 / 2(n+1)^2. When the foreign firm does not enter the home country's market, the equilibrium level of production of each domestic firm is the standard Cournot equilibrium output, A / n+1 units, and the total domestic output in this case is 

X^C = An / n+1. (4) 

Each domestic firm profits now are ¼[d]^C = A^2 / (n+1)^2. 

We turn now to examine the strategies that the domestic firms will select in equilibrium. As one may expect the equilibrium strategies are affected by the interplay between tariffs, the level of fixed costs and the size of the country H market. Specifically, the outcome depends on the size of the export deterring level of output A-t relative to the FDI-deterring level of output A - 2 sqrt{I}. Thus, for each level of setup cost I there exists a corresponding tariff rate such that A - t = A - 2sqrt{I} or equivalently, 

t = 2 sqrt{I}. (5) 

• If t >= 2 sqrt{I}, then from the vantage point of the foreign firm, as stated in Proposition 1, FDI dominates the exports option. 
• If t < 2 sqrt{I} both FDI and exporting options are viable. This leads to several interesting distinctive equilibrium outcomes. 

First we consider the case when barX^E >= A-t > X^C. Total domestic output in the Exports regime is large enough to deter exports, whereas the Cournot equilibrium output is not sufficient to deter exports. In view of the foreign firm's best response, neither X^C nor barX^E will not be optimal since by choosing the former exports will occur whereas the latter exceeds the level of output that is needed to prevent exports. Thus in equilibrium the domestic firms will produce the minimal level of output that is needed to prevent exports, that is the limit output. We thus have 

Proposition 2. Deterred Exports and Blockaded FDI: If A / n+1 > t >= A / 2(n+1) the domestic firms will deter exports by jointly producing A-t units of output.  

Next we examine the case when X^0 >= barX^E < A-t. The total domestic output produced in the Exports regime allows exports and by Proposition 1, the foreign firm now prefers exports over the FDI. We note that the combined domestic output which renders the foreign firm indifferent between FDI and exporting option, 

X^0 = A - t/2 - 2I/t (6) 

is decreasing in I. Let hatI[t] be the value of the fixed cost I, such that the domestic output in Exports equilibrium X^E renders the foreign firm indifferent between exports and FDI, i.e., X^0(t,I) = X^E(t). [9] Thus we have 

Proposition 3. Exports Accommodation and Blockaded FDI: If t < A / 2n+1 and I >= hatI(t), the domestic firms will accommodate exports and each will produce A+t / n+1 units of output.  

Finally, when barX[H]^E < X[H]^0(t,I) the total domestic output in the Exports regime makes both exports and FDI viable. Although the foreign firm now prefers the FDI option over exports what will emerge in equilibrium depends on whether the setup cost is ``low'' or ``high'' i.e., I < tildeI or I > I* (t), respectively. [10] 

To show this suppose that FDI is the equilibrium outcome. As in the earlier examined FDI game each domestic firm produces A / n+1 units of output. To determine whether the FDI accommodation is an equilibrium outcome, we inquire whether it is worthwhile for any domestic firm d to unilaterally deviate from FDI accommodation and preempt FDI. Such deviation involves an increase in its output that is sufficient to raise total domestic production up to the level X[H]^0. For each t >= A / 2n+1 we define by tildeI(t) the value of entry cost I which makes each domestic firm indifferent between FDI accommodation and the (unilateral) increase of its output that leads to FDI deterrence. 

Proposition 4. FDI Accommodation: If t < A / 2n+1 and I < tildeI(t), the domestic firms will accommodate FDI and each will produce A / n+1 units of output.  

Let us turn to the second case where the domestic firms accommodate exports and choose to prevent FDI. Since in Exports equilibrium the total domestic output X^E is smaller than X[H]^0, for FDI to be deterred this output needs to be X[H]^0. Denote by I*(t), the value of setup costs which makes every domestic firm indifferent between preventing FDI by jointly producing X^0 and a unilateral reduction of its output that leads to FDI. We thus have 

Proposition 5. Exports Accommodation: If t < A / 2n+1 and I >= I*(t), the domestic firms will accommodate exports by jointly producing X[H]^0 units of output.  

2.2. Discussion

When the tariff is sufficiently low, (column (a)), the domestic firms accommodate exports. Over this range of tariffs, the domestic firms combined output increases as the tariff rises whereas the foreign firm's exports decline. When the tariff reaches a critical level the domestic firms produce a combined level of output X[H]^0(t,I), that although deters FDI it still accommodates exports, (column (b)). The incentive to deter FDI stems from the following consideration. Tariffs at any level, endow the domestic firms with a competitive advantage, since they raise the foreign rival total marginal cost relative to the case where the foreign firm supplies the market with goods produced behind the tariff wall. To gain the competitive advantage the domestic firms need to produce a sufficiently large level of output such that the foreign firm will not find it desirable to jump the tariff wall. As tariffs further rise the foreign firm best response is to build a foreign plant and thus jump the tariff wall, (column (d)). Proceeding to higher levels of tariffs, the equilibrium outcome is somewhat surprising. Now the domestic firms find it desirable to regain the competitive advantage relative to the foreign firm by producing the limit output, X[H]^0 which deters FDI, (column (f)) but they accommodate exports. 

Finally, when tariffs are very high, the foreign firm's entry via FDI or exports is blockaded, (column (i)). 

The upshot of this analysis is that one needs not expect to find a simple positive correlation between tariffs and the propensity to engage in foreign direct investment. We have shown that for some range of tariffs, set up cost, home country market size, and the number of domestic firms, as tariffs rise FDI will occur; as tariffs rise further exports will occur. 

The model that we presented so far focused solely on the effects of tariffs and ignored other types of trade barriers. Some non-tariff barriers, such as import quotas also affect the foreign firms choice between direct investment and exports.[11] The more restrictive import quotas are the more likely is that foreign firms will engage in direct investment. Other forms of non-tariff barriers, such as harmonization standards and local content requirements may differ from quotas in their effects on the choice between exports and foreign direct investment. These barriers to trade often lead to higher total marginal cost of servicing the foreign market. Therefore, these barriers affect FDI and exports in a fashion that is similar to the effects of tariffs. The analysis and conclusions regarding the effects of tariffs are readily applicable to some NTBs as well. 

3. Empirical Investigation

To test the hypothesis of a non-monotonic relationship between foreign direct investment and imports and trade barriers, ideally one would like to obtain observations of the industry equilibrium levels of foreign activity over time. FDI and imports are expected to vary as tariffs and the fixed costs of establishing a plant in the industry vary. If data were available we would have liked to design a test for each industry separately. In practice it is difficult to obtain such data; first tariffs changes over time are rare and and second it is difficult to discern how long it take for FDI flows and imports to adjust to their new equilibrium in response to a tariff change. 

Since we cannot conduct the test for each industry separately, we instead test the hypothesis using a cross section of manufacturing industries, in the United States. The theory suggests that tariffs (t) and fixed cost of investment (I) determine the extent of FDI and/or import activity in an industry. We use the ratio of FDI over imports in the industry as the dependent variable. 

The null hypothesis will be the non-monotonic relationship between tariffs and the ratio of FDI over imports in the industry while controlling for other factors such as concentration ratios, changes in bilateral exchange rates, fixed costs of investments, and non-tariff barriers. 

The empirical model is 

FDI / IMP = f(I, t, CR, EXRATE, NTB) (7) 

We expect that the ratio of foreign direct investment to imports (FDI/IMP) in an industry is negatively related to the fixed cost involved in FDI (I). For a range of low tariffs, this ratio is positively related to tariffs (t), whereas for a range of higher tariffs the relationship should be negative in industries in which strategic considerations among producers are likely to be important. Strategic interactions among domestic producers in the industry is captured by CR and an interactive term, t*CR. The higher the degree of domestic concentration, CR, the lower is the expected value of FDI over imports. Non-tariff barriers (NTB) should be positively related to the level of foreign direct investment if they take the form of quantitative restrictions on imports. However, if the effects of NTBs is to raise the marginal cost of the exporting firms, their effect will be similar to that of tariffs.[12] We also incorporate an additional factor that was not explicitly included in the theoretical model, the exchange rate (EXRATE). A depreciation of the home currency raises the foreign firm's cost of exporting and thus will have a positive effect on the ratio of FDI over imports.[13] 

3.1 Description of Data

The hypothesis to be tested reflects the equilibrium situation in narrowly defined industries where interactions among competitors, domestic and foreign, are likely to be important. Therefore, we are interested in using the most disaggregated information available. The source of information on foreign direct investment is the International Trade Administration (ITA) of the Department of Commerce. The main advantage in using this data source, as opposed to the alternative source, the Bureau of Economic Analysis (BEA) data, is its higher degree of industry level disaggregation. The ITA reports foreign direct investment transactions according to the four-digit SIC industry classification.[14] ITA also reports the names of the foreign and the domestic companies involved in the transaction, and the country of origin. 

Out of the available four-digit manufacturing industries we selected a sample of 219 industries where most foreign direct investment activity took place.[15] From this sample we selected only the combinations of industry and country for which at least one FDI transaction was observed during the sample period, 1984-1991. Based on this selection procedure we identified 625 industry/country combinations that include FDI transactions from 19 countries. The sample includes a panel of annual observations for each industry/country pair for which positive FDI occurred at least in one year during the sample period. The FDI data therefore takes either the value of 0 if no investment occurred by a given country in that industry and year, or a positive value equal to the country's total value of investment. 

The import data consist of values of imports in each four-digit manufacturing industry. The value of imports in an industry are reported by country of origin. The source is the Census Bureau of the Department of Commerce. The import data is aggregated up to the four digit SIC classification. There are however some industries at the four-digit SIC classification for which import transactions are not reported. Due to this missing data on imports and tariffs the sample was reduced to 4111 observations. 

Tariffs are the ad-valorem tariff equivalent of duties collected in each industry. To compute tariff rates we used the ratio of duties collected over the total value of imports by country of origin for each industry.[16] We also computed an alternative measure for import tariffs; this is the ratio of tariffs collected over the value of dutiable imports for each industry. By construction, this measure always yields a higher tariff rate than the first one. Since the correlation between these two measures is very high, 0.93, we decided to use the first ratio only. The information on tariffs is annual and is available only for the period 1989-1991. Since there were no major changes in tariff rates during the sample period we used the 1989 tariffs as a proxy for prior years. 

The data on non-tariff barriers represent the percentage of all U.S. tariff categories included in a given four-digit SIC industry that had at the border some type of non-tariff restriction on trade during 1988. The source is the UNCTAD data base on trade restrictions. The original data consists of the number of tariff subcategories under each SITC (Revision 2) industry classification that had some sort of non-tariff restriction. We developed a method to convert the SITC data to the corresponding SIC classification.[17] These data have a number of important shortcomings. First, it only reports non-tariff barriers existent at the border. It includes the most common restrictions such as quotas, voluntary exports restraints, non-automatic licensing, harmonization standards, etc. However, it does not include other important barriers such as distribution restrictions, state monopolies, countervailing duties and anti-dumping provisions. Second, what is stated in the data is whether non-tariff restrictions exist or not; it does not state the extent to which these restrictions are binding.[18] Third, a NTB restriction can have very different effects when imposed on different industries. For instance, quotas on imports of steel in the U.S. existed in this period; however, according to the International Trade Commission, these quotas where not binding during 1991.[19] The NTB variable is an index that can take values ranging from 0 to 100, where 100 means all product categories in that 4-digit SIC industry were subject to NTB restrictions. 

To capture the degree of oligopolistic competition in the industry we use the 1987 industry concentration ratio for U.S. industries. The source of this information is 1987 Census of Manufactures. We employed the four and eight firm concentration ratios, and both measures yield very similar results. 

As a measure of the initial investment required in the industry we use plant size (I) defined as the ratio of total 1987 shipments in each industry over the total number of U.S. establishments that employ at least 20 employees. These data were obtained from the 1987 Census of Manufactures. Previous studies have found that the average size of a foreign entry in a domestic industry is higher than the average domestic entry. We decided to pick establishments which employ at least 20 employees, rather than all the establishments, in order to avoid possible distortions due to a disproportionate amount of small establishments in certain industries. Using the total number of establishments does not affect the results that we report later. 

Finally, the exchange rate, EXRATE, is the bilateral real exchange rate index between the U.S. dollar and the currency of the country from which FDI and imports originate. where 1985 is the base period. [20] The exchange rate is defined as the price of U.S. dollars in terms of foreign currency. Thus, a higher exchange rate means an appreciation of the dollar. 

In Table 2 we report some descriptive statistics of these variables. The average industry tariff for all observations in the sample is 4.05% while the average measure of NTB's is 18.62%. Only four industries in the sample have tariffs that are equal to zero while 77 industries have zero non-tariff barriers.[21] The highest ad-valorem tariff equivalent in the sample is 28.77%, and there are five industries that were subjected to 100% non-tariff barriers.[22] The four-firm concentration ratio in the sample ranges from 2% to 100% with a mean of 38.78%. The ratio of foreign direct investment to imports varies significantly across the sample and within an industry across years. Actual FDI occurred in slightly more than 30% of the industry/country/year observations; 1255 out of 4001 observations have positive values. The values of positive foreign direct investment by industry range from 0.991 to 8555 million dollars, with a mean annual value of 65.35 million dollars. 

In Table 3 we report the correlations between FDI, imports, the exchange rate and the two measures of trade protection. Tariffs and non-tariff barriers have a positive correlation, 0.37. The simple correlation between FDI and the two measures of protection is small and in both cases is negative. Imports and NTBs are positively correlated, 0.089, whereas the correlation between imports and tariffs is -0.07. FDI and imports are positively correlated, 0.132, suggesting that industries with higher imports also have higher FDI activity. The weak positive correlation between imports and FDI suggests that certain industries have a higher degree of international economic activity than others. 

Before we present the empirical results a number of issues regarding the sample selection must be addressed. We have restricted the sample to industries with high levels of FDI activity and imports. These industries reflect a wide range of U.S. manufacturing activity and do not form a random sample of manufacturing industries in the U.S. Therefore, any estimates obtained do not necessarily generalize to overall U.S. economic activity and should be interpreted with caution. Second, the theoretical model was based on the presumption that FDI and imports are substitute ways of serving foreign markets. Hence, we selected industries where both FDI and imports exist. Although, industry equilibria with zero export or zero FDI activity are also possible they have not been included in the sample. 

The information on the level of trade protection in an industry is constant in the sample. The information on tariffs and NTBs are available for each industry for a single year (1989 and 1990 respectively). Therefore, all variation in these variables will come from variability across industries. As stated above, this variation will suffice to identify the effect of trade barriers on the industry equilibrium given that other exogenous variables impacting the industry equilibrium are controlled for. 

3.2. Empirical Results

In Tables 4, 5, and 6 we report the results of the empirical analysis. In each table, we report the results of six different specifications of the regression. In the first three columns of each table the dependent variable is the ratio of FDI over imports. As defined, the dependent variable captures most directly the implications of the theoretical model, but has some shortcomings: it might not appropriately account for industry differences. Therefore, in columns (4)-(6) we redefine the dependent variable as the ratio of FDI normalized by domestic investment over imports normalized by domestic shipments for each of the 219 four digit SIC industries. Columns (1) and (4) in each table depict linear specifications of the regression specified in equation (6). In columns (2) and (5) we report the results for a more general specification that allows for a non-linear relationship between the dependent variable and barriers to trade. These columns include a quadratic specification of non-tariff barriers (NTB^2) and an interactive term to capture the relationship between tariffs and industry concentration (CR*t). Columns (3) and (6) are identical to columns (2) and (5) except that the concentration ratio (CR) is dropped. 

In Table 4 we report the results of the regression for the ratio of FDI over imports by industry and country of origin. Although the FDI data compiled by the International Trade Administration is very disaggregated it is not complete. The values of a significant number of FDI transactions that actually occurred are not reported. Therefore, our measure of FDI underestimates the amount of activity that actually took place. Also as explained above, the use of annual data might misrepresent the substitutability between FDI and imports. 

To address these concerns, we estimated the missing FDI values by computing the mean value of the available transaction in a given industry based on a five year window, centered around the year when the transaction took place. We assigned these estimates to transactions for which values were not available.[25] 

The coefficient of the exchange rate (EXRATE) has a negative sign and is always significant implying that an appreciation of the dollar is associated with less FDI. The negative correlation is consistent with previous findings that the depreciation of the dollar is associated with higher FDI in the United States.[24] 

The coefficient of industry concentration (CR) has a negative sign and is significant which suggests that FDI was less likely to take place in highly concentrated industries. This finding is at odds with some previous studies which suggested that high industry concentration is associated with high profitability and thus is conducive to more FDI. The coefficient of non-tariff barriers (NTB) is always positive and significant in the quadratic specification. The coefficient of (NTB^2) is negative and significant which confirms the prediction that the relationship between the ratio of FDI over imports and NTBs is non-linear. The tariff coefficient (t) is negative and sometimes significant (in the linear specification). The coefficient of the interactive term (CR*t) is negative and significant. The inclusion of this interactive term leads to a change in the sign of the tariff coefficient; it is now positive and sometimes significant (Columns (5) and (6)). 

These results are consistent with the theoretical explanation that higher tariffs lead to more foreign direct investment, although the level of foreign direct investment is lower in highly concentrated industries. For high tariffs and high concentration of domestic producers the level of imports is higher and the level of FDI is lower when compared with industries that have similar tariffs but lower concentration level of domestic producers. 

FDI activities in the U.S. were not evenly distributed across countries. A small number of industrialized countries undertook most FDI activities in the United States. To obtain more refined information on the behaviour of FDI originating from these countries we select a subsample that includes four countries which undertook most foreign direct investment (Japan, United Kingdom, Germany and Canada). These four countries captured 75% of the average annual value of FDI in the U.S.[25] 

In Table 5 we report the regression results for these four countries. In all the specifications presented in this table the coefficients of (I), (EXRATE), (t), (CR*t) and (NTB^2) are always significant. All the explanatory variables except (I) have a sign that is consistent with the predictions of the theoretical model. The coefficient of plant size, (I), is positive. The coefficient of (NTB) is still insignificant in the linear specification of the regression but it becomes significant once we include the non-linear term (NTB^2). The industry concentration (CR) coefficient is negative and significant except in Column (2) where the interactive term (CR*t) is incorporated in the regression. These results reconfirm the differential effects of trade restrictions on FDI and imports across various degrees of industry concentration. 

The data also reveals that FDI by Japanese companies in the U.S. mostly took place in highly concentrated industries. In these industries the average plant size is larger than the average plant size of all manufacturing industries. These facts are easily captured by the correlation between (FDI) and (CR) and (FDI) and (I). The respective correlation coefficients for Japan are 0.283 and 0.381. For the other three major countries the corresponding correlation coefficients are -0.039 and 0.206. Similarly, the correlation between imports and industry concentration is 0.426 for Japan and 0.130 for the other three countries; the correlation between imports (IMP) and plant size (I) is 0.908 for Japan and 0.279 for the other three large countries. These simple correlations reveal the relative importance of Japanese FDI and imports in highly concentrated U.S. industries where non trade related barriers to entry, as captured by plant size, are important. These are precisely the industry characteristics that best resemble the assumptions of the theoretical model and therefore, we expect that this will be reflected in the results of the test. Furthermore, the amount of Japanese FDI in the U.S. for the period 1984-1991 was larger than that of any other country.[26] Thus, we also test the predictions of the theoretical model with regard to FDI by Japanese firms. In Table 6 the regression results for Japan show that the coefficients of (I), (t) and (NTB^2) have the expected sign and are significant. The coefficient of (NTB) is always positive and is significant in the specifications that include the variable (NTB^2). Surprisingly, the coefficient of the exchange rate is always insignificant. [27] The coefficient of (CR) is always significant but its sign changes from negative to positive once the interactive term (CR*t) is also included. The change in the sign of the coefficient reflects the fact that when strategic aspects are controlled for, (captured by the term (CR*t)) Japanese FDI in the U.S. is more prevalent in concentrated industries. Finally, the interactive term (CR*t) is always negative and significant confirming the importance that strategic behaviour plays with regard to FDI and imports from Japan in the U.S. markets. 

There is a broader conclusion that emerges from the analysis conducted so far. The estimated coefficients of tariffs and the interaction between industry concentration and tariffs presented in Tables 5 and 6, enable us to state the following: In industries with low concentration of domestic producers the relationship between tariffs and the ratio of FDI over imports is positive; however, in highly concentrated industries the relationship between the ratio of FDI/Imports and tariffs is negative. The critical level of concentration that separates between industries where tariffs have a positive/negative effect on FDI varies across regression specifications. For the four largest countries, the critical values of these coefficients (reported in Table 5) vary from 35.6 to 54.8. For Japan the range of critical values is higher and varies from 50.2 to 71.2. 

Our analysis indicates that there is a negative relationship between tariffs and FDI in highly concentrated industries. Based on the above estimates, a four firm concentration ratio of 60 seems to be an appropriate lower bound of the range of highly concentrated industries. It is noteworthy that in the sample of 219 industries 17% of these industries have a concentration ratio that is higher than 60 percent. These industries account for almost a quarter of total FDI activity in U.S. manufacturing. 

4. Concluding Remarks

We tested this hypothesis using a sample of U.S. manufacturing industries and found that the nature of the relationship between the ratio of foreign direct investment over imports and tariffs depends on the degree of industry concentration. This relationship is negative in highly concentrated industries and positive otherwise. We also found that NTBs are positively related to the ratio of foreign direct investment over imports. 

These results confirm the differential effects of trade restrictions on FDI and imports across various degrees of industry concentration. Our findings differ significantly from those of previous studies that ignored the role that strategic considerations may have on the mode in which firms penetrate foreign markets. 

Appendix A

¼[f]^E - ¼[f]^I = 1/4 [(A - X[H] - t)^2 - (A - X)^2 + 4I] = 1/4 [(t^2 - 2t(A - X[H]) + 4I]. (A.1) 

The expression in (A1) is increasing in X[H] and iff t >= 2 sqrt[I] and X[H] >= A - 2 sqrt[I], this difference is non-positive and the foreign firm will prefer the FDI option. On the other hand when t < 2 sqrt[I], there exists a value X^0 = A - t/2 - 2I/t (whenever positive) that renders the foreign firm indifferent between exports and FDI, i.e., 

¼[f]^E (X[H]^0(.),t) = ¼[f]^I (X[H]^0(.),I) (A.2) 

Then if X[H] < X^0, the foreign firm will prefer the FDI option, and if X[H] >= X^0 it will prefer the exports option. 

Proof of Proposition 2: Since the combined equilibrium domestic output in the Exports game is X^E = (A+t) / n+1, the inequality X^E >= A-t yields t >= A / 2n+1. We shall show that in this case in equilibrium the domestic firms will deter exports by producing total output of A-t. 

Let each domestic firm produces x units of output. If X[H] = nx > A-t then t < A / n+1 implies that the n-tuple(x,...,x) is not an equilibrium; a slight change in the output of each firm would still deter exports and increase the profits of the deviating firm. If X[H] < A-t then exports are accommodated and in equilibrium the domestic output is X^E >= A-t. It follows that the n-tuple (x,...,x) is not an Exports equilibrium and a slight change in the output of each firm would still allow exports and increase profits. It follows, therefore, that in equilibrium the total output of the domestic firms X is A-t. 

Assume now that x = A-t / n. The n-tuple (x,...,x) will constitute an equilibrium if no domestic firm finds it beneficial either to increase its output while still deterring exports or to allow exports by reducing its output. Since the best response of any domestic firm to other firms' output (n-1)x in the Cournot game is [A-(n-1)x] / 2 which is less than x for x = A-t / n >= A / n+1, it follows that no domestic firm would benefit by increasing its output. Since the best response of firm d to the combined output (n-1)x in the Exports game is [A - (n-1)x + t] / 2 >= x, it follows that no domestic firm would benefit by reducing its output and thus accommodate exports. 

Proof of Proposition 3: Let t < A / 2n+1 and I >= hatI(t). The n-tuple of the domestic firms' outputs (barx[1]^E,..., barx[n]^E) is the only candidate for an equilibrium. The only threat of deviation from this n-tuple would be the willingness of one of the firms to produce a larger amount of output and thus deter exports. The minimal level of output that would guarantee that exports are deterred is hatx[d] = A-t - (n-1 / n+1)(A+t) = (2A / n+1) - (2nt / n+1). Let us first show that the profits of the deviating firm are decreasing for all x[d] > hatx[d]. For x[d] >= hatx[d], consider the function ¼[d] = x[d](A - X[-d]^E - x[d]) = x[d](A - [n-1 / n+1]A - x[d]), which is decreasing for x[d] >= A / n+1. However, since hatx = (2A - 2nt) / n+1 > A / n+1 when t < A / 2n+1, it follows that if firm d decides to deter exports, its optimal choice should be hatx[d]. It remains to show that the choice of hatx[d] does not benefit firm d. That is, 

(2A - 2n[t^2]) / n+1 <= (A+t)^2 / 2(n+1)^2 (A3)  

It is easy to verify that the inequality in (A3) holds since 

(A+t)^2 / 2(n+1)^2 - ([2At - 2n[t^2]2] / n+1) = [A - (2n+1)t]^2 >= 0. 

Thus, the n-tuple of the domestic firms' outputs (A+t / n+1, ..., A+t / n+1) is, indeed, an equilibrium. 

Proof of Proposition 4: Let t < A / 2n+1 and I < tildeI(t). We shall derive the conditions under which FDI accommodation emerges as an equilibrium outcome. As in the equilibrium of the FDI game discussed earlier the output of each domestic firm is A / n+1. The only threat of deviate from this n-tuple would be the willingness of one of the firms to produce a larger amount of output and deter FDI. The minimal level of output that would guarantee FDI deterrence is tildex[d](t,I) = X[H]^0(t,I) - [(n-1)A / n+1]. Let us first show that the profits of a deviating firm d are decreasing for all x_d > tildex[d](t,I), which would imply that if firm d decides to preempt FDI it should choose the output tildex[d](t,I). For x[d] >= tildex[d](t,I) we have 

¼[d] = x[d] [A - (n-1 / n+1)A - x[d] - 1/2(A - (n-1 / n+1)A - x[d] - t)] = x[d]/2 [(2A / n+1) + t - x[d] ], 

which is decreasing for x[d] >= (A / n+1) + t/2. Since tildex[d] = X[H]^0(t,I) - [(n-1)A / n+1] > (A + nt) / n+1, it follows that if firm d decides to deter FDI, its optimal choice should be tildex[d](t,I). To complete the proof of the proposition, it remains to compare the profit of the domestic firm d at the FDI equilibrium with the profits it earns when it unilaterally raises the total domestic output to X[H]^0(t,I) and thus it preempts FDI. From the definition of tildeI it follows that if I < tildeI(t), then each domestic firm will choose to accommodate FDI, which, indeed, is an equilibrium outcome. 

Proof of Proposition 5: Using the same line of arguments as in the proof of Proposition 3 it follows that when the domestic firms accommodate exports they produce the output X = X[H]^0. The n-tuple (y,...,y) where y = X^0 / n will constitute an equilibrium if no domestic firm finds it beneficial either to increase its output while still preventing FDI or to allow FDI by reducing its output. Since the best response of firm d to output (n-1)y in the Exports game is (A - (n-1)y + t) / 2 <= y for y >= A+t / n+1, it follows that no domestic firm would benefit by increasing its output. 

Since the best response of firm d to the combined output X[-d] in the FDI game is (A - (n-1)y) / 2, it follows that a domestic firm will benefit by allowing FDI if its profits are greater than when it prevents FDI. That is 

y(A - ny + t)/2 < [A - (n-1)y]^2 / 8 

Thus, a necessary and sufficient condition for firm d not to benefit from allowing exports is 

[ (n+1)A + 2t - sqrt[4(n+1)At + 4t^2] ] / (n+1)^2 <= y <= [ (n+1)A + 2t + sqrt[4(n+1)At + 4t^2] ] / (n+1)^2 

Since ny = X^0 = A - 2I/t - t/2, we can derive I*(t) as the minimal value of I such that for a given t the above inequality holds: 

I*(t) = [ 2(n+1)At - (n^2 + 6n + 1)t^2 - 2nt sqrt[4(n+1)At+ 4t^2] ] / 4(n+1)^2 

To complete the proof of this proposition, it remains to observe that for all I >= I*(t), the n-tuple (y,...,y) with y = X^0/n, constitutes an equilibrium in which the domestic firms accommodates exports. 

(i) hatI(t) > tildeI(t) > I*(t) for all t <= A / 2n+1. 

(ii) Each of the three curves hatI(t), tildeI(t) and I*(t) intersects the curve I = t^2 / 4 at two points, one is the origin t=0 and the other is hatt, tildet and t*, respectively, where A / 2n+1 = hatt > tildet > t* > 0. 

(iii) All three functions I = hatI(t), I = tildeI(t), and I = I*(t) are concave and have an interior maximum in the intervals [0, hatt], [0, tildet] and [0,t*]. 

To prove part (i) we note that, A - t/2 - 2I/t = n(A+t)/n+1, yielding 

hatI(t) = [2At - (3n+1)t^2] / 4(n+1). 

In addition, 

[ X[H]^0(t, tildeI(t)) - (n-1/n+1)A ] [ A - X[H]^0 (t, tildeI(t)) + t] / 2 = A^2 / 2(n+1)^2, 

which yields two roots: 

X{H]^0 (t, tildeI(t)) = [ 2nA + (n+1)t ± sqrt[ (n+1)^2 t^2 + (4n+4)At ] ] / 2(n+1). 

Since X[H]^0 (t, tildeI(t)) > X[H]^E(t) = ( n / n+1 )(A+t) it follows that 

X[H]^0 (t, tildeI(t)) = [ 2nA + (n+1)t + sqrt[ (n+1)^2 t^2 + (4n+4)At ] ] / 2(n+1). 

By using the condition that the foreign firm is indifferent between exports and FDI as stated in (A2), we obtain 

tildeI(t) = [ 2At - 2(n+1)[t^2] - t sqrt[ (n+1)^2 t^2 + (4n+4)At ] ] / 4(n+1). 

It is easy to see that hatI(t) >= tildeI(t) for all t. Moreover, the equation tildeI(t) = I*(t) has two roots, t=0 and t = [ (9n^2 + 6n + 1) / 4n(n^2-1) ]A > (1 / 2n+1) A. Since tildeI'(0) > I*(0), it follows that tildeI(t) > I*(t) for all t < A / 2n+1. 

Part (ii) Follows from (i) and from the fact that the equation hatI(t) = t^2/4 has a smallest positive root at hatt = A / 2n+1. 

Part (iii): Concavity of the three functions is verified by simple algebra. It remains to observe that all three functions hatI, tildeI and I* have a negative slope at t=0 and are decreasing at the points hatt, tildet and t*, respectively. 

• Caves, R. E., (1971), ``International Corporations: The Industrial Economics of Foreign Investment'', Economica, 38, 1-27. 
• Caves, R. E., (1982), Multinational Enterprise and Economic Analysis.. Cambridge University Press, Cambridge. 
• Cushman, D., (1985), "Real Exchange Rate Risk, Expectations and the Level of Direct Investment", Review of Economics and Statistics, 67, 297-308. 
• Cushman, D., (1988), "Exchange Rate Uncertainty and Foreign Direct Investment in the United States", Weltwirtschaftliches Archiv, 124, 322-334. 
• Donnenfeld, S., and S. Weber, (1993), "Exports versus Foreign Direct Investment", Working Paper, York University. 
• Ethier, W. J., (1992), "Multinational Firms and the Theory of International Trade", working paper no. 30, International Economics Research Center, University of Pennsylvania. 
• Froot, K. A., and J. Stein, (1991), "Exchange Rate and Foreign Direct Investment: an Imperfect Capital Markets Approach", Quarterly Journal of Economics, 106, 191-217. 
• Graham, E. M. and P. Krugman, (1991), Foreign Direct Investment in the United States, Washington, D.C.: The Institute for International Economics. 
• Horst, T., (1972), "Firm and Industry Determinants to Invest Abroad: An Empirical Study", Review of Economics and Statistics, 54, 258-266. 
• Horseman, I. J. and J. R. Markusen, (1987), ``Strategic Investment and the Development of Multinationals'', International Economic Review, 28, 109-121. 
• Klein, M. W., and E. Rosengreen, (1992), "The Real Exchange Rate and FDI in the U.S.: Relative Wealth versus Relative Wage Effects", NBER wp no. 4192. 
• Levinsohn, J. A., (1989), ``Strategic Trade Policy when Firms can Invest Abroad: When are Tariffs and Quotas Equivalent?'', Journal of International Economics, 27, 129-146. 
• Lipsey, R. E., (1993), "Foreign Direct Investment in the United States: Changes over Three Decades", ed. K. A. Froot, Foreign Direct Investment, The University of Chicago Press, Chicago. 
• Mann, C. L., (1993), "Determinants of Japanese Direct Investment in U.S. Manufacturing Industries", Journal of International Money and Finance, 12, 523-541. 
• Motta, M., (1992), ``Multinational Firms and the Tariff-Jumping Argument'', European Economic Review, 36, 1557-1571. 
• Ray, J. E., (1977), ``Foreign Direct Investment in Manufacturing'', Journal of Political Economy, 85, 283-297. 
• Ray, J. E., (1989), "The Determinants of Foreign Direct Investment in the United States: 1979-1987", in R. Feenstra, ed. Trade Policies for International Competitiveness, Chicago University Press, 53-77. 
• Salvatore, D., (1991), ``Trade Protection and Foreign Direct Investment in the U.S.'', The Annals of the American Academy, 516, 91-105. 
• Smith, A., (1987), ``Strategic Investment, Multinational Corporations and Trade Policy'', European Economic Review, 31, 89-96. 
• Trifler, D., (1993), "Trade Liberalization and the Theory of Endogenous Protection: An Econometric Study of U.S. Import Policy", Journal of Political Economy, 101, 138-160. 

    |EA and IB|EA and ID |EA and IA| IA   |EA and ID |EA and IB|ED & IB|EB & IB|
    |   (a)   |   (b)    |   (c)   | (d)  |   (f)    |   (g)   |   (h) |   (i) |
-------------------------------------------------------------------------------|
    |         |          |  n(A+t) |      |          |         |       |       |
    |         |          |  ------ |      |          |         |       |       |
    |  n(A+t) |X[0] = A -|   n+1   | An   |X[0] = A -| n(A+t)  |       |  An   |  
    |  ------ |t/2 - 2I/t|    or   | ---  |t/2 - 2I/t| ------  | A - t | ----- |
X[H]|   n+1   |          |   An    | n+1  |          |  n+1    |       |  n+1  |
    |         |          |  ------ |      |          |         |       |       |
    |         |          |   n+1   |      |          |         |       |       |
----|---------|----------|---------|------|----------|---------|-------|-------|
    |         |          |A-t(2n+1)|      |          |         |       |       |
    |         |          |---------|      |          |         |       |       |
    |A-t(2n+1)| A-t(2n+1)|  2(n+1) |  A   |A-t(2n+1) |A-t(2n+1)|       |       |
    |---------| ---------|   or    |------|--------- |---------|   0   |   0   |
x[f]| 2(n+1)  |  2(n+1)  |    A    |2(n+1)| 2(n+1)   | 2(n+1)  |       |       |
    |         |          | --------|      |          |         |       |       |
    |         |          |  2(n+1) |      |          |         |       |       |
--------------------------------------------------------------------------------

(a) t <= A / (2n+1) and I >= hatI(t)
(b) t <= A / (2n+1) and I >= I*(t)
(c) t <= A / (2n+1) and I < tildeI(t)
(d) t <= A / (2n+1) and I < I*(t)
(e) same as (c)
(f) same as (b)
(g) same as (a)
(h) A/(n+1) > t >= A / 2(n+1)
(i) t > A/(n+1)

Variable             |  Mean  | Standard  | Minimum | Maximum  | Obervations |
                     |        | Deviation |         |          |  >0  |  =0  |
----------------------------------------------------------------------|------|
FDI*                 |  65.35 |  290.46   |  0      |  8555    | 1255 |      |
---------------------|--------|-----------|---------|----------|------|------|
Imports*             | 258.65 | 1224.9    |   .997  | 26600    | 4001 |    0 |
---------------------|--------|-----------|---------|----------|------|------|
Tariff Rate (t)**    |   4.05 |    3.46   |  0.00   |    28.77 | 3769 |  232 |
---------------------|--------|-----------|---------|----------|------|------|
Non Tarrif           |  18.62 |   27.46   |  0      |   100    | 2699 | 1302 |
Barriers (NTB)**     |        |           |         |          |      |      |
---------------------|--------|-----------|---------|----------|------|------|
4-Firm Concentration |  38.78 |   19.57   |  2      |   100    |  --  |  --  |
Ration (C4)**        |        |           |         |          |      |      |
---------------------|--------|-----------|---------|----------|------|------|
Plant Size (I)*      |  54.28 |  101.36   |  2.9    |   830.2  |  --  |  --  |
---------------------|--------|-----------|---------|----------|------|------|
Exchange Rate        |  76.81 |   16.14   | 53.72   |   129.7  |  --  |  --  |
(EXRATE)             |        |           |         |          |      |      |
------------------------------------------------------------------------------

* In millions of U.S. dollars.
** In percentage terms.

Variable         | FDI    | Imports | Tarriff | Non Tariff | Exchange |
                 |        |         |  Rates  |  Barriers  |   Rate   |
----------------------------------------------------------------------|         
FDI              |  1.00  |         |         |            |          |
-----------------|--------|---------|---------|------------|----------|
Imports          |  0.132 |  1.00   |         |            |          |
-----------------|--------|---------|---------|------------|----------|
Tariffs Rate (t) | -0.021 | -0.072  |  1.00   |            |          |
-----------------|--------|---------|---------|------------|----------|
Non Tariff       |  0.042 |  0.089  |  0.371  |    1.00    |          |
Barriers (NTB)   |        |         |         |            |          |
-----------------|--------|---------|---------|------------|----------|
Exchange Rate    | -0.007 | -0.032  | -0.012  |    0.026   |   1.00   |
(EXRATE)         |        |         |         |            |          |
-----------------------------------------------------------------------

  
        |           FDI/IMP         |     (FDI/DMINV)/(IMP/SHIP)    |
        |   (1)   |   (2)  |   (3)  |    (4)   |    (5)   |    (6)  |  
------------------|--------|-------------------|----------|---------|
PSIZE   |  .007   |  .007  |  .006  |   .234   |   .234   |   .244  |
        |  .005   |  .005  |  .004  |   .166   |   .167   |   .160  |
--------|---------|--------|--------|----------|----------|---------|
EXRATE  | -.234*  | -.229* | -.229* | -7.971*  | -7.841*  | -7.834* |
        |  .046   |  .046  |  .045  |  1.650   |  1.644   |  1.644  |
--------|---------|--------|--------|----------|----------|---------|
NTB     |  .019   |  .187* |  .185* |   .697   |  6.697*  |  6.726* |
        |  .018   |  .054  |  .054  |   .651   |  1.961   |  1.957  |
--------|---------|--------|--------|----------|----------|---------|
NTB2    |         | -.002* | -.002* |          |  -.075*  |  -.075* |
        |         |  .001  |  .001  |          |   .023   |   .023  |
--------|---------|--------|--------|----------|----------|---------|
TAR     | -.252** |  .336  |  .336  | -4.072   | 20.038** | 18.112* |
        |  .144   |  .234  |  .234  |  5.171   | 12.270   |  8.415  |
--------|---------|--------|--------|----------|----------|---------|
CR      | -.053*  | -.025  |        | -1.578** |   .316   |         |
        |  .025   |  .041  |        |   .892   |  1.464   |         |
--------|---------|--------|--------|----------|----------|---------|
CR*TAR  |         | -.014* | -.014* |          |  -.572** |  -.519* |
        |         |  .005  |  .005  |          |   .310   |   .310  |
--------|---------|--------|--------|----------|----------|---------|
N. Obs. |  4002   |  4002  |  4002  |   3994   |   3994   |   3994  |
---------------------------------------------------------------------

Each regression also includes country and year dummies.
Standard errors reported under the coefficients.
*   Significant at the 5 percent level.
**  Significant at the 10 percent level.
*** See the appendix for a list of countries included in the analysis.

  
        |           FDI/IMP          |     (FDI/DMINV)/(IMP/SHIP)    |
        |   (1)   |   (2)   |   (3)  |    (4)   |    (5)   |    (6)  |  
------------------|---------|-------------------|----------|---------|
PSIZE   |  .011*  |  .011*  |  .010* |   .416*  |   .376*  |   .400* |
        |  .004   |  .004   |  .004  |   .137   |   .137   |   .130  |
--------|---------|---------|--------|----------|----------|---------|
EXRATE  | -.161*  | -.159*  | -.159* | -5.368*  | -5.299*  | -5.304* |
        |  .034   |  .034   |  .034  |  1.223   |  1.214   |  1.215  |
--------|---------|---------|--------|----------|----------|---------|
NTB     |  .007   |  .128*  |  .126* |   .107   |  4.280*  |  4.345* |
        |  .014   |  .041   |  .041  |   .485   |  1.477   |  1.474  |
--------|---------|---------|--------|----------|----------|---------|
NTB2    |         | -.002** | -.001* |          |  -.052*  |  -.053* |
        |         |  .0005  |  .0005 |          |   .017   |   .017  |
--------|---------|---------|--------|----------|----------|---------|
TAR     |  .041   |  .537*  |  .633* |  8.139*  | 35.620** | 31.866* |
        |  .111   |  .264   |  .181  |  3.940   |  9.345   |  6.400  |
--------|---------|---------|--------|----------|----------|---------|
CR      | -.055*  |  .012** |        | -1.851*  |   .604   |         |
        |  .019   |  .006   |        |   .675   |  1.095   |         |
--------|---------|---------|--------|----------|----------|---------|
CR*TAR  |         | -.015   | -.014* |          |  -.681** |  -.581* |
        |         |  .031   |  .004  |          |   .230   |   .141  |
--------|---------|---------|--------|----------|----------|---------|
N. Obs. |  2606   |  2606   |  2606  |   2603   |   2603   |   2603  |
---------------------------------------------------------------------

Each regression also includes country and year dummies.
Standard errors reported under the coefficients.
*   Significant at the 5 percent level.
**  Significant at the 10 percent level.
*** Includes:  Canada, Germany, Japan and United Kingdom.

  
        |           FDI/IMP          |     (FDI/DMINV)/(IMP/SHIP)     |
        |   (1)   |   (2)   |   (3)  |    (4)   |     (5)   |    (6)  |  
------------------|---------|-------------------|-----------|---------|
PSIZE   |  .021*  |  .012** |  .022* |   .905*  |    .444   |   .960* |
        |  .006   |  .006   |  .006  |   .310   |    .306   |   .282  |
--------|---------|---------|--------|----------|-----------|---------|
EXRATE  | -.008*  | -.006*  | -.004  |  -.338*  |   -.215   |  -.143* |
        |  .044   |  .043   |  .043  |  2.112   |   2.045   |  2.079  |
--------|---------|---------|--------|----------|-----------|---------|
NTB     |  .005   |  .184*  |  .186* |   .386   |   9.795*  |  9.916* |
        |  .018   |  .055   |  .056  |   .890   |   2.672   |  2.713  |
--------|---------|---------|--------|----------|-----------|---------|
NTB2    |         | -.002*  | -.002* |          |   -.124*  |  -.120* |
        |         |  .001   |  .001  |          |    .031   |   .032  |
--------|---------|---------|--------|----------|-----------|---------|
TAR     |  .727*  | 3.016*  | 1.784* | 37.924*  | 151.61*   | 88.767* |
        |  .166   |  .422   |  .260  |  8.014   |   20.310  | 12.560  |
--------|---------|---------|--------|----------|-----------|---------|
CR      | -.053   |  .185*  |        | -2.346** |   9.450*  |         |
        |  .027   |  .050   |        |  1.313   |   2.400   |         |
--------|---------|---------|--------|----------|-----------|---------|
CR*TAR  |         | -.060*  | -.026* |          |  -2.975*  | -1.246* |
        |         |  .011   |  .006  |          |    .522   |   .279  |
--------|---------|---------|--------|----------|-----------|---------|
N. Obs. |  1000   |  1000   |  1000  |    998   |    998    |   998   |
-----------------------------------------------------------------------

Each regression also includes country and year dummies.
Standard errors reported under the coefficients.
*   Significant at the 5 percent level.
**  Significant at the 10 percent level.

Footnotes

1. U.S. Department of Commerce(1993), Foreign Direct Investment in the U.S.: An Update.

 

2. See Salvatore (1991) for additionaldiscussion on trade barriers and FDI in the U.S. for the periodcovering 1970's and 1980's. See also Graham and Krugman (1991)for a comprehensive survey of foreign direct investment in theU.S.

 

3. For an early discussion on theeffects of tariffs on the decision to export or to get behind thetariff wall and produce abroad in the context of imperfectlycompetitive markets, see Horst (1971) and Caves(1982).

 

4. See also Horstman and Markusen(1987), Smith (1987) and Motta (1992). For a more recent andclosely relatedexamination of alternative modes of supply of goods to aforeign market, i.e., via exporting or setting upproduction facilities abroad, see Donnenfeld and Weber (1993). These papers focus on strategic considerations in the context offoreign direct investment. In all these papers save the last onean established MCis facing potential competition from a single domestic entrant.Donnenfeld and Weber (1993) considerthe case where the established firms are domestic whereasthe potential entrant is a foreign firm. 
5. The assumption that marginal cost are the same for all firms, domestic and foreign, is notnecessary, but it greatly simplifies computations. What is important is that the foreign firm'stotal marginal cost (including tariffs) when it is exporting is higher than in the case of FDI. Theconsequence of this assumption is that the foreign firm is at a competitive disadvantage relativeto the domestic firms, when it exports to the home country's market rather than engaging inFDI by setting up a plant in the home country. 
6. We assume that in thecase where the foreign firm is indifferent between the exportsand FDI, it chooses exports. 
7. Inthe appendix we present the proofs of all the propositions thatfollow. 
8. The Nash-Cournot levels of outputwill be chosen by the domesticfirms if total output is sufficiently large to inducetheforeign firm to stay out of the market. This happenswhen nA / n+1 >= A-t or t >=A / n+1. 
9. The properties of the functions hatI(t) are examined in the Appendix. 
10. The properties of the functions tildeI(t), and I* are examined in the Appendix. 
11. See Levinsohn (1989) who examined the implications of the possible occurrence of FDI for the equivalence between tariffs and quotas in imperfectly competitive markets. 
12. We presume that NTB is an exogenous variable here. Trifler (1993) shows that if one accounts for the endogeneity of NTBs its negative effect on imports is even stronger. 
13. Changes in the exchange rate affect the cost of exporting relative to the cost of servicing foreign markets via FDI. In the theoretical model the effects of the exchange rate were captured by the tariff, and for the sake of simplicity it was not explicitly modeled. In previous studies Ray (1989) and Mann (1993) investigated the effects of exchange rate changes on FDI. These studies, however, ignore the importance of strategic behaviour by domestic firms which is the focus of this study. 
14. The BEA reports information that is equivalent to the three-digit SIC. 
15. Most activity means that in these industries the reported number of transactions was largest. This criteria leads a very similar result to the alternative of using the value of FDI in the industry. A list of the SIC industries used appears in Appendix B. 
16. We are thankful to Hugh Arce of the U.S. International Trade Commission for providing the data. 
17. To accomplish the conversion we used the Bureau of the Census: U.S. Foreign Trade Statistics, Classifications and Cross Classifications tables, 1978. 
18. Thus if NTBs are so restrictive that prohibit imports, this will not be captured by the index that we use. Given, this shortcoming we decided not to weight industry NTBs by the observed level of imports in each industry. Consequently, we might underestimate the pervasiveness of NTBs in some U.S. industries. 
19. See ITC: The Economic Effects of Significant U.S. Import Restraints, 1993. 
20. The source is International Financial Statistics. We deflated the nominal exchange rates (series ae from the IFS) by the U.S. and each country's consumer price index (series 64 from the IFS) to obtain a real exchange rate. For Taiwan, the exchange rate data source is the Federal Reserve data base. Tariffs are defined as the ad-valorem tariff equivalent. 
21. These numbers refer to all exporting countries in an industry. The industries with zero tariffs are all in SIC 27, printing and publishing, except for SIC 3273, ready mixed concrete 
22. Four of these industries were in the textile sector (SIC 23) and the fifth was SIC 3523 (farm machinery and equipment) 
23. For the beginning and end periods in our sample, we also used a five year window. These windows encompass the closest available years and did not center around the year that the transaction took place. We also estimated the missing FDI values based on a three year window and this led to very similar results. 
24. For a similar conclusion regarding the effects of exchange rate on FDI in the U.S. see Froot and Stein (1991). Cushman (1985), (1988) examined the effects of the exchange rate on FDI in the U.S by a group of five major industrialized countries. 
25. The combined amount of FDI by Japan, E, and Canada in the U.S. totalled 9.5 billion dollars in 1983, 29.4 billion dollars in 1986, 43.1 billion dollars in 1989 and 10.5 billion dollars in 1991. The source is Foreign Direct Investment in the U.S.: An Update, 1993, U.S. Department of Commerce. 
26. During this period Japanese investment accounted for 25% of total FDI in the U.S.(Source U.S. FDI: an Update. U.S. Department of Commerce). 
27. Ray (1989) and Mann (1993) also find that the exchange rate had no effect on the Japanese FDI in the U.S. 

Re-engineering of Business Processes in Multinational Corporations

Katholische Universitat Eichstatt, Germany 

Working Paper No. 95-4 

The research for this paper was supported by a grant from the Carnegie Bosch Institute and was presented at the Institute's International Research Conference, November 2, 1994. 

Re-engineering of Business Processes in Multinational Corporations

Prof. Dr. Michael Kutschker

Abstract

Contents

• Introduction 
• International Business Processes 
• - Management of Ongoing Business Processes 
• - Improvement of Business Processes 
• - Re-engineering of Business Processes 
• Management of International Business Processes 
• - Coordination 
• - New Information Technologies 
• Re-engineering of International Business Processes 
• - Characteristics 
• - Information Technology -
•  Deeper Structure 
• - Internationality 
• - Complexity 
• Conclusion

Introduction

Both the characteristics of international management and the process orientation underlying the philosophy of BPR have been the reasons to present a paper at this conference. Fifty in- depth interviews in 30 large Multinational Corporations (MNCs) of Germany and Switzerland constitute the background for the authorÕs comments on Business Process Re-engineering in an international context. 

International Business Processes

Management of Ongoing Business Processes

Improvement of Business Processes

Re-engineering of Business Processes

However, in most cases the holistic approach ends at national borders. There are two possible reasons for ignoring the international dimension of business processes. Either the redesign of international processes is deemed to be unimportant or the international dimension is perceived not to change the character of BPR substantially. In view of the increasing internationalization of many industries, the first reason is rather academic. Therefore only the second reason may be accepted, raising the question: What is special about international BPR? The answer to this question is one of the objectives of this paper. 

Practitioners as well as academics and consultants have differing views about business processes. In the survey conducted for this research the interviews showed considerable variation in their understanding of business process issues. Some companies had improved or redesigned some isolated business processes, others had changed process systems, and only very few had introduced a comprehensive process re-organization, decomposing the ongoing activities of the company into a well defined set of business processes. Some authors stress organizational aspects of processes, others concentrate on aspects of improving processes or Business Process Re-engineering. Consultants often sell old concepts under the name of the new concept, BPR. Although perceptions and understanding of business processes are different, their common focus is to optimize the efficiency of an organization. Efficiency can be increased by a planned change of appropriate processes, thus shifting the attention of organization theory from structure to process. 

Management of International Business Processes

Coordination

Obviously, each subsidiary has its distinctive set of context factors, implying that a firm has to control different headquarter-subsidiary relationships in different ways. Empirical results confirm that firms are successful when they adapt the coordination instruments to the subsidiaries' specific situation. Moreover, it may be assumed that the coordination pattern of the headquarter-subsidiary relationship also varies with the character of strategic business units and of functional departments . 
A process orientation assumes that managers adapt the application of coordination instruments to the specifics of the business process. Therefore, the above mentioned four business processes were analyzed with respect to their political and structural character. The interviews with managers in German and Swiss MNCs clearly supported the view that managers do vary their use of coordination instruments according to the character of business processes. Personal instruments such as face-to-face meetings, personal instructions, and visits are more often used in political and ill-structured processes like strategic planning. In contrast, managers often apply technocratic instruments of coordination such as rules, programs and written documents in operational and well-structured decisions, such as physical distribution and warehousing. The processes of product launch and budgeting have a mix of personal and operational elements. In these processes managers vary the use of coordination instruments over the course of time, preferring face-to-face meetings in value-intensive soft phases and exchanging formalized documents in the operational phases of budgeting or product introduction. 

Managers construct the reality of their firms, subsidiaries, departments, and strategic business units. They design the structures of their units, develop organizational cultures, expand the internationality of their departments, and decide on "appropriate" coordination instruments in general. However, this self-created static frame of contingencies can only deliver a partial explanation of organizational behavior. Within the business processes, individuals constantly construct reality anew, accepting, ignoring or inventing new contingencies on the base of their perceived and assessed reality. So, it is not so much the absolute character of a business process that determines the selection of coordination instruments, but rather the perception of process complexity that dictates the use of coordination instruments. The interviews strongly support the view of "bringing mind back in . 
New Information Technologies

New Information Technologies (IT) are said to have a major impact on the coordination of headquarter-subsidiary-relations. New IT such as electronic mail, corporate and public data bases, application systems, fax, video and computer-conferencing , are considered to be some of the driving forces of internationalisation. 

However, very little academic research has focused on the interface between International Management, Organization Theory and Information Systems. Only a few authors have considered the fit between global business strategy and global IT strategy. Even though the strategic importance of IT is asserted, few studies closely investigate the relationship between state of the art applications of IT in MNCs and their impact and importance for coordinating dispersed activities and business processes. Research linking IT and coordination focuses either on domestic inter-unit coordination , or lacks empirical content. Some studies have the right research focus, but they are outdated because of the rapid change of IT. 

Regarding the enabling role of IT for the internationalization process, the interviews conducted produced mixed results. On the one hand more companies than expected have developed worldwide communication networks. These networks include E-mail as well as internal fax networks. On the other hand, the stereotype that IT pushes globalization was not supported. Firms change IT to facilitate the management of international business processes and renew the communications hardware when higher levels of internationalization ask for it. IT follows internationalization, but not vice-versa. 

New IT influences the interaction between headquarters and subsidiaries and may have an impact on the use of coordination instruments. In many interviews the influence of IT on coordination in MNCs was discussed. IT helps to solve problems which are intensified by the international scope of business processes: geographical distances that have to be overcome, scattered members in a decentralized organization who need to create and process information in many places, and different time zones between senders and recipients of information that pose additional problems. 

Our hypothesis that the impact of IT varies among the four business processes has been confirmed. Operational, well-structured processes like logistics tend to be more supported by IT than political, ill-structured processes like strategic planning. 

The perception of the surveyed respondents was that new IT does not lead to a substitution of coordination instruments. In those companies using global networks for the exchange of data and written information, the frequency and intensity of personal visits abroad have not decreased. The major reason for this lies in the fact that confidence and personal relationships can not at all be established by computer or video-conferencing. 

The four business processes do not only vary concerning the use of IT and coordination mechanisms, but also have different international orientations. An initial assumption was that internationalization will be realized by a well designed and orchestrated strategic planning process. Surprisingly, operational business processes are much more mutually dependent and linked than political processes. Operational processes appear to be better designed for the purpose of international coordination than strategic planning processes. Applying Perlmutter's framework , strategic planning is more ethnocentric with a strong orientation towards centralized decision-making, whereas R&D or logistics processes resemble more a geocentric network of mutually coordinating partners. However, the greater bulk of the sample is far away from having realized the vision of a transnational network organization - at least at the strategic level. 

The research was started with a traditional view on coordination instruments which can be considered as central subjects of the theory of international management. Throughout the interviews it was recognized that the question of international coordination is certainly interesting to managers. It was also evident that managers do not bother very much about traditional coordination instruments. Centralization, standardization, or formalization were found to be less important than questions of international team-building or the participation of foreign managers in strategic decisions. 

Decision arenas provide opportunities to exchange values, opinions, data, and test theories and thereby form organizational identity. Thus corporate culture controls and coordinates activities. The interviews show that traditional coordination mechanisms are supplemented or even replaced by such "new" forms of coordination. Norms and organizational culture play an increasingly important role in managing business processes. 

The internationalization of firms increases the dynamics and complexity of their relations with the environment. Rapid external change makes MNCs so vulnerable that they cannot fully rely on adaptive structural changes. They have to organize their business processes in ways that allow greater flexibility. Organizing for more flexibility means deliberately to design or re-design existing international business processes. 

Re-engineering International Business Processes

See Handelsblatt (1994), No. 117, June 21, 1994.
See for instance Davenport and Short (1990), Hammer (1990), Harrington (1991), Knorr (1991), Barton
(1993), Davenport (1993), Hammer and Champy (1993).

 These interviews have been carried out together with Professor Krcmar, Chair of Information Systems at the 
University of Stuttgart-Hohenheim,  and  two research assistants, Ulrich Hanfeld  and  Dr. Bettina Schwarzer, 
both being financed by the generosity of the Carnegie Bosch Institute.

  The terms business process re-engineering  and  business process redesign are used synonymously throughout 
the paper.

  See for instance Stopford and Wells (1972), Franko (1976), Doz (1980), Welge (1981). 
  The criteria for selection are taken from literature pertaining to individual  and  collective problem solving. See 
for instance March and Simon (1958). The selection was performed to obtain a large variance representing the 
spectrum of processes found in organizations.
  From an external point of view this situation leads to disjointed incrementalism; see Lindblom (1964).
  See on interface problems for instance Lim and Reid (1992) or Brockhoff and Hauschildt (1993).
  See Harrington (1991), Short and Venkatraman (1992), O`Sullivan and Geringer (1993).
  See Kane (1986), Belmonte and Murray (1993), Davenport (1993), Hammer and Champy (1993).
  The problem of coordination mechanisms originates in the bureaucratic model of Max Weber. Here we refer 
to Kenter's classification of coordination instruments. See Kenter (1985). 
  For exhaustive reviews of the literature see Khandwalla (1975), Welge (1981), Baliga and Jaeger (1984), 
Egelhoff (1988), Martinez and Jarillo (1989), Wolf (1994).
  See Prahalad and Doz (1987).
  See Brandt and Hulbert (1976), Prahalad and Doz (1981), Ghoshal and Nohria (1989), Roth and Nigh (1992).
  See Macharzina (1992), Sullivan (1992), Kutschker (1994).
  See Brandt and Hulbert (1977), Welge (1981), Kenter (1985).
  This was the title of an article stressing the role of the manager; see Pondy and Boje (1976).
  For a complete listing  and  interpretation see Conger (1988).
  See Manheim (1990), Jarvenpaa and Ives (1991).
  See for instance Jarvenpaa and Ives (1990), Jarvenpaa and Ives (1991), Ives and Jarvenpaa (1991).
  See for one of these studies Samiee (1984), Sethi and Olson (1991), Palvia and Saraswat (1992). 
  See Conger (1988), Moynihan (1985).
  See Malone (1988).
  See Mandell (1975), Brandt and Hulbert (1976), Mandell and Grub (1979).
  See Keen (1991), Manheim (1991).
  See Heenan and Perlmutter (1979).
  See Bartlett and Ghoshal (1989).
  See Etzioni (1968), Deal and Kennedy (1987), Bartlett and Ghoshal (1989), Nohria and Ghoshal (1994).
  See on flexibility Boudette (1989), Bleicher (1990), Maljers (1992), Amburgey, Kelly and Barnett (1993). 
  See Striening (1988), Hammer (1990), Gilbert and Siong (1993), Harrison and Pratz (1993), Girth (1994).
  See Gaitanides (1983), Peters (1988), Davenport and Short (1990), Holst (1991), Davenport (1993).
  See Kane (1986), Davenport (1993), Scheer (1993).
  See on process owners Striening (1988).
  See Pall (1987), Vantrappen (1992), Blattberg and Deighton (1993), Hammer and Champy (1993).
  See Handelsblatt-Karriere (1994), No. 29, July 29-30,1994, Fischer et. al. (1994).
  See Davenport (1993).
  See Gaitanides (1983), Bellmann (1991), Striening (1988).
  See Simon (1989), Fromm (1992).
  The sample consisted of an undifferentiated sub-sample of German MNCs  and  two Swiss  and  German sub-
samples in the electronics  and  the pharmaceutical industries.
  For instance Davenport, Ernst&Young or Hammer and CIC or Scott-Morgan and ADL.
  See for example Fischer et al. (1994) who give a disillusioning report.
  See Davenport (1993).
  See Kane (1986).
  See Brandes et al. (1989), Davenport and Short (1990), Hammer (1990), Venkatraman (1991), Scheer (1993), 
Benjamin and Levinson (1993), Hauser and Thurmann (1993).
  See for instance Schwarzer, Krcmar and Kutschker (1993), Schwarzer (1993), Schwarzer and Krcmar (1994).
  See Etzioni (1968). For the interplay of data, values,  and  theory in science see Galtung (1978), Kirsch 
(1991).
  The distinction between surface structures  and  deeper structures is based on ChomskyÕs concept on the 
syntax of a language; see Chomsky (1965), Kirsch (1991), Kirsch (1992).
  See for instance Kirsch, Esser and Gabele (1979), Kirsch (1991), Hedberg (1981), Meffert (1984), Stata 

(1989).

  See Handelsblatt Karriere (1994).
  See however Scott-Morgan (1994).
  See on co-orientation Kirsch (1992).
  See Etzioni (1968), Ouchi (1981), Nohria and  Ghoshal (1994).
  See Eisenhardt (1989), Yin (1990), Wollnik (1992), Schmid (1994).
  See for instance Adler (1991).
  See Porter (1985), Porter (1986), Bartlett and Ghoshal (1988), Bartlett and Ghoshal (1989).
  See Bellmann (1991), Wildemann (1993)