Journal Issue
Share
Article

The Role of Foreign Direct Investment in the External Adjustment Process

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1979
Share
  • ShareShare
Show Summary Details

This Paper is Concerned With the role of foreign direct investment (FDI) in the external adjustment process. In particular, it investigates the extent to which flows of FDI are affected by changes in countries’ relative competitiveness and real levels of aggregate demand. To this end, a simple model is developed to explain how a multinational firm locates its production facilities among different countries. This location model is then incorporated into a neoclassical investment function to explain the world-wide plant and equipment expenditures by multinational firms. The factors influencing the distribution of financing for these expenditures are then discussed, and a model of direct investment flows is derived. The results indicate that, in general, FDI flows are strongly influenced by changes in both real levels of demand and by countries’ relative competitiveness.

The plan of the paper is as follows: Section I begins with a brief review of the size of FDI flows and their relative importance for different countries. Section II discusses the possible effects of real exchange rate changes on the profitability of production in different countries and develops the model of plant and equipment expenditures by the multinational firm. Section III discusses the influences on the financing of these expenditures, and Section IV derives the final estimating equation for FDI and discusses its empirical application. Section V presents the results from fitting this equation for the FDI inflows and outflows of the United States, the United Kingdom, Japan, and the Federal Republic of Germany. Section VI contains a brief summary of the major results and conclusions.

I. The Relative Importance of Foreign Direct Investment

Over the last two decades, there have been a number of significant changes in the importance of FDI flows in the economies of the four countries considered in this paper. These trends are illustrated in Table 1, which gives net flows of FDI as a proportion of total exports of goods and services and as a proportion of gross domestic fixed capital formation. Japan and the Federal Republic of Germany have emerged as significant net exporters of FDI while there has been a decline in the relative importance of the United States, although net FDI outflows from the United States still remain greater than those from the other major industrial nations, both in absolute size and as a percentage of export earnings. The other major feature illustrated by the table is the continuing low level of FDI inflows into Japan, which is caused by Japan’s stringent controls on inward direct investment. (Japan’s licensing procedures for FDI are described in Bank of Japan (1971).) These controls make it impossible to explain FDI inflows to Japan using any conventional model, and consequently no attempt has been made to include these flows in the empirical analysis.

Table 1.Four Countries: Relative Importance of Foreign Direct Investment Flows During Selected Periods
Country and Period 1InflowOutflowNet Flow 2Net Flow as Per Cent

of Exports of Goods

And Services 2
Net Flow as Per Cent

of Gross Fixed Capital

Formation 2
millions of SDRsper cent
United States
1959–613662,685–2,319–8.4–2.7
1969–711,0377,120–6,083–9.5–4.6
1975–772,92910,740–7,811–5.7–3.2
United Kingdom
1959–61483624–141–0.9–1.2
1969–71522895–373–1.2–1.7
1975–771,2662,714–1,448–2.3–3.9
Germany, Fed. Rep.
1960–61220147+73+0.5+0.4
1969–71786909–123–0.3–0.3
1975–771,2012,048–847–0.9–1.0
Japan
1959–612874–46–1.0–1.3
1969–71125307–182–0.8–0.3
1975–771011,529–1,428–2.2–0.9
Sources: International Monetary Fund, Data Fund, for exports of goods and services. United Nations, Statistical Office, Yearbook of National Accounts Statistics (New York), various issues, for gross fixed capital formation. FDI flows are taken from the sources given in Appendix II.

The figures in each row are averages for the years shown.

A (+) denotes an inflow, and a (–) denotes an outflow.

Sources: International Monetary Fund, Data Fund, for exports of goods and services. United Nations, Statistical Office, Yearbook of National Accounts Statistics (New York), various issues, for gross fixed capital formation. FDI flows are taken from the sources given in Appendix II.

The figures in each row are averages for the years shown.

A (+) denotes an inflow, and a (–) denotes an outflow.

There are a number of interesting differences in the geographic and industrial distribution of the direct investment of the four major source countries. For instance, Japanese direct investments have tended to be more heavily concentrated in less developed countries (LDCs) than the investments of the other three. (About 60 per cent of the total stock of Japanese FDI was located in LDCs at the end of 1974, as compared with around 30 per cent for the other three countries. See Ishimine (1978).) Until recently, Japanese investments tended to concentrate in the primary and tertiary sectors, with most investment in the latter sector directly connected with facilitating exports to the host country (e.g., investment in the wholesale trade). However, in recent years, the appreciation of the yen and increased protectionist pressures on Japanese goods have led to an increased emphasis on direct investments in manufacturing. An example taken from the Japanese textile industry will be discussed further in the context of the location model to be developed later. In contrast, the Federal Republic of Germany’s direct investments are heavily concentrated in Western Europe (which accounts for around one half the total stock of German investments) and in manufacturing industry, with relatively little investment in primary, extractive industries. U. S. and U. K. investments tend to be more widespread geographically and industrially, although a relatively large share of the stock of U. K. direct investments is located in the developed primary producers (just under 30 per cent of the total stock by the end of 1974).

II. Development of a Model of Plant and Equipment Expenditures by a Multinational Firm

The aim here is to develop a simple theoretical model of plant and equipment expenditures that incorporates the effects of changes in countries’ relative competitiveness, including the impact of exchange rate changes. Most recent empirical models have tried to explain overseas plant and equipment expenditures by using some version of the same neoclassical investment function developed by Jorgenson (1967) to explain domestic investment expenditures. If product and factor markets, including the market for old capital goods, are perfectly competitive and capital is completely malleable, then determination of the firm’s optimal capital stock can be treated as a single-period, “myopic” decision problem, and the stock can be derived from the first-order condition that the marginal revenue product of capital must equal its rental cost. With a Cobb-Douglas production function Q = AKαLβ, the first-order condition becomes

where p denotes the price of output, ν denotes the rental cost of capital, and K* denotes the desired capital stock. Q and L denote the levels of output and labor input, respectively.

There are a number of problems common to all applications of this approach, most of which stem from the attempt to treat capital investment decisions as a single-period decision problem. In practice, typical capital investment goods cannot easily be resold or converted to other uses once installed, so expectations of future demand and factor prices have an important influence on present investment decisions. In the empirical application of these models, this objection is not overwhelming, since expectations formation can be built into the model as part of the lagged adjustment process of actual toward desired capital stocks. Nevertheless, the fact that most investments realize their returns over a number of years has important implications for the effects of exchange rate changes and countries’ relative rates of inflation on a multinational firm’s location decisions, as will be discussed later on.

For instance, consider how the level of output, Q, in the present model is determined. Most previous empirical applications have simply used a country’s gross domestic product (GDP) or its manufacturing output as a proxy for the aggregated output of all foreign-owned subsidiaries in that country. However, this ignores one of the crucial decisions facing a multinational firm: namely, why it should locate production in that particular country rather than elsewhere. To take account of this, a simple model of a firm’s location decision will be developed. First, consider the effects of inflation and exchange rate changes on the profitability of a single subsidiary of a firm, say a subsidiary located in the United Kingdom. Assume that the firm is truly multinational and aims to maximize profits measured in some basket of currencies. For convenience, assume this basket is the special drawing right (SDR). Let

  • Π = the subsidiary’s profits, expressed in SDRs

  • x = the exchange rate (pounds per SDR, so that an increase in x represents a devaluation of the pound)

  • Q = the subsidiary’s output

  • c(Q,x) = the subsidiary’s cost function, expressed in SDRs

  • p(Q,x) = the price of output, expressed in SDRs

All values are expressed in SDRs, since profits in SDRs are what interests the multinational firm. If the product is a traded good and the United Kingdom forms only a small part of the total world market for that good, then its SDR price will be determined by world market forces and will not be affected by U. K. inflation or changes in the pound’s exchange rate (at least in this partial equilibrium framework). If the product is a non-traded good, then U. K. inflation or changes in that country’s exchange rate will be completely reflected in the SDR price of the good (i.e., a 10 per cent U. K. inflation will raise the SDR price by 10 per cent provided there is no change in the exchange rate; and a 10 per cent devaluation will lower the SDR price by 10 per cent). In practice, most goods fall between the two extremes, being neither perfectly traded nor completely insulated from world market forces. The same is true for the prices of the firm’s inputs. So, in general, it is not possible to say whether an exchange rate change will have a larger impact on the SDR-denominated price of the product or on its SDR-denominated costs of production. It depends upon whether the firm’s output is more, or less, traded than the inputs it uses to produce that output.

The firm’s profits are

so the first-order condition for profit maximization is

Differentiating totally gives

so that

The second-order condition for profit maximization requires that the expression in the denominator be negative, so dQdx has the same sign as

(where MR and MC are marginal revenue and marginal cost in SDR terms, respectively, both decreased or left unchanged by a devaluation). As can be seen from the previous discussion of traded and nontraded goods and inputs, this expression is of ambiguous sign, so the effect of a devaluation of the foreign currency on the foreign subsidiary’s output is, strictly speaking, ambiguous. The foreign output will be increased or decreased by a devaluation, depending whether the SDR-denominated marginal revenue curve is shifted inward by less or by more than the SDR-denominated marginal cost curve. However, in most cases, it can be argued that the subsidiary’s final output is more likely to be a traded good susceptible to international market pressures than are its inputs, particularly its labor inputs. This would especially be so for those goods that are supplied partly from the production of foreign subsidiaries and partly from the firm’s home country. In such cases, a devaluation of the foreign currency would have more effect on the SDR-denominated costs of production than on the SDR-denominated prices, and so would tend to increase the output of the foreign subsidiary. (However, there are some cases where this would not be true, such as that of foreign-owned utility companies located in developing countries that sell a nontraded product and import a high proportion of their capital equipment and perhaps even their skilled labor.)

A similar analysis to the above can be done for the effects of foreign-country inflation on the output of the subsidiary. If the firm’s output is more of a “traded” good than the inputs the firm uses to produce that good, then inflation will shift the subsidiary’s SDR-denominated marginal cost curve outward by more than its marginal revenue curve, and foreign production will decline.

To derive a more specific expression for the output of the foreign subsidiary, a particular case of a firm producing “traded” output will be considered. Suppose the firm aims to minimize the SDR cost of supplying three markets (countries 1, 2, and 3) with quantities Q1, Q2, and Q3, respectively, and can meet this supply from production facilities in the three countries that produce outputs X1, X2, and X3, respectively. Suppose the subsidiaries in countries 1 and 2 produce enough to satisfy their domestic markets and are also actual or potential exporters to country 3, and suppose further that their exports face tariffs of t1 and t2, respectively. (Note here that t1 represents country 3’s tariffs on country l’s exports.) If c1 (X1), c2 (X2), and c3 (X3) are the SDR-denominated cost functions, then the firm aims to minimize its total SDR costs

subject to supplying the market

The first-order conditions for this are

where λ is the Lagrange multiplier.

For particular forms of the cost functions, it would be possible to solve this system of linear equations to derive expressions for each subsidiary’s output as a function of the costs of production of each subsidiary, the level of demand in each country, and the tariff levels. To indicate the direction of these effects, we can differentiate totally the first-order conditions and use Cramer’s rule to solve the resulting system of equations to obtain expressions for the change in each subsidiary’s level of output

The second-order conditions require that the denominator be positive so that, provided there are not increasing returns to scale, a subsidiary’s production is a positive function of demand in each market and of import tariffs in its host country and a negative function of import tariffs in other countries to which it exports. (Recall that t1 and t2 represent country 3’s tariffs on imports from subsidiaries in countries 1 and 2, respectively.) An increase in the costs of one plant’s production relative to the production costs of other plants will have a negative effect on its level of production, so a subsidiary’s production can be written as a function of

where CAV is the cost of production in each subsidiary relative to some weighted average of other subsidiaries’ costs of production and Q is a measure of the aggregate demand for the firm’s products (subsidiary production plus imports from the other plants).

Incorporating the model for the subsidiary’s output into the expression for the desired stock of capital (and ignoring changes in tariffs for the present) gives (assuming a linear relationship)

where CAV is the exchange-adjusted relative cost of production for the subsidiary and Q is a measure of aggregate demand for the firm’s products. In the empirical application, this will be proxied by the domestic demand for manufactured goods in the country concerned.

Actual capital stock will adjust to desired capital stock with a lag, so plant and equipment expenditures will be some lagged function of past desired and actual capital stocks. The lag functions will be discussed further in Section IV, which deals with the empirical application of the model.

As discussed earlier, the firm’s investment decisions are, in practice, based on a much longer time horizon than the one period assumed in the neoclassical model because of the absence of proper markets for secondhand capital goods. Therefore, it could be argued that firms will ignore diverging rates of inflation that alter the relative costs of production in different countries, since they would assume that these divergences would ultimately be reflected in exchange rate changes that would eliminate any such temporary advantages. However, divergent trends in costs of production can persist for a number of years—long enough to influence investment decisions—as Chart 1 shows. For instance, there has been a downward shift in relative unit labor costs in the United States versus those in Japan and the Federal Republic of Germany largely because of exchange rate movements in the early 1970s, and this trend has continued, so that relative labor costs of the United States are distinctly lower in the 1970s than they were in the 1960s.

Chart 1.Four Countries: Relative Unit Labor Costs in Manufacturing, Adjusted for Exchange Rate Changes, 1961–78

(1968=100)

Source: International Monetary Fund, International Financial Statistics, various issues.

An interesting specific example of the effects of changing relative costs of production on the international choice of location and direct investment expenditures is the textile industry in Japan and Southeast Asia. The appreciation of the yen in 1971, combined with rapid wage inflation in the Japanese textile industry during the early 1970s, led to a major loss in competitiveness for Japan-based production. This is illustrated in Chart 2, which gives the exchange-adjusted changes in labor costs per unit of output in the cotton spinning industries of Japan and Korea and indicates a major deterioration in Japan’s position from 1971 onward. This was soon reflected in the export performances of the two countries, as seen in Table 2.

Chart 2.Japan and Korea: Comparison of Labor Costs Per Bale of Yarn in the Cotton Spinning Industry, 1965–76

(In U. S. dollars)

Source: Woo (1978).

Table 2.Korea and Japan: Exports of Cotton Yarn, 1970 and 1975(In tons)
JapanKorea
1970197519701975
Exports to Asian countries4,0213,2693,85525,604
Exports to other countries3,0946,6839992,383
World total7,1159,9524,85427,987
Source: Woo (1978).
Source: Woo (1978).

A similar deterioration took place in the relative position of other sections of the Japanese textile industry (see, for instance, “International Competitiveness of Japan’s Textile Industry” (1976)), and this led to a rapid increase in overseas direct investments by the industry, most of which was concentrated in Southeast Asia, as is shown in Table 3. In the six years from the end of 1969 to the end of 1975, the stock of Japanese direct investments in Southeast Asian textile industries increased almost tenfold.

Table 3.Japan: Stock of Direct Investment in Overseas Textile Industries, 1969 and 1975(In millions of U. S. dollars)
End of 1969End of 1975
Southeast Asian countries70.6678.0
Other countries69.3338.6
World total139.91,016.6

III. Financing of Foreign Direct Investments

This section discusses the influences on the multinational firm’s financing of its international distribution of plant and equipment expenditures. One possible approach would be to treat the international allocation of both the firm’s assets and liabilities as a simple portfolio diversification problem, no different from that for other long-term capital flows. However, this would ignore the essential nature of direct investments, which are undertaken for industrial-organization type reasons (e.g., to reap greater economic rents on scarce industry-specific factors, such as technological information and management expertise). The approach used here will be to assume that the firm’s capital assets are first allocated according to its overall industrial strategy and that the firm then devises the optimal financing plan for these assets.

Suppose, in a simple two-country model, that the firm has amounts K1 and K2 of capital assets invested in countries 1 and 2, respectively, and wishes to choose the amounts D1 and D2 to be financed by borrowing in countries 1 and 2, respectively. For simplicity, assume the firm is interested in returns denominated in the currency of country 1. (The results would not be materially altered if the currency of country 2 or a basket of currencies were used.) Let K = K1 + K2 denote the total of assets to be financed; r1*,r2* denote the after-tax rates of return in each country, expressed in country l’s currency (both are random variables); i1*,i2* denote the interest rates on borrowing in each country, expressed in country l’s currency (i2* is a random variable); i1, i2 denote the interest rates on borrowing in each country, expressed in that country’s currency; and let e denote the expected rate of change of the exchange rate (i.e., the percentage rate of change in the price of country 2’s currency in terms of the currency of country 1; e is a random variable). So,

Using a mean-variance approach, the firm aims to choose D2 so as to minimize some function of the expected cost of borrowing

and of the variance of the overall portfolio

where D2 is treated as a negative asset. Thus, the firm will seek to occupy some point on the “efficient frontier” of portfolio choices, where for any given expected cost of borrowing, the variance is at a minimum. (The approach here follows that of Feldstein (1968) and Hartman (1979).) Which point on the efficient frontier of portfolios is chosen will then depend on the firm’s (or, rather, its shareholders’) preferences between risk and return. Each of the efficient portfolios is characterized by the minimization of variance V subject to some constraint on the cost of borrowing (i.e., D2 is chosen to minimize V subject to i = i0). Solving the Lagrangian for this gives the following expression for borrowing in country 2 (see Appendix I for details):

where var denotes variance and cov denotes covariance.

The coefficients of the second and third terms in this expression are those that would result if the rates of return denominated in country l’s currency, r1* and r2*, were each in turn regressed on changes in the exchange rate e. They reflect the extent to which returns in each country (still denominated in the currency of country 1) are sensitive to exchange rate changes. Therefore, these coefficients depend upon the relative sensitivity of the price of output and the costs of production to exchange rate changes, which has already been discussed in Section II. For any pair of countries, these two coefficients will probably not change very rapidly over time, since they depend on the structural attributes of the two economies.

The Lagrangian multiplier λ represents the firm’s marginal rate of substitution between the cost of borrowing and the variance of the overall portfolio. If the firm’s (really, the shareholders’) indifference curves are relatively flat within the range of actual fluctuations in interest differentials, then λ can be treated as a constant, so that the equation for desired borrowing in country 2’s currency becomes

where α=λ2

and where the coefficient γ will be close to zero if country 2 is a small part of the world market, so that changes in the exchange rate between it and country 1 do not significantly affect returns in country 1. The first term is the interest rate differential between the two countries including the expected change in the exchange rate, weighted by the variance of the probability distribution on exchange rate changes (in practice, this variance is unlikely to change rapidly over time, except when there is a switch from fixed to floating rates).

In any case, there are good reasons for believing that the whole first term may well tend toward zero. In a world of perfect capital mobility, capital flows (both portfolio and direct) are sufficiently large and respond rapidly enough to eliminate any potential covered interest differential, and a country’s domestic interest rate is determined by the world interest rate plus expectations regarding the country’s exchange rate. Arbitrage between the capital markets of the industrial countries is sufficiently rapid for observed interest differentials on liabilities of the same risk and maturity to be insignificant. There are two ways in which this arbitrage could occur:

(1) The arbitrage could be done solely by the movement of portfolio capital, which responds more rapidly to any potential interest differential than FDI flows do. In this case, any ex ante differential has no effect on FDI flows, since it is eliminated before they can adjust, and there is no need to include an interest-differential term in the estimated model.

(2) The arbitrage could be accomplished by adjustments of both portfolio capital and the financing component of FDI flows. In this case, the observed ex post interest differential will still be zero, but now any ex ante differential will have had an effect on FDI flows, although it will not be observed empirically because of an extreme simultaneous-equation bias.

An attempt was made to test whether the financing component of FDI flows responds rapidly enough to interest differentials to play a role in the arbitrage process. This involved using the excess flow demand for money in an economy as a proxy for the ex ante interest differential and including it in the direct investment equations. This is an adaptation of the Kouri and Porter (1974) model of capital flows in a small, open economy under fixed parities in which net capital inflows are determined by the excess flow demand for money. (For instance, if a country has a large positive excess demand for money, there will be a tendency for interest rates to rise at home, causing a positive ex ante covered interest differential and thereby inducing a net capital inflow that eliminates the differential. If part of this net capital inflow consists of changes in net FDI, then the excess demand for money will show up in the estimating equations for FDI flows.) The components of the excess flow demand for money (see Kouri and Porter (1974)) were included in the FDI inflow and outflow equations for the United Kingdom, the Federal Republic of Germany, and Japan. (The model is not really applicable to the United States, which is clearly not a “small” country.) However, the terms were never significant and frequently had the wrong sign, even when estimations were limited to the period of fixed exchange rates. This indicates that arbitrage of interest differentials takes place through portfolio capital flows and that FDI flows do not respond rapidly enough to temporary interest differentials to be significantly affected by them.

Therefore, the desired stock of foreign borrowing can be treated as a fairly stable proportion of the firm’s plant and equipment expenditures in that country, that is,

where β is a constant.

IV. An Empirical Application of the Model of Foreign Direct Investment Flows

This section discusses an empirical test of the simple model for the effect of changes in relative competitiveness on direct investment flows that uses data on the inflow and outflow of FDI for the United States, the United Kingdom, the Federal Republic of Germany, and Japan. Recall that the flow of direct investment funds that affects the balance of payments is the part of plant and equipment expenditures that is financed outside the host country. (In practice, part of the plant and equipment expenditures financed by local borrowing may still enter the FDI statistics if the borrowing is done by the multinational firm’s head office and the funds are then on-lent to its local subsidiary.) 1 If the firm’s stock of physical capital adjusts to its desired level with a lag, then a change in the desired stock of capital may generate investment expenditures over a number of periods, reflecting the various adjustments lags—that is,

(See Coen (1971) for a more detailed discussion of this specification.) If the local financing component is a constant proportion β of the desired capital stock, then the flow of FDI is

where It denotes the flow of actual physical investments in the firm’s plant and Bt denotes the flow of new local borrowing by the subsidiary to partially finance that investment.

Determinants of the Desired Capital Stock, K*

From the previous discussion,

where r denotes the rate of interest, δ the rate of depreciation, and pq the price of the firm’s output relative to the price of capital goods. In practice, it is difficult to get reliable data on shifts in this last ratio, especially in a multinational context, and it is here assumed to be constant.2 The aggregate demand Q for the multinationals’ output is represented by the index of final demand for manufactures in the host country. The interest rate used in the cost of capital is the long-term government bond rate in the source country.3 The rate of depreciation used in the cost of capital was taken to be a uniform 10 per cent. Experimentation with different rates did not greatly alter any of the regression results.

The host country for the capital outflow equations and the source country for the inflow equations are weighted composites of the major industrial countries4 with the weights assigned to each country determined by its share of the relevant stock of direct capital assets. These composite countries account for at least two thirds of the total stock of assets in each case, with the exception of the host countries for U. K. and Japanese FDI outflows—which accounted for only 40 per cent and 30 per cent, respectively, of total foreign direct assets—since a larger proportion of the total FDI stock of these two countries is located outside the industrial economies. For this reason, the estimated equations for these two FDI flows will be somewhat less reliable than for the others.

The Measure of Relative Competitiveness, CAV

The measure used here is the index of unit labor costs in each country, adjusted for exchange rate changes. Obviously, there may be a number of objections to this, as it is only a partial measure of trends in production costs. Prices of material inputs and other cooperating factors of production may move differently, and the measure may be distorted by differing trends in capital/labor ratios in the different countries. However, it is the only direct measure of changes in production costs that is available for a large number of countries, and the alternative, which would be to use an index of goods prices as a proxy for production costs, would raise even more problems. This is because the goods will, in general, be internationally traded, so that competition in world markets will tend to keep prices uniform regardless of relative shifts in the costs of production in different countries. In addition, there is usually much greater variation in the basket of goods produced in different countries than there is in the types of labor used, so that comparing goods prices between countries would lead to a much greater index number problem than comparing labor costs would. (See Artus (1978) for a more detailed discussion of the choice of price indices when comparing countries’ relative competitiveness.) A number of regressions were re-estimated using the GDP deflator for value added in manufacturing, adjusted for exchange rate changes, as an alternative measure, and none of the results were significantly altered.

The Lag Structure in the Adjustment of Capital Stocks

Recall that the general specification of the lag structure is that actual investment in any one period is the result of accumulated past decisions to alter the desired capital stock and that such decisions take time to implement. So

To estimate this empirically, it is necessary to use a more specific lag structure. One possibility is to specify that the lags decline at a constant geometric rate—that is,

So

Following the usual argument (see, for example, Coen (1971), p. 148), this can be written

which is the Koyck lag specification. (An alternative rationale for the Koyck specification would be that the firm aims in each period to adjust by a constant proportion, b, of the gap between the desired and actual capital stocks. The previous derivation is preferred here, since it serves better to show the connection between the Koyck lag and the Almon lag specifications.)

An alternative possibility is to specify that the lags Wi follow a quadratic distribution and to estimate them using the procedure developed by Almon (1965). In the actual estimation, the equations for capital flows to and from the United States and the Federal Republic of Germany give a better fit using the Almon lag, while the Koyck lag gives better results for the United Kingdom and Japan.

A further issue is whether the degree of capacity utilization in the host economy will have any effect on the adjustment process. In the simple neoclassical investment model, there can be no unutilized capacity because perfect markets exist for secondhand capital goods that allow the firm to sell any unwanted equipment and take decisions on a myopic, one-period basis. In practice, most capital equipment cannot be resold once it has been installed, so that the actual capital stock will adjust more slowly to its desired level after a sudden turndown in demand, since the only adjustments will be made as equipment becomes obsolete and is scrapped. Similarly, any subsequent increase in demand will result in less new investment than the simple model would predict, since unutilized capacity can be brought back into production before new investment becomes necessary.

A number of attempts were made to incorporate a measure of capacity utilization into the model, both as a separate variable and as part of the adjustment process of actual to desired capital stocks. However, none of these attempts fitted the data as well as the simpler model already developed.

Therefore, the two forms in which the model will be estimated are:

where Q and CAV denote, respectively, the final demand for manufactures and relative unit labor costs in the host country, and r denotes the long-term interest rate in the source country. A (–1) appearing after a variable indicates that it is lagged one period. (The host country for capital-outflow equations and the source country for capital-inflow equations are weighted averages of the major industrial countries with the weights assigned to each country equal to its share of direct investment in the country concerned. See Appendix II for details.) The rate of depreciation, δ, is set at 10 per cent, and FDK denotes the foreign direct capital stock in the host country.

Institutional Influences on Foreign Direct Investment Flows

Most countries have some type of controls over the movement of FDI. These, and a number of other institutional factors, may have to be taken account of in the empirical model:

(1) The U. S. Foreign Direct Investment Program (FDIP) was in effect, with varying degrees of stringency, from the beginning of 1965 to mid-1973. Its aim was to limit the strain on the U. S. balance of payments of FDI outflows, and it imposed quantitative controls on the amount of FDI which could be undertaken by a firm in any one year. The quotas took the form of a proportion of the firm’s FDI in a geographic area during a specified benchmark period. More liberal quotas were allowed for investments in developing countries. The Program is represented in the U. S. capital-outflow equation by a dummy variable for the period of its operations.

(2) Japanese controls over FDI inflows were so stringent during most of the period of estimation that the model developed above is not applicable.

(3) Most new outflows of direct investment capital from the United Kingdom (but not reinvested earnings) had to be financed through purchases of foreign exchange on the U. K. Investment Currency Market, which was a type of dual exchange market for foreign exchange resulting from the purchase or sale of overseas assets (both direct and portfolio) by U. K. residents. (See Woolley (1977) for a detailed description of the market.) During the balance of payments crises of 1965 and 1966, a number of measures were taken to reduce direct investment outflows by tightening the Investment Currency Market (e.g., reducing allocation of official foreign exchange to the Market, instituting compulsory surrender of a portion of the proceeds from the sale of overseas assets at the ordinary, rather than the “investment currency,” exchange rate). (See Boatwright and Renton (1975) for a more detailed discussion.) However, attempts to estimate these effects empirically, either via dummy variables or by including the “investment premium” over official exchange rates in the model, did not give significant results. This was probably because reinvested earnings were not included in the Investment Currency Market, and firms responded to the Market’s restrictions by increasing their financing out of these earnings. (Boatwright and Renton (1975) report that the share of unremitted profits of overseas subsidiaries in total U.K. direct investment rose from about one half in the period 1962–65 to around two thirds after 1966.)

(4) In addition to the longer-term influences on FDI flows, firms have also used short-term switches in interbranch indebtedness as a means of speculation on exchange rate changes. These effects have been picked up using a series of dummy variables for the periods of speculative activity. (See Appendix II for further details.)

V. Results

The estimated inflow and outflow equations for each of the four major countries (with the exception of Japan, for which no equation could be fitted for inflows) are given in Tables 4 and 5. A number of factors common to the equations should be noted:

(1) All equations are estimated for semiannual data.

Table 4.United States and Federal Republic of Germany: Foreign Direct Investment Flows Estimated with Quadratic Almon Lag Specifications 1
Dependent Variable

and Period of

Estimation
ConstantΣi=0nαi[Q¯(rδ)Q¯(1)(r(1)+δ)]Σi=0nβi[CAV(r+δ)CAV(1)(r(1)+δ)]FDK(–1)U.S.

FDIP
DummiesR2D-W

and

ρ
(sum αi)(sum βi)
U. S. FDI Inflow 19611–1977 II–619.0261.1 *

(2.20)
–674.9 *

(1.89)
0.084

(1.24)
348.0* DMY761

(3.66)
0.5712.10

0.55
U. S. FDI Outflow 19611–1977 II2,368.62,476.6*

(5.31)
–2,963.5*

(8.45)
–0.012

(1.59)
–330.7*

(2.64)
–2,270.0* DMY1AI

(9.57)
0.9591.85

–0.85
German FDI inflow 19611–1977 II–51.7893.6*

(2.85)
331.8

(1.43)
0.112*

(6.48)
425.7* DMY15U

(2.25)
0.8102.42
German FDI outflow 19611–1977 II–140.1522.8*

(2.22)
–434.6*

(2.90)
0.105*

(8.82)
0.8552.08

Figures in parentheses are t-statistics; * and * denote significance at 1 per cent and 5 per cent levels, respectively. D-W denotes the Durbin-Watson statistic. ρ denotes the estimated first-order autocorrelation coefficient in equations with the Cochrane-Orcutt transformation. U. S. FDIP is the dummy variable for the U. S. Foreign Direct Investment Program. A (–1) appearing after a variable indicates that the variable is lagged one period. The Roman numerals appearing after the dates denote half years.

Figures in parentheses are t-statistics; * and * denote significance at 1 per cent and 5 per cent levels, respectively. D-W denotes the Durbin-Watson statistic. ρ denotes the estimated first-order autocorrelation coefficient in equations with the Cochrane-Orcutt transformation. U. S. FDIP is the dummy variable for the U. S. Foreign Direct Investment Program. A (–1) appearing after a variable indicates that the variable is lagged one period. The Roman numerals appearing after the dates denote half years.

Table 5.United Kingdom and Japan: Foreign Direct Investment Flows Estimated with Geometric Koyck Lag Specifications 1
Dependent Variable and Period of EstimationConstantQ¯(r+δ)CAV(r+δ)FDK(–l)DummiesR2D-W and ρ
U. K. FDI inflow504.076.6*–80.6*0.074–260.6* DMY67110.6952.08
1964 I–1977 II(1.73)(3.31)(0.26)(2.71)
235.7* DMY701
(2.04)
U. K. FDI outflow3,579.7108.5–93.8*–0.122*511.4* DMY73U0.6712.07
1964 I–1977 II(1.65)(2.95)(2.04)(3.75)0.19
Japanese FDI outflow705.1150.4*–234.6*–0.081*–209.7* DMY74U0.5871.68
19641 I–1977 II(3.70)(4.50)(2.25)(2.97)0.49

Figures in parentheses are t-statistics; * and * denote significance at 1 per cent and 5 per cent levels, respectively. D-W denotes the Durbin-Watson statistic, ρ denotes the estimated first-order autocorrelation coefficient in equations with the Cochrane-Orcutt transformation. A (–1) appearing after a variable indicates that the variable is lagged one period. The Roman numerals appearing after the dates denote half years.

Figures in parentheses are t-statistics; * and * denote significance at 1 per cent and 5 per cent levels, respectively. D-W denotes the Durbin-Watson statistic, ρ denotes the estimated first-order autocorrelation coefficient in equations with the Cochrane-Orcutt transformation. A (–1) appearing after a variable indicates that the variable is lagged one period. The Roman numerals appearing after the dates denote half years.

(2) All values are measured in millions of SDRs.

(3) The investment flows and capital stocks are all deflated to constant-price terms, in each case using the index of capital goods prices in the host country, converted into SDRs. This was made necessary by the inclusion of the relative price (labor cost) terms in the equations (if bias caused by omitted variables was to be avoided).

(4) The relative unit labor cost variable CAV is so defined that the index of labor costs of the host country is always in the numerator, which implies that the expected coefficient of the CAV term is always negative (rising relative costs of production reduce direct capital inflows).

The exact definition and source of each variable used are given in Appendix II.

The simple model fits the data quite well for each of the four countries. A quadratic lag structure, estimated using the Almon procedure, fits best for the U. S. and German data, with investment flows typically depending on demand and relative cost changes over the previous six periods (three years). The lag for FDI inflows to the Federal Republic of Germany seems to be a little longer (around four years). The constant, geometrically-declining Koyck lag seems to fit best for U. K. and Japanese data. There was significant first-degree serial correlation in the initial regressions, except those for the Federal Republic of Germany, and this has been dealt with using the Cochrane-Orcutt transformation.

Table 6 gives estimates of the cumulative effects over the first four years of a 1 per cent increase in the real level of demand or in relative labor costs. The following general points can be made about the results:

Table 6.Four Countries: Cumulative Effects on Foreign Direct Investment (FDI) Flows of a 1 Per cent Increase in Host Country’s Real Level of Demand (Q) or in Relative Unit Labor Costs (Cav) 1(in per cent)
Country and FDI Flow1 Per Cent

Increase in

Real Demand
1 Per Cent

Increase in

Relative Labor Costs
United States
Inflow0.39–0.84
Outflow1.53–1.49
Federal Republic of Germany
Inflow0.760.29
Outflow0.59–0.45
United Kingdom
Inflow1.60–2.48
Outflow0.51–0.26
Japan
Outflow1.83–1.86

The figures in the table give the percentage changes in the total flow of FDI over a four-year period following a permanent 1 per cent increase in real demand or in relative labor costs (i.e., for the United Kingdom, a 1 per cent increase in demand causes total FDI inflows over the next four years to rise by 1.6 per cent).

The estimates are calculated on the assumption that the rate of interest, r, is a constant 10 per cent. A lower rate of interest would lead to larger elasticity estimates.

The figures in the table give the percentage changes in the total flow of FDI over a four-year period following a permanent 1 per cent increase in real demand or in relative labor costs (i.e., for the United Kingdom, a 1 per cent increase in demand causes total FDI inflows over the next four years to rise by 1.6 per cent).

The estimates are calculated on the assumption that the rate of interest, r, is a constant 10 per cent. A lower rate of interest would lead to larger elasticity estimates.

(1) The coefficients for the demand term were positive and significant at the 5 per cent level in all of the equations, which shows that increased demand in the host country leads to an increased FDI inflow.

(2) The coefficients for the relative unit labor cost term were negative and significant at the 5 per cent level for all equations except the one for FDI inflows to the Federal Republic of Germany. Thus, increased costs of production in the host country, relative to costs of production in the rest of the world, lead to a reduced FDI inflow.

(3) FDI outflows from the United States and Japan seem to be much more responsive to changes in relative labor costs or in the real level of demand overseas than are those of the United Kingdom and the Federal Republic of Germany.

(4) FDI inflows to the United Kingdom are more responsive to changes in the level of demand and relative competitiveness than are FDI inflows to the United States and the Federal Republic of Germany. (To a large extent, of course, the results for the U. S. FDI outflow and U. K. FDI inflow are linked, since U.S. firms account for over half of all foreign direct capital in the United Kingdom.)

(5) The U. S. Foreign Direct Investment Program had a significant effect in reducing FDI outflows from the United States. For the period of its operation, it reduced outflows by an estimated SDR 330 million per period, or by an estimated SDR 660 million per year (compared with actual outflows of around SDR 7 billion per year).

VI. Summary and Conclusion

This paper has been concerned with investigating whether foreign direct investment flows play any role in the external adjustment process. It has shown that, in principle, an increase in the real exchange rate can have either a positive or negative effect on the profitability of a multinational firm’s production in a particular country. However, if, in general, the firm’s output is more of a traded good than the inputs used to produce that good, there is a presumption that profitability and the level of production will increase as the real exchange rate decreases. Therefore, a simple model of the location decision of a multinational firm has been constructed and incorporated into a neoclassical investment function to obtain a model of FDI flows. This has been fitted to data on the FDI flows for four major industrial countries. The results indicate that these flows are responsive to changes in both real levels of demand and in relative competitiveness, although there are quite large variations, from country to country, in the degree of response.

APPENDICES

I. The Financing of Foreign Direct Investment in a World of Uncertain Exchange Rates

The financing problem for the firm is to choose D2 to minimize variance V, subject to i = i0. The Lagrangian for this is

So the first-order condition gives

But it*=i2+e, and only e is a random variable, so

Therefore, D2=λ[i2+e¯i1]2var(e)K1cov(r1*,e)var(e)K2cov(r2*,e)var(e)

This is based on Hartman (1979).

II. Sources and Definitions of Variables Used in Regressions

(All values have been converted to millions of SDRs.)

Host country: For outward flows of direct investment from each of the four source countries for which regressions are estimated, the host country is a weighted average of the major industrial countries,5 with weights equal to the relative shares of each country in the direct capital assets of the source country as of the end of 1970.

Source country: Similarly, for inflows of direct investment to each of the four countries, the source country is a weighted average of the industrial countries, with the weights in each case being equal to each country’s relative shares in the foreign-owned assets within the particular host country as of the end of 1970.

FDI: Inward and outward flows of direct investment for the United States, the Federal Republic of Germany, the United Kingdom, and Japan, respectively. In each case, flows are converted to constant 1970 prices by the SDR-denominated index of capital goods prices in the host country. All series include reinvested earnings. Sources are:

United States—Department of Commerce, Bureau of Economic Analysis, Survey of Current Business, various issues.

Federal Republic of Germany—Statistische Beihefte zu den Monatsberichten der Deutschen Bundesbank, Reihe 3: Zahlungsbilanzstatistik, various issues.

United Kingdom—Central Statistical Office, Economic Trends, various issues (excluding the transactions of oil and insurance companies).

Japan—Bank of Japan, Foreign Department, Balance of Payments Monthly, various issues.

FDK: The inward and outward stocks of direct investment for the United States, the Federal Republic of Germany, the United Kingdom, and Japan, respectively. These series were calculated using a simple perpetual inventory model, starting from benchmark stock figures derived from various national sources (e.g., Krägenau (1975); United Kingdom, Department of Industry (1977); and United States, Department of Commerce, Bureau of Economic Analysis, Survey of Current Business, various issues.

One problem is the choice of a suitable rate at which to depreciate the old capital stock. In fact, depreciation is already largely accounted for by the way in which the direct investment statistics are derived. They are calculated as the sum of new capital outflows and the home firm’s share of the reinvested profits of the foreign affiliate. But depreciation is deducted before the affiliate’s profits are calculated, so that the published direct investment figures do not measure gross investment overseas by the home firm. Therefore, a further depreciation element needs to be included in the calculation of the capital stock figures only to the extent that the affiliate’s depreciation provision does not reflect true depreciation. Since this is an unknown element, the stock figures in this study have been calculated on the assumption that depreciation has already been fully accounted for in the derivation of the investment statistics. Experimentation with different capital stock series, calculated using positive rates of depreciation in the perpetual inventory model, did not greatly alter any of the regression results.

Q: The index of final demand for manufactured goods in the host country (1970=1.0). Final demand is defined as output in manufacturing minus exports of manufactures plus that portion of manufactured imports estimated to belong to final demand. The indices are taken from the data base of the Fund’s World Trade Model.

r: The long-term government bond rate in the source country. This information is taken from International Monetary Fund, International Financial Statistics.

CAV: This is an index (1970=1.0) representing unit labor costs in each of the four major industrial countries relative to the unit labor costs in a composite country representing the world market. The composite country that has been chosen is a weighted average of the 12 major trading countries, with weights equal to the share of each country’s currency in the SDR basket. The individual country’s indices of unit labor costs are taken from the data base of the World Trade Model. The index is so defined that the unit labor cost of the host country is always in the numerator. Therefore, the expected sign of the relative competitiveness term in the regression is always negative.

FDIP: A dummy variable representing the U. S. Foreign Direct Investment Program. The variable takes a value of 1 from 19651 to 1973II and a value of 0 during other periods (Roman numerals denote half years).

DMY 76, DMY 74 I, etc.: Dummy variables representing periods of speculative activity. They take a value of 1 for the date specified, and a value of 0 for all other dates. The short-term shifts in interbranch indebtedness picked up by the dummy variables are similar to the “leads and lags” phenomenon in the payments for imports and exports, except that the transactions now take place between different branches of the same multinational company and so show up in the foreign direct investment statistics rather than showing up as shifts in the stock of trade credit or as other short-term capital movements. DMY 741 also picks up a fall in the U. S. FDI outflow caused by a Middle Eastern country’s purchase of stock in the local affiliate of a U. S. oil company (see United States, Department of Commerce, Bureau of Economic Analysis, Survey of Current Business, Vol. 57 (August 1977), p. 40).

BIBLIOGRAPHY

Mr. Goldsbrough, economist in the External Adjustment Division of the Research Department when this paper was prepared, is currently economist in the African Department. He is a graduate of Cambridge University and received his master’s degree and doctorate from Harvard University.

Note that the portion of the subsidiary’s reinvested earnings that belongs to the parent company is treated as financing from outside the host country and is included in the definition of FDI flow for each of the four countries considered in the empirical analysis.

An attempt was made to construct this variable using a weighted average of host country movements in the GDP output deflator relative to the capital goods deflator. (If the capital goods deflator was not available, some suitable wholesale price index for capital goods was used.) However, the simpler model treating this term as a constant gave a better fit, and the estimated parameters were not significantly altered. This has also been the experience with most applications of the neoclassical model to domestic investment expenditures.

If investment decisions were truly taken according to the “myopic” model, the firm would have no reason to borrow for more than one period at a time, and a short-term rate would be appropriate. However, since capital goods cannot easily be resold once they are installed, the myopic model is not strictly applicable, and most capital expenditures are, in fact, financed by long-term borrowing.

Austria, Belgium, Canada, France, the Federal Republic of Germany, Italy, Japan, the Netherlands, Sweden, Switzerland, the United Kingdom, and the United States.

The 12 countries used in the various weighting schemes are Austria, Belgium, Canada, France, the Federal Republic of Germany, Italy, Japan, the Netherlands, Sweden, Switzerland, the United Kingdom, and the United States.

Other Resources Citing This Publication