Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach

The conventional view, which the authors attribute to the “modern view of international capital markets” dating to Mundell (1968) and the industrial-organization theory of FDI (Hymer 1960, Kindleberger 1969, Vernon 1966), holds that with mobile capital, exchange-rate movements cannot impart any systematic cost-of-capital advantage to domestic or foreign investors – so the composition of a country’s capital-account surplus (bonds vs. direct investment, inflows vs. outflows) should be indeterminate with respect to the exchange rate (Sec. 1). Against this, the authors present a regression of detrended FDI inflows (as a share of U.S. GNP) against the real value of the dollar since 1973: a 10 percent dollar depreciation is associated with roughly $5 billion of additional FDI inflows (standard error under $2 billion), a relationship the authors note is not merely a 1980s phenomenon – excluding the 1980s from the sample actually raises the estimated coefficient.

The model’s core claim is that when there are informational asymmetries about an asset’s payoffs, external financing is more costly than internal financing, so changes in an entrepreneur’s own net wealth translate directly into changes in his demand for (willingness to bid for) information-intensive direct investment (Sec. 1). To the extent foreigners hold more of their wealth in non-dollar form, a dollar depreciation raises foreigners’ wealth relative to domestic entrepreneurs’, lowering foreigners’ relative cost of capital and allowing them to bid more aggressively for assets. The authors offer a stylized numerical illustration: a U.S. and a Japanese investor bidding for an office building worth $100 million in expected rents, each able to borrow up to 90 percent of the purchase price at 10 percent interest; at an exchange rate of 200 yen/dollar the U.S. investor’s $7 million cash lets him bid up to $70 million (beating the Japanese investor’s $50 million maximum bid from 5 million yen-equivalent dollars of wealth), but if the dollar depreciates to 100 yen, the Japanese investor’s dollar wealth doubles to $10 million and he now wins the bidding at up to $100 million.

The authors adopt Townsend’s (1979) costly-state-verification approach: a risky asset’s profit x is observed costlessly by the entrepreneur who manages it, but an external financier must pay a monitoring cost c to observe it, which makes the optimal financing contract a standard debt contract with repayment D (Sec. 2). With asset ability X (uniformly distributed profit on [0, X]), the lender’s expected return given contracted repayment D is R_D^L = D - D^2/(2X) - cD/X (eq. 2), implying entrepreneurs face credit rationing: the maximum obtainable loan is L_max = (X-c)^2 / (2X(1+r)), strictly below the full expected present value of the asset X/(2(1+r)) whenever c > 0. Proposition 1 characterizes the entrepreneur’s reservation (bid) price P(X,w,c) as a three-region function of wealth w: in Region 1 (low w) the entrepreneur is credit-rationed and his bid rises one-for-one with wealth; in Region 2 his bid still rises with wealth but at a decreasing rate as incremental wealth substitutes for costly external funds; in Region 3 (w (cid:18) X/(2(1+r))) the entrepreneur can self-finance the full value of the asset and further wealth has no effect on his bid. When c = 0 (no agency costs), the reservation price collapses to the frictionless present value X/(2(1+r)), independent of wealth entirely.

Proposition 2 shows that for c > 0, the probability phi_i(e,c) that the domestic entrepreneur wins a given auction is a strictly declining function of the exchange rate e (dollar depreciation), and that this effect is strictly stronger for more informationally intensive (higher-c) assets; when c = 0 the exchange rate has no effect on the auction outcome at all (Sec. 2.4). The underlying asymmetry is that a domestic entrepreneur’s dollar wealth is fixed in dollar terms and so does not change with e, while a foreign entrepreneur’s wealth, held in foreign currency, has a dollar value that rises one-for-one with depreciation – so foreigners’ reservation prices rise proportionately with e regardless of their own wealth level, while credit-constrained domestic entrepreneurs’ reservation prices rise less than proportionately (or not at all, if they are already in Region 3). The authors note that for assets without informational asymmetries (c = 0, e.g., passive stocks and bonds), all agents’ reservation prices rise proportionately with depreciation, so exchange-rate changes confer no bidding advantage on anyone – the prediction that FDI, specifically, should respond to the exchange rate while passive portfolio flows should not.

The authors show, via a worked numerical example (X^d = 10, X^f = 5.40, c = 3, w^d = 0.30, W^f = 3.00), that a depreciation-induced shift in asset ownership from domestic to foreign control can in principle lower both domestic and aggregate welfare, if it moves a credit-rationed domestic entrepreneur (Region 1) out of ownership of an asset he manages more productively than the foreign entrant, such that “an asset worth 5.80 in domestic hands is sold ’too cheap’” (Sec. 3). The authors stress this welfare-reducing possibility is special, arising only when the domestic entrepreneur is credit-rationed (Region 1): if his wealth instead places him in Region 2 or 3, he can always at least match the net value the asset would generate under his own management, which rules out welfare-reducing FDI by construction; the result is also weakened by introducing multiple competing foreign bidders. The authors further caution that the model is silent on two features with potentially offsetting welfare implications – it captures only foreign acquisitions of existing assets, not new capital formation by foreign-owned firms (which could have positive spillovers), and it abstracts from product-market/oligopoly effects of reduced foreign capital costs.

Section 4 considers five alternative hypotheses: (i) a tax-timing story (1981 depreciation-allowance changes favoring domestic buyers, reversed by the 1986 Tax Reform Act), which the authors do not dispute “may well be” partly true but note cannot explain why the exchange-rate/FDI relationship “was as strong before 1980 as after”; (ii) a portfolio-diversification story in which FDI is a fixed share of overall capital inflows that simply revalues with the exchange rate – which the model’s own prediction (passive portfolio flows should NOT respond to the exchange rate, since c = 0 for such assets) is explicitly designed to distinguish from, and which the authors test directly in Section 5; (iii) sticky asset prices creating temporary bargains – which the authors argue is both logically insufficient (sluggish prices would attract domestic as well as foreign buyers equally, absent the informational friction) and, properly understood, is actually implied by their own model (informational asymmetries make asset-price responses sluggish by construction); (iv) “tariff-jumping” in response to anticipated protectionism – which the authors argue cannot explain foreign acquisitions of existing assets specifically, since protection raises returns for American and foreign investors alike; and (v) a brand-loyalty/consumer-habit story (Americans “hooked” on Japanese goods during dollar strength, inducing later FDI as the dollar falls) – which the authors find economically “sound” and plausible but defer to the data to assess its importance.

Using quarterly 1973-1988 Balance-of-Payments data regressed on the real exchange rate, a relative-wealth (income) proxy, and a time trend, the authors report that direct investment is the only category of capital inflow statistically negatively correlated with the dollar’s value; foreign official assets also show a negative but statistically insignificant coefficient, while the point estimate for portfolio inflows into corporate stocks and bonds is positive and not significantly different from zero. In the annual-data regressions (Table 2), the coefficient on the exchange rate for Direct Investment is -0.8325 (robust standard error 0.2386, significant at the 1 percent level) in the simplest specification and -0.7424 (robust SE 0.2340) with a relative-income control added, while the coefficients for Foreign Official Assets, U.S. Treasury Securities, and Corporate Stocks and Bonds are statistically indistinguishable from zero. The authors note the results are robust to estimating on annual rather than quarterly data (point estimates somewhat larger, standard errors somewhat higher, as expected) and to including or excluding the relative-income control.

Disaggregating FDI inflows into 13 industries using Balance-of-Payments data, all 13 estimated exchange-rate coefficients are negative, five statistically significant, with the strongest effects concentrated in manufacturing, particularly chemicals – the authors note that since chemical products are not plausibly subject to the consumer “brand addiction” mechanism of alternative (v), this result suggests more is driving the pattern than that explanation alone. Using International Trade Administration (ITA) data that distinguishes acquisitions of existing assets from genuinely new capacity, the authors find all eight ITA-classified FDI transaction types show negative exchange-rate coefficients; the two largest categories by value – Mergers and Acquisitions (51 percent of 1987 FDI) and Equity Increases (20 percent) – are both statistically significant, while New Plant and Plant Expansion categories (together only about 5 percent of 1987 FDI) are also significantly negative but economically small. The authors conclude “it therefore appears that our overall results do not merely reflect foreigners’ participation in the rise and fall of exchange-rate sensitive sectors” – i.e., the dollar-FDI link reflects primarily the types of transactions (ownership transfers) that best fit the paper’s informational-asymmetry mechanism, not merely coincident capacity expansion in exchange-rate-sensitive industries.

Examining FDI inflows into the United States, United Kingdom, West Germany, Canada, and Japan using IMF Balance of Payments Yearbook data, most regression coefficients remain negative, but only the coefficient for Germany offers evidence the authors characterize as “strongly supportive” of the model. The authors are explicit that the aggregate Balance-of-Payments data available for other countries do not permit them to establish whether foreign inflows into those countries are driven by acquisition-type activity to the same extent as U.S. inflows, so they present this cross-country evidence as suggestive rather than as fully testing the model outside the U.S. context.

The Conclusion states plainly: “we have not attempted to provide a comprehensive theory of FDI. Although the exchange rate adds some explanatory power to the experience of the U.S. (and perhaps West Germany), there are obviously many other forces at work,” pointing specifically to the long-run upward trend in foreign ownership of U.S. assets (more than tripling over the decade preceding the paper), about which “our model sheds no light.” The paper frames its contribution narrowly: providing a model capable of explaining the correlation between exchange-rate movements and FDI documented in Figure 1 and confirmed in cross-sectional/time-series regressions, not a complete account of the secular rise in U.S. inbound FDI.