International Capital Mobility in the 1990s

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 472

June 1994

INTERNATIONAL CAPITAL MOBILITY IN THE 1990s

Maurice Obstfeld

NOTE: ‘nternational Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussicn Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors.

ABSTRACT

This paper surveys the performance of international capital markets and the literature on measuring international capital-mobility. Three main functions of a globally integrated and efficient world capital market provide focal points for the analysis. First, asset-price arbitrage ensures that people in different countries face identical prices for a given asset. Second, to the extent that the usual market failures allow, people in different countries can pool risks to their lifetime consumption profiles. Third, new saving, regardless of its country of origin, is allocated toward the world’s most productive investment opportunities. The paper evaluates the international capital market's performance of these roles by studying data on international interest-rate differences, international consumption correlations, international portfolio diversification, and the relation between national saving and investment rates. The conclusion is that while international capital mobility has increased markedly over the last two decades, international capital movements remain less free than intranational

movements, even among the industrial countries.

International Capital Mobility in the 1990s Maurice Obstfeld*

Over the past two decades, global trade in financial assets has been spurred by advances in communication and transaction technologies, by the creation of new financial products, and by a widespread trend toward deregulation of domestic and international capital-market activities. In almost all respects, the consequences of these developments remain controversial.?

In theory, the potential benefits of international capital mobility are clear. Individuals gain the opportunity to smooth consumption by borrowing or diversifying abroad, and world savings are directed toward the world’s most productive investment opportunities. The size of these gains, and the extent of their attainment in practice, remains uncertain and furnishes an active research area in which answers are urgently needed. High on the policy agenda in a number of countries is a choice between further integration into regional and world capital markets and the retention of traditional macroeconomic policy options.

This paper surveys the performance of international capital markets and the literature on measuring international capital mobility. Section 1 reviews the main functions and implications of capital mobility. Section 2 examines recent evidence on the world capital market's ability to arbitrage prices of similar assets. The market’s record in allowing countries to diversify risks is taken up in section 3. Section 4 focuses on interpreting divergences

between national saving and domestic investment rates. Section 5 concludes.

*The author is Professor of Economics at the University of California at Berkeley, and was a visiting scholar in the International Finance Division when this paper was revised. The paper will appear in Peter B. Kenen (ed.), Understanding Interdependence: The Macroeconomics of the Open Economy, Princetcn University Press, 1994. This paper represents the views of the author and should not be interpreted as reflecting those of the Board of Governors of the Federal Reserve System or other members of its staff.

*For an excellent overview of the expanding range of international financial markets, see Goldstein et al. (1993).

1 Free International Capital Mobility: Definition and Implications

Capital is freely mobile within a multi-country region when its residents face no official obstacles to the negotiation and execution of financial trades anywhere and with anyone within the region and face transaction costs tnrat are no greater for parties residing in different countries than for parties residing in the same country. The definition implies that national authorities do not interpose themselves between transaction partners from different countries, other than through the provision of a nationality-blind legal framework for contract enforcement.

Actual conditions may differ from this ideal of free international capital mobility. Governments can impose taxes on cross-border financial flows and payments, including certain types of reserve requirements, as well as quantitative limits and outright prohibitions. The mere possibility of such measures can discourage international capital movement, as can official "moral" suasion in which threats of formal regulation may be implicit. “he prospect of partial or full government expropriation of foreign-owned assets lowers the financial openness of some economies. Differences in language and business practice can raise the cost of an international financial deal relative to that of a similar bargain between residents of the same country.

In measuring the strength of such barriers to international capital movement, an essential comparative benchmark is the ideal case of perfect international capital mobility, in which capital is free to move internationally and transaction costs are literally zero. This section therefore reviews the main implications of perfect capital mobility, implications that will be compared with recent experience in the sections to

follow. A main theme of the discussion is that such comparisons are selclom

ct

straightforward: many commonly used barometers of capital mobility are based on strong, often questionable, auxiliary assumptions about the world.

The Law of One Price

Perhaps the most basic implication of perfect capital mobility is that an asset’s price must be the same wherever it is sold. With sufficiently detailed data, it would be possible to test this implication directly on a wide array of assets. In practice, however, most tests of the law of one price examine the prices in different localities of a narrow set of closely comparable assets, namely, claims on specified future currency payments.

The dollar price of $1 to be delivered in country A one period from today is 1/(1 + i$), where i4 is the one-period nominal dollar interest rate in country A. In country B on the same date, the nominal dollar interest rate is if. Under perfect capital mobility, the price of a future dollar is the same no matter where the claim to the dollar is located. Thus, the equality i3 = i? holds true (as does the corresponding equality for any other currency).

Empirical studies have pursued this implication of perfect capital mobility by comparing nominal currency interest rates in different financial centers, for example, the interest rates on large dollar certificates of deposit sold in New York and those on London Eurodollar deposits of the same maturity. Strictly speaking, such assets do not guarantee the same payment in all states of nature--for example, the unregulated offshore Eurodollar market may be more prone to a generalized financial crisis than is the onshore U.S. money market. Nonetheless, the relation between nominal interest rates on the same currency in different financial centers is probably the least ambiguous of the commonly used indicators of international capital mobility.

In contrast, little can be learned about international capital mobility

from cross-country comparisons of nominal or real uncovered returns on differei:t currencies. Such tests are uninformative about capital mobility because they necessarily appeal to auxiliary maintained assumptions that may or may not be valid independently of the degree to which capital is mobile.

To illustrate, let ifs be the one-period dollar interest rate in New York, if the corresponding rate in the London Eurodollar market, if, the nominal deutschmark interest rate in Frankfurt, if, the Eurodeutschmark interest rate, and x, the subsequent one-period percentage change in ‘she dollar price of deutschmarks.

Consider how information about capital mobility is embedded in the ex post difference in dollar returns between dollar deposits in New York and deutschmark deposits in Frankfurt, if§ - if, - xy. Let E(*) denote a conditional expectation. If one decomposes the preceding dollar return

differential into

USE E OE E g ~ 1g) + Ug - toy - EXg py $/om ~ *gvpm? * “py

T T T T

) + (Ex

(i ~

if) DM’

U.S. onshore- foreign-exchange expectation German offshoreoffshore dif- risk premium error onshore differential ferential

it becomes apparent that all direct information about international capital mobility is contained in the two onshore-offshore differences. Perfect capital mobility has the clear implication that both of the onshore-offshore interestrate differentials above must be zero; but the implications of perfect capital mobility for foreign-exchange risk premia and exchange-rate forecast errors are much less obvious.

The risk premium links expected returns on assets (such as different-

denomination Eurocurrency deposits) that are identical in location and in all

o* er respects except for currency of denomination. As stressed in my 1986 pa; zr, however, hypotheses about the relative returns on two London deposits can yield no direct information on capital mobility among financial centers.

It is similarly difficult to think of a Significant direct link between capital molsility and the exchange-rate forecast errors of market participants. Conceivably, the degree of capital mobility affects the information-revelation process in foreign-exchange markets, with some impact on the distribution of forecast errors, but no definite hypotheses concerning such effects have been advanced, let alone tested.

Thus, only with the aid of specific and probably irrelevant maintained hypotheses about the risk premium and expectations can one glean information about capital mobility from ex post uncovered return differentials such as is - itu - Xspy- Tests based on international differences in real interest rates-domestic nominal rates less expected domestic inflation--would require even more maintained auxiliary hypotheses than those based on uncovered nominal returns (see Obstfeld, 1986, for a detailed discussion). A more direct approach, yielding results vastly easier to interpret, is to analyze the one observable and relatively unambiguous indicator of capital mobility, the onshore-offishore interest differential.? Results based on this indicator are

reported ir. Section 2.

*Tests of covered interest parity between different countries, such as those reported by Giavazzi and Pagano (1985) and Frankel (1993), can be formulated so that they are equivalent to comparisons of onshore and offshore interest rates in the same currency. To return to the example, let fy», be the one-period forward premium for deutschmarks in terms of dollars quoted in the London market. Eurocurrency arbitrage ensures that ie = ify + Lypyr 80 the covered differential ipy + fypy - if between the Frankfurt deutschmark market and the Eurodollar market is identical to the onshore-offshore deutschmark differential if, - if,.

Consumption Insurance Capital mobility allows countries to trade differential consumption risks; the effect is to provide mutual insurance against purely idiosyncratic national consumption fluctuations. In practice, consumption insurance is provided by trade in a wide array of contingent and noncontingent securities: a crossborder exchange of common stock, for example, will alter the statistical distribution of both trading partners’ future consumptions. The insurance function of international capital markets is best illustrated, however, by assuming that countries can trade a set of Arrow-Debreu securities, one of which entitles its owner to a specified payment on a particular date if, and only if, a well-defined event, or "state of nature," occurs.

Figure 1 illustrates the effect of trade in such securities for a world in which there are two countries peopled by representative agents, A and B, two states of nature, 1 and 2, and in which consumption of a homogeneous nonproduced output is the only argument in utility functions. At the endowment point £, country A is relatively well-endowed with state 1 consumption and country B with state 2 consumption; that is, state 1 is relatively more favorable to the fortunes of country A; state 2, more favorable to those of country B. Otherwise, the two countries are, for simplicity, portrayed as being identical. If the free exchange of Arrow-Debreu securities is allowed, country A exports and country B imports securities that pay off in state 1; to balance this trade, country A imports and country B exports securities that pay off in state 2. At the resulting free-trade allocation, point F, both countries have raised their utilities by reducing the variability of consumption across states of nature.

Note that this outcome is predicted by the classical principle of

State 2 consumption

State 1 consumption

FIGURE 1: Trade across states of nature

6a

comparative advantage, whereby a country exports the good the domestic autarky price of which is relatively low.’ The relative price of the two available Arrow-Debreu securities can be identified with the price of state 1 consumption in terms of state 2 consumption. As usual, the free-trade price, shown as p in Figure 1, lies between the countries’ autarky prices; and, ina trading equilibrium, the countries have equated their marginal rates of substitution across states to p, and thus to each other.

The key statistical implication of an efficient allocation of consumption risks is that countries’ marginal utilities of consumption are proportional and, thus, perfectly correlated across states of nature. Notice that this proportionality holds true if and only if national marginal rates of substitution across any two states of nature coincide.‘

The preceding empirical prediction stems from two distinct assumptions: that there is free and costless international asset trade, and that the available set of securities available to trade is complete, so that all consumption risks are insurable. In theory, either of these two assumptions can fail independently of the other; in practice, it is clear that the existence of informational asymmetries and limits to enforcement restricts the extent to which individuals can contract to share risks. Even under perfect

capital mobility, there thus may be no close ex post association between

3Svensson (1988) places this result in a generalized setting.

‘To formalize the one-period example in Figure 1, let c4(s,) be the consumption of a representative individual from country A in state j (j = 1, 2,-.-, N) and let U*[c*(s,), c4(s,),---, C4(Sy)] be country A’s utility from its contingent consumption plans. Then, with similar notation for country B, marginal utilities are proportional if, for some constant A > 0 and for every state j, Uj = Au, where U; is a partial derivative with respect to state-j consumption. But this condition implies the international equalizat:ion of marginal rates of substitution between any states j and 1, that is, uA/uS = u?/U?. To show the converse, define A = U,/4/U,’.

national consumption levels. Other things being equal, however, increasing international capital mobility should entail an increasing tendency for positively correlated consumption comovements among countries. Evidence related to this prediction is discussed in Section 3.

The International Allocation of Investment

If the set of state-contingent assets people trade is sufficiently rich, perfect capital mobility leads to an efficient international allocation of investment: at the margin, a decision to invest a unit of output in country B rather than in country A should not affect the expected value of the flow of future world output.

The clause concerning the richness of the available asset menu is crucial, because the expected value of world output is the sum of output realizations in different states of nature weighted by state-contingent output prices. If the required set of state-contingent assets does not exist, people generally will not have common marginal rates of consumption substitution across all states of nature, and there is no automatic presumption that investment will be efficiently allocated throughout the world.®

In a world of uncertainty and incomplete markets, it therefore can be difficult to judge how close global investment patterns are to those that free Capital mobility would imply. Researchers hoping to assess capital mobility from this perspective have been forced to rely on very rough measures of constrained investment efficiency.

A number of studies attempt to compare, directly or indirectly through

‘Under restrictive theoretical conditions, an efficient complete-markets allocation can be reached even when a complete set of state-contingent assets is not traded. For different examples, see Rubinstein (1974) and Cole and Obstfeld (1991).

an examination of capital-output ratios, the marginal contribution of installed capital to national outputs. In the presence of capital installation costs, however, this marginal product of capital need not be the same everywhere at every moment. What should be observed under capital mobility is a tendency for time-averaged marginal products of capital in various countries to converge. Correspondingly, world investment should flow disproportionately toward countries where capital is relatively more productive.

A controversial way of evaluating the efficiency of the global allocation of investment is proposed by Feldstein and Horioka (198() and Feldstein (1983). They argue that the productivity of capital in a country is not systematically linked to the determinants of its saving rate and infer that national saving and domestic investment rates should not be systematically associated either if capital is internationally mobile. Other things being equal, a rise in a country’s saving rate should cause a current-account surplus that directs the freed investable resources) toward their most efficient worldwide uses, and an increase in the productivity of a nation’s capital should cause a current-account deficit that draws in savings from abroad. Feldstein and Horioka’s conclusion that this picture does not match the postwar facts has spawned a large literature, which is reviewed in

Section 4 below.

2 Evidence on the Law of One Price Section 1 argued that the least ambiguous evidence on international. capital mobility comes from a comparison of nominal interest rates on onshore and

offshore loans of the same currency. Under perfect capital mobility, the

interest rate on a three-month French franc deposit in Paris, for example, should equal that on a three-month French franc deposit in London.

Numerous studies have compared onshore-of fshore interest differentials or the related covered interest differentials; partial surveys are in Frankel (1993) and Obstfeld (1986). Frankel (1993, table 2.4) reports statistics on the size and variability of covered interest differentials for a range of industrialized and developing countries over the period from September 1982 to April 1988. His conclusion is that by that period, departures from perfect capital mobility, indicated by short-term covered interest differentials, were small for a number of countries (Popper, 1993, reaches the same conclusion regarding long-term differentials). Included in Frankel’s group of financially open economies are Austria, Belgium, Canada, Germany, Hong Kong, Japan, the Netherlands, Singapore, Sweden, Switzerland, and the United Kingdom. For other economies in Frankel’s sample, most glaringly Greece, Mexico, and Portugal, substantial barriers to capital movement apparently remained during the period from 1982 to 1988. This latter group also includes France, Ireland, and Italy, European Community (EC) members (now members of the European Union, or EU) that adopted timetables for capital-account liberalization as part of the single-market program set out in the EC’s Single European Act of 1987.

Table 1 summarizes a set of more detailed and up-to-date data for four industrialized countries, France (panel A), Italy (panel B), Germany (panel C), and Japan (panel D). For each currency, the onshore interest rate is the three-month domestic interbank rate, and the offshore rate is the three-month rate in the London Euromarket. Rates are expressed as basis points per year. Daily Reuters data covering January 1982 through April 1993 (as reported by

Data Resources, Inc.) are used. Because these data did not appear to be

10

Table 1

Domestic Interbank versus Eurocurrency Three-Month Interest Rates:

Daily Data, January 1,

1982, to April 30, 1993

(in basis points at an annual rate)

Period Jan. 1, 1982- Jan. 31, 1987

Feb. 1, 1987- June 30, 1990

July 1, 1990- May 31, 1992

June 1, 1992- April 30, 1993

Period Jan. 1, 1982- Jan. 31, 1987

Feb. 1, 1987- June 30, 1990

July 1, 1990- May 31, 1992

June 1, 1992- April 30, 1993

Period

Jan. 1, 1982- Jan. 31, 1987

Feb. 1, 1987- June 30, 1990

July 1, 1990- May 31, 1992

June 1, 1992- April 30, 1993

-50 (262)

29 (48)

56 (29)

36 (49)

17 (17)

(10)

(9)

(13)

if, ~ rf, it ~ iF, -254 -267 (375) (375) -10 -23 (20) (19) 1 -11 (11) (7) -3 -35 (40) (45) B. Italy Ii, - if in ~ if, -89 -124 (311) (308) 48 -14 (47) (49) 63 9 (36) (29) 28 -8 (50) (43) Cc. Germany Ibu ~ Ibu ibm ~ Ibm 16 5 (17) (18) 3 -8 (10) (12) -5 -18 (8) (9) 5 -6 (12) (12)

10a

214 (336) (17)

-20 (10)

-32 (36)

15 (265)

-91 (47)

-111 (37)

-73 (62)

°E tpm ~

-28 (16)

-15 (10) (8)

-18 (13)

G 1pm

Onshore Ask-Bi.d

13 (3)

13 (4)

12 (8)

32 (20)

Onshoire Ask-B:i.d

34 (10)

62 (20)

55 (24)

36 (42)

Onshore Ask-Bid 11

(4)

10 (2)

13 (2)

11 (2)

Offshore Ask-Bid

40 (49)

13 (10)

19 (5)

34 (38)

Offshore Ask-Bid

74 (57)

43 (7)

47 (6)

45 (33)

Offshore Ask-Bid 13

(3)

13 (3)

13 (1)

13 (2)

Period iy - it i - if ay - If a} - iy Onshore Offshore Ask-Bid Ask-Bid

Jan. 1, 1982- -7 n.a. -20 n.a. n.a. 13

Jan. 31, 1987 (28) (28) (4)

Feb. 1, 1987- -60 n.a. -68 n.a. n.a. 8

June 30, 1990 (33) (33) (3)

July 1, 1990- 9 n.a. 2 n.a. n.a. 7

May 31, 1992 (37) (37) (3)

June 1, 1992- 17 n.a. 10 n.a. n.a. 7

April 30, 1993 (19) (19) (2)

Note: Numbers in parentheses are standard deviations. Subscripts denote asset currency of denomination, franc (Fr), lira (Li), deutschmark (DM), yen (¥); superscripts denote asset location, Lordon Eurocurrency market (E£), France (F), Italy (I), Germany (G), and Japan (J). Underbars denote bid interest rates (the rates banks pay on deposits); overbars denote ask interest rates (the rates at which banks lend funds). Data are daily except for weekends and holidays.

10b

completely accurate, suspicious observations were checked against the daily reports in the Financial Times of London and corrected when necessary.

Many empirical studies ignore the existence of information on both the ask and bid rates of interest at which banks stand ready to supply and accept funds.*® Ask and bid prices are important data in comparing rates of return internationally, however, because the rates at which interbank transactions actually take place are bracketed by the ask-bid spread. In addition, use of the distinct ask and bid rates allows the researcher to test a wider rance of hypotheses about financial market links.

Under free capital mobility, bank borrowers have the option of usirg whichever market is cheapest, and bank lenders can place funds wherever they get the highest net return. Thus, borrowing rates should be the same in all centers where borrowing at the ask rate is occurring, lending rates should be the same in all centers where lending at the bid rate is occurring, and the ask-bid spread should thus be the same in all centers where both activities are occurring at the ask and bid rates.

The first two columns of numbers in Table 1 compute period daily averages of differences between onshore and offshore bid (denoted by an underbar) and ask (denoted by an overbar) rates of interest on loans of domestic currency. As above, the subscripts on the nominal interest rate i refer to currency of denomination, and the superscripts refer to location, either the home country (F for France, I for Italy, G for Germany, J for

Japan) or the offshore Eurocurrency market (symbolized by the letter E). The

‘The price of current money in terms of future money at which a bank is willing to supply current funds is 1 plus the ask rate; it always exceeds 1 plus the bid rate, which is the price of current money in terms of future money that a bank stands ready to pay for current funds.

11

last two columns of Table 1 report average onshore and offshore ask-bid spreads, which must be the same if ask and bid rates are the same onshore and off. Trke use of period averages is not ideal, because large positive and negative daily observations could cancel when the average is taken. The standard deviations given in parentheses below the average return differences offer @ rough idea of the extent to which such cancellation has occurred. Figures’ 2 to 6, which graph the daily data on onshore-offshore bid differences expressied in percentage points per year, also contain some of this information.’

In principle, two financial centers linked by free capital mobility could have different ask rates, if banks are not lending at the ask rate in one ceriter, or bid rates, if no deposits are being taken at the bid rate in one center. This situation seems unlikely to prevail for any Jength.of time, however, and so should not be too problematic for analyzing the period averages reported in the table. In reality, of course, interbank transactions often clo not occur at ask or bid rates. As a stronger test, Table 1 also reports: the returns to a hypothetical arbitrageur who borrows in one center at the ask rate and lends in the other center at the bid rate. The third column is the return to borrowing offshore and lending onshore; the fourth column is the ret:urn to borrowing onshore and lending offshore. Because such arbitrage opportunities would always be exploited, hypothetical arbitrage profits are an unambiguous indicator of capital-market segmentation and must always be absent under free capital mobility. Obviously, the indicators in Table 1 are not independent of each other. For example, offshore-to-onshore arbitrage at ask

and bid rates is profitable only if the onshore bid exceeds the offshore bid

7In comparing these figures, be aware that their left-hand scales differ.

12

FRENCH INTEREST RATE DIFFERENTIAL Onshore—Offshore Bid

Jul-8 2 Jun-84 = Jun-86 =~ Jun-88—Jun—90. Jun 92 Jun-8 3. Jun-85 Jun-87 Jun-89 Jun-9 1

FIGURE 2: French france onshore-offshore bid interest rate differential, January 1982-April 1993

FRENCH INTEREST RATE DIFFERENTIAL Onshore—Offshore Bid

T tha any ami ae ul

Jan-9 2 Apr—-92 Aug—92 Dec—92 Apr—-93

Mar —9 2 Jun-—9 2 Oct-92 Feb—93

FIGURE 3: French franc onshore-offshore bid interest ° rate differential, January 1992-April 1993

12b

ITALIAN INTEREST RATE DIFFERENTIAL Onshore—Offshore Bid

!

Jul-8 2 Jun—8 4 Jun-86 Jun-88 Jun-90 Jun—-9 2 Jun—8 3 Jun-85 Jun—8 7 Jun-89 Jun-9 1

FIGURE 4: Italian lira onshore-offshore bid interest rate differential, January 1982-April 1993

12¢

GERMAN INTEREST RATE DIFFERENTIAL Onshore—Offshore Bid

6 Jul-8 2 Jun-8 4 Jun-8 6 Jun-88 Jun-90 Jun-9 2 Jun—8 3 Jun-85 Jun-—8 7 Jun—-8 9 Jun-9 1

FIGURE 5: German mark onshore-offshore bid interest rate differential, January 1982-April 1993

JAPANESE INTEREST RATE DIFFERENTIAL Onshore—Offshore Bid

Wy yi ‘i i Wyk j fi -0.5 a eae nen i i “Ty

Jul-8 2 Jun-8 4 Jun-86 Jun-8 8 Jun-—90 Jun-9 2 Jun—8 3 Jun-85 Jun-87 Jun-89 Jun—9 1

FIGURE 6: Japanese yen onshore-offshore bid interest rate differential, January 1982-April 1993

12e

and the offshore ask-bid spread is sufficiently small.

The first period analyzed in the table extends through the entry into force of the Single European Act in January 1987. For France (panel A) there is evidence of significant barriers to capital mobility during this period. Average ask and bid rates of interest on French franc loans are much higher offshore than onshore, and the average profitability of hypothetical onshoreto-offshore arbitrage operations is substantially positive. The interpretation of these results is that France maintained controls on capital outflows that kept domestic interest rates below Eurocurrency rates, particularly around realignments (Giavazzi and Pagano, 1985). The especially high divergences occurring around realignments are apparent in Figure 2. Note also that the ask-bid spread is lower onshore than offshore, consistent with the relative thinness of the Eurofranc market in the first half of the 1980s.

The last three periods shown in Table 1 begin roughly around the last French realignment within the European Monetary System’s Exchange Rate Mechanism (ERM) (February 1, 1987), the deadline for abolition of French capital controls under the Single European Act (July 1, 1990), and the month of the surprise initial Danish rejection of the Maastricht Treaty on European monetary and political union (June 1, 1992). This last event set off a period of turbulence in exchange markets that culminated in the "flotation" of ERM currencies on August 2, 1993.

In all three of these periods, the average onshore-offshore difference is on the order of 10 basis points in magnitude for both bids and asks. Hypothet.ical arbitrage profits are negative on average, and average ask-bid spreads are much closer in the two markets. Clearly, the integration of

onshore and offshore money markets is much higher than before 1987.

13

The final period, that of the ERM crisis, is clearly more turbulent than the previous two: the standard errors of returns are much higher, as are askbid spreads. As Figure 3 (an enlargement of the data from January 1992 to April 1993) shows, some large gaps between onshore and offshore bid rates emerged during September 1992, when the franc first came under concerted attack by speculators. Similar data have been identified as evidence cf lingering capital controls by some commentators ("A Funny Thing Happer.ed," The Economist, October 10, 1992, p. 97).

Note in Figure 3 that the onset of ERM turbulence is the dividirg point between a period in which onshore bid rates usually exceed offshore rztes by a small amount, and a period in which the reverse is true. This pattern would be consistent with a shift from a regime in which the market attaches a small but positive probability to future capital controls, to a regime in which mild official discouragements to capital outflow are actually in place. Between September 1992 and April 1993 there are, however, only four instances of pure profits from onshore-to-offshore arbitrage, all in 1992: on September 22 and 23, on November 24, and on December 1.

The case of Italy (panel B of Table 1) also shows evidence of restricted capital mobility before February 1987. Average offshore bid and ask rates both exceed onshore counterparts, and there exists a small mean (15 basis point) profit from undertaking a hypothetical onshore-to-offshore arbitrage.® As Giavazzi and Pagano (1985) observed using a shorter data sample, domestic Italian interest rates diverge less from the corresponding offshore rates than

do French domestic rates during this initial period. Nonetheless, the data are

*The large standard error on this small mean value implies episodically large notional profit opportunities.

14

consistent with the view that Italy, like France, restricted capital outflows and thus held domestic interest rates artificially low. As in the case of France, the ask-bid spread before February 1987 is higher offshore.

The next subperiod, February 1, 1987, to June 30, 1990, shows some convergence to offshore conditions: average onshore rates now rise a bit above average offshore rates, average arbitrage opportunities disappear, and the absolute difference between mean offshore and onshore spreads narrows.

After July 1, 1990, average onshore rates actually rise further above offshore rates, and apparent opportunities for profitable offshore-to-onshore (that is, inward) arbitrage open up (see Figure 4). Italy adopted a narrow ERM band for the lira in January 1990 and then removed its remaining capitalaccount restrictions in May. Subsequently, Italy’s desire to avoid realignment clashed increasingly with the lira’s ongoing real appreciation and with the growth in domestic public debt. The onshore interest premium may have reflectid market fears that capital controls might be reimposed in the future to shore up Italy’s increasingly strict interpretation of its ERM commitments. Consistent with this view is the behavior of the average onshore premium after June 1, 1992, a period that includes Italy’s abandonment of the ERM for a float on September 17, 1992: the average onshore premium drops and average arbitrage profits disappear as one key motive for reimposing capital controls evaporat:es. After September 1, 1992 one (probably spurious) instance of a pure profit from outward arbitrage occurs on January 4, the first business day of 1993.

Panel C of Table 1 shows that, before February 1987, Germany's onshore interest: rates were on average slightly above offshore rates, consistent with

official measures discouraging capital inflow (see also Figure 5). There is

15

even a slight average profit from hypothetical inward arbitrage during this period. Ask-bid spreads, however, are essentially the same in the onshore and offshore markets throughout the full sample period.

In all three subperiods after February 1987, onshore and offshore rates are very close on average and mean arbitrage profits are negative. Some large daily onshore premia emerge during the fall 1992 ERM crisis, however: over the period from September 1992 to April 1993, offshore-to~onshore arbitrage appears profitable on 51 out of 242 business days! This pattern may reflect continuing government intervention in the capital markets. Goldstein et al. (1993, p. 56) mention the "gentlemen’s agreement" whereby the Bundesbank may impose high marginal reserve requirements on loans in excess of a certain limit to German banks from their London branches.

For Japan (panel D of Table 1), a less complete set of data were available from Reuters. The available data show a very small average difference between onshore and offshore bid rates over the first sample subperiod, consistent with Japan’s substantial liberalization of capital movements in December 1980.°

Surprisingly, the subperiod beginning with February 1987 shows é. 60 basis-point average excess of offshore over onshore bid rates; Figure 6 makes clear that this differential is much too long-lived to ascribe to the time-ofday difference in the Japanese onshore and offshore data. Ueda (1993, p. 19) suggests that, before November 1988, the Bank of Japan used heavy administrative guidance to separate the interbank loan market from both the

onshore certificate of deposit market and the Euroyen market; during the

‘Marston (1993a) examines differences in Japanese and U.S. short-term interest rates and reviews related literature.

16

subpericd in question, the Bank of Japan wished to hold interbank rates below onshore and offshore open-market rates.' Thus, the onshore-offshore gaps in Figure 6 indicate a segmentation within the domestic financial market that, as a side effect, insulated part of that market from global forces.

Over the last two subperiods, the mean onshore bid exceeds the mean offshore bid by relatively small amounts. The ask-bid spread in the Euroyen market is so slim that even the small onshore bid premium implies positive average arbitrage profits from borrowing offshore and investing onshore. These divergerices grow stronger in the period starting with June 1992. In light of the data’s imperfections, it is unwise to put much weight on these numbers as indicators of capital-market restriction. Faced with a punctured "bubble" economy and a rising yen in these years, however, Japanese officials did have incentives to discourage capital inflows through informal means.

What conclusions follow from these and similar data for other industrial countries? For the four countries in Table 1 as well as for others such as the United States and United Kingdom that have liberalized international financial transactions, there are extremely close links between onshore and offshore money markets, links that increased in strength over the 1980s. The data also show, however, that actual or prospective government interventions remain a significant factor in times of turbulence. Industrial-country governments still have instruments that can drive at least temporary wedges between onshore and offshore interest rates. European countries that have not completely opened their capital accounts, such as Greece, Ireland, Portugal,

and Spa:n, openly retain such instruments; all four used them during the ERM

In November 1988, the bank took measures liberalizing the interbank market.

17

IRISH INTEREST RATE DIFFERENTIAL Onshore—Offshore Bid

}—

—6 Oct-86 Oct-87 Oct-88 Oct-89 Oct-90 Oct-91 Oct-92 Apr-87 Apr-88 Apr-89 Apr-90 Apr—-91 Apr—-92 Apr—-93

FIGURE 7: Irish punt onshore-offshore bid interest rate differential, October 1986-April 1993

17a

currency crisis that began in 1992 (Goldstein et al., 1993; Committee of Governors, 1993). Even these countries tend to have strong links to world Capital markets. For example, Ireland’s onshore and offshore interest rates were close on the whole from the late 1980s through 1992 (see Figure 7). Matters are different in the developing world, where much higher explicit or implicit barriers to capital flows remain common. Discussions of financial liberalization and international interest-rate linkages for developing regions can be found in Calvo, Leiderman, and Reinhart (1993),

Glick and Hutchison (1990), and Mathieson and Rojas-Suarez (1993).

3 The Diversification of Global Consumption Risks

Researchers have taken several approaches to studying the world capital market’s success in helping countries trade consumption risks to achieve a mutually preferable allocation of consumption across states of nature. Some look directly at national or regional correlations in consumption. Others look at the extent of trade in explicitly state-contingent assets. As will become apparent in the discussion, the implications of such data for capital mobility are ambiguous unless specific and strong side assumptions are made about the functioning of domestic and international capital markets. Much recent research is aimed at testing these assumptions, and, as difficult as the task is, it is justified by the need to understand better the current and potential risk-allocation role of world capital markets.

International Consumption Correlations

A simple maximization problem illustrates how global consumption allocations would behave in the ideal case of perfect international trade in a complete

set of state-contingent assets. Because the resulting allocation is Pareto-

18

optimal, its properties can be read off from the first-order conditions that a world planner would derive in maximizing a social welfare function lirear in national utilities.

An analytically convenient starting point is the assumption of a representative national agent for each country. This assumption, which will be discussed further below, amounts to supposing that risks have already been shared optimally within each country, leaving only the remaining gains from trade between countries as the analytical focus. Country j‘s representative agent maximizes (from time t = 0) the expected utility function

Up = El S'w(cr, x)

20

where 5 e€ (0,1) ig a discount factor, ci) (as before) is consumption of an internationally tradable good, and x/ is consumption of a nontradable good (possibly leisure) ." :

Given N countries and fixed country welfare weights w, J = 1, 2,..., N,

the planner maximizes the social welfare function N W = Yo wd j=l

by distributing the tradable consumption available on each date, and in each state, among the N countries. If c¥ is world tradable consumption on date t, a

necessary condition for distributing it efficiently among countries is

‘This formulation already imposes strong restrictions on national utility functions (for example, time and state separability), and more will be imposed later. Without some assumptions on preferences, however, no observable implications of international risk sharing could be derived.

19

wu (ci, x/) = wuj(c/, x/) (for all countries jandl), (1)

where u,(c,x) is a partial derivative with respect to c. Equation (1) implies that, for tradable goods, marginal rates of substitution across states of nature are equalized internationally in an efficient allocation. Because nontradables cannot be shifted among countries, however, the corresponding condition on marginal utilities from the nontradables need not hold.

To derive more specific predictions from (1), suppose that no

nontradables x are consumed and that utility functions have the specific form ul(c/, x4) = (1 -R,)1(c4)'*% . Then, if ¢, = loge, - loge,,, (1) implies G] = (R,/R,) G) 5 (2)

that is, with isoelastic preferences, logarithmic growth rates of consumption are perfectly correlated ex post in all countries. If countries have different (fixed) rates of time preference, equation (2) will contain a constant term, but the perfect correlation prediction will still hold true.

If capital is internationally mobile but asset markets are incomplete, conditicns weaker than perfect correlation will characterize the relation between countries’ ex post intertemporal marginal rates of substitution. As noted akove, informational asymmetries generate moral hazards; these or other problems can make certain risks uninsurable. In the extreme case in which only a riskless consumption-indexed bond is traded among countries, expected, but not ex fost, intertemporal marginal rates of substitution will coincide internationally. This case is the one analyzed in stochastic versions of the

life-cycle/permanent-income hypothesis. If only nominally risk-free bonds are

20

traded, expected intertemporal marginal rates of substitution for money (rather than consumption) will be equalized (Kollman, 1992; Obstfeld, 1989). More generally, ex post cross-country differences in intertemporal marginal rates of substitution will be uncorrelated with any random variables on which international contracts can be written. Under incomplete markets asset trade allows the sharing of some, but not all, risks.

To compare reality against the predictions of the specific completemarkets model just set out, Table 2 examines the correlations of national annual real private consumption growth rates, measured in per capita terms, with rest-of-world per capita private consumption growth over two eras in the development of world capital markets, 1951 to 1972 and 1973 to 1988.” The consumption data come from the Penn World Table assembled by Summers and Heston (1991); the "world" shown in Table 2 consists of countries with continuous 1950-88 data rated of quality C- or above by Summers and Heston.

All the correlation coefficients, denoted p(é,é”), where c” is rest-of-world real per capita consumption, are below the value of 1 that would obtain with a common world isoelastic utility function were capital perfectly mobile and markets complete. Several regularities in the results are, however, apparent.

For the post-1973 period--a period during which the volume of international financial transactions has increased enormously relative to

world output--consumption growth in industrial countries is, on average,

“The current model implies that each country’s consumption growth is perfectly correlated with world consumption growth if all countries have the same value of R,. Looking at correlations with world consumption growth, rather than at the customary pairwise consumption-growth correlations, economizes on the number of estimates reported. This procedure also has some potential statistical advantages (see Obstfeld, 1994a).

21

Table 2 Consumption and Output Correlations: International Data, 1951-1972 and 1973-1988

Correlation 1951-1972 Correlation 1973-1988

Country p(é,é") P(¥,9") p(é,¥) p(é,é") P(¥, 9") p(é,¥)

Industrial countries EU members

Belgium 0.50 0.47 0.66 0.49 0.58 0.81 Denmark 0.09 -~0.04 0.75 0.60 0.39 0.80 France 0.26 0.41 0.64 0.50 0.56 0.71 Germany -0.11 0.31 0.78 0.72 0.87 0.68 Greece -0.10 0.03 0.69 0.13 0.41 0.56 Ireland 0.58 0.58 0.77 0.48 0.57 0.76 Italy -0.02 0.35 0.62 0.27 0.61 0.90 Luxembourg 0.14 -0.18 0.20 0.21 0.73 0.19 Netherlands 0.49 0.27 0.77 0.56 0.59 0.75 Portugal -0.10 0.18 0.55 0.06 0.44 0.89 Spain -0.33 0.01 0.90 0.32 0.39 0.93 United Kingdom 0.29 0.49 0.60 0.59 0.66 0.81 Others

Australia 0.39 0.06 0.88 -0.00 0.72 0.66 Austria 0.33 0.27 0.59 0.29 0.55 0.71 Canada 0.43 0.42 0.71 0.10 0.30 0.90 Finland 0.20 0.34 0.82 0.19 0.06 0.46 Iceland 0.17 -0.18 0.91 0.05 0.27 0.85 Japan 0.06 0.43 0.57 0.62 0.71 0.86 New Zealand 0.38 -0.07 0.81 -0.03 0.16 0.77 Norway 0.36 0.01 0.56 0.05 0.37 0.43 Sweden 0.27 0.07 0.74 0.18 0.04 0.36 Switzerland 0.32 0.50 0.56 0.64 0.53 0.79 United States 0.26 0.19 0.59 0.31 0.67 0.81

Developing countries

Argentina 0.00 0.02 0.96 -0.04 0.25 0.92 Bolivia -0.07 0.34 0.59 0.29 0.08 0.74 Chile -0.32 0.02 0.69 0.44 0.62 0.85 Colombia 0.28 0.11 0.89 0.29 0.51 0.79 Costa Rica 0.15 -0.09 0.89 0.63 0.65 0.95 Cyprus 0.20 -0.04 0.62 0.64 0.64 0.92 Dominican Rep. 0.03 0.10 0.92 0.11 0.26 0.88 Ecuador -0.01 0.19 0.63 -0.17 0.05 0.67 El Salvador 0.38 0.21 0.89 0.56 0.44 0.95

2la

Table 2 (continued)

Correlation 1951-1972 Correlation 1973-1988 Country p(é,é") P( 9,9") p(é,9) p(é,é") p(¥, 9") p(é,F) Guatemala -0.28 -0.40 0.81 0.39 0.48 0.95 Honduras 0.16 0.20 0.58 0.54 0.68 0.91 India -0.13 -0.09 0.59 -0.13 -0.16 0.93 Kenya -0.04 0.24 0.93 -0.08 0.20 0.82 Mexico -0.01 0.22 0.92 -0.27 0.02 0.98 Morocco -0.18 -0.05 0.94 0.22 -0.04 0.62 Paraguay 0.13 -0.21 0.78 -0.32 0.01 0.93 Pakistan 0.03 0.33 0.59 -0.20 0.06 0.44 Peru 0.11 0.35 0.60 -0.26 -0.18 0.94 Philippines 0.03 -0.15 0.77 -0.06 -0.12 0.80 South Africa 0.39 0.20 0.85 -0.49 -0.10 0.88 Thailand -0.27 -0.23 0.94 0.51 0.61 0.84 Trinidad&Tobago-0.20 -0.09 0.69 -0.30 -0.33 0.95 Turkey -0.13 0.21 0.96 0.06 -0.18 0.86 Uruguay 0.17 0.42 0.95 0.09 0.28 0.90

Note: The numbers p(é,é"), or p(y,y”), are simple correlation coefficients between the annual change in the natural logarithm of the country’s real per capita consumption (or output) and the annual change in the natural logarithm of the rest of the world’s real per capita consumption (or output), with the "world" defined as the sample liszed in the table. National per capita consumptions and outputs were calculated using variables 1, 3 and 6 listed in appendix A.1 of Summers and Heston (1991). The numbers p(¢,y) are

correlations between each country’s log consumption per capita and log out:.ut per capita changes.

,

21b

somewhat more highly correlated with rest-of-world consumption growth than is consumption growth in developing countries. Within the group of industrial countries, however, there are sharp differences.

For a narrow majority of EU members, domestic and world consumption growth are relatively strongly correlated; Greece, Portugal, and Spain, which still maintain capital controls, as well as Italy, which did so through early 1990, are in the minority, as, surprisingly, is Luxembourg. For virtually all EU countries, and most dramatically for Germany, the correlation coefficient ‘rises between the first and second subperiods. Multiple regressions show that this last result persists even after one controls for possible parallel responses to the two OPEC oil-price shocks (see Obstfeld, 1994a, for further discussi.on).

For industrial countries outside the EU, the consumption correlations tend to be lower in the recent period except for Switzerland and Japan. Moreover, apart from those two countries, there is a tendency for the correlations to decrease, not increase over time. To explain the contrast with the EU countries would require a country-by-country analysis. One general factor, however, may be the exchange-rate regime: these countries opted for greater exchange-rate flexibility than the EU countries in the early 1970s partly hecause they desired to decouple domestic from world consumption growth. The Japanese example shows, however, that floating exchange rates, and even capital controls (which persisted in Japan through 1980), need not rule out a strong coherence between domestic and world consumption growth.

One way to highlight the change in German and Japanese consumption behavior after 1973 is through a simple regression. Let y denote country j’s

real per capita GDP, inv its real per capita investment, and g/ its real per

22

capita government spending. Absent international asset markets, domest.ic per

capita consumption c would be limited to y - inw - g. The regression Gr = a + a6," + a,Alog(y/ - inv! - g/) +

gives an indication of whether consumption growth is more strongly assaciated with global or with domestic factors.’ The Summers-Heston data lead to the

following results:

Germany Japan 1951-1972 a, = -0.18 , a, = 0.76 a, = -0.15 , a, = 0.76

(0.33) (0.13) (0.37) (0.13) 1973-1988 a, = 1.07 ,a,=9.02 a = 1.18, a, = 0.35

(0.32) (0.20) (0.42) (0.26)

The regressions show a stunning reversal for both countries. In the earlier period, national consumption growth is insignificantly correlated with world consumption growth but moves nearly one-for-one with the growth of GDP net of investment and government spending. From 1973 on, the opposite is true.

A fundamental identification problem is suggested by the columns iin Table 2 labeled pP(9,9"), which report correlations between national per capita output growth rates and rest-of-world per capita Output growth. For most: of the industrial countries, these correlations rise between the two subperiods shown. Thus, while any increase over time in the correlation between national and world consumption growth could be due to increased risk sharing through the international capital market, it could also be explained by other mechanisms, such as a naive Keynesian consumption function in which consumption merely tracks current Output or by one of the richer behavioral

models discussed by Carroll and Summers (1991). The Table 2 correlations

eee

°see Obstfeld (1994a) for more discussion of this equation and its estimation.

23

p(¢,¥) between domestic output and consumption growth are high in most cases, although often they are well below unity.

Again, only country-by-country analysis can resolve this question. For example, tests reported in Obstfeld (1994a) suggest that the high post-1973 correlation of Japanese with world consumption growth may reflect only the high correlation coefficient between world consumption and Japanese output (0.72), coupled with the high correlation of Japanese consumption and output. In contrast, German output growth also has a very high correlation coefficient with world consumption growth (0.84) yet adds no significant explanatory power to a regression of German on world consumption growth. These regressions are somewhat: analogous to those Campbell and Mankiw (1991) examine in modeling departures from the permanent-income theory.

Aniong the developing countries in Table 2, a few have reasonably high post-1973 correlation coefficients with world consumption growth--notably, Chile, Cyprus, Thailand, and a few Central American countries. But this is not the norm. Note that the developing countries with high post-1973 values of p(é,é”) also have high values of p(¥,y").

Before drawing strong conclusions from Table 2 about feasible gains from risk sharing, recall that (2) was based on some restrictive auxiliary assumpt:ons, for example, the assumption that nontradables are not consumed. If some consumption goods are nontradable, there is no necessity for national consumpt:ions to be perfectly correlated: risks relating to the consumption of nontraded goods may be impossible to share (Stockman and Dellas 1989). At best, consumption of tradables will obey (2) if the utility function w(c,»)

is separable (but still isoelastic in c). In more complicated models, even

24

this simple property can fail despite complete markets. '4

By investigating the stochastic consequences of a labor-leisure tradeoff and/or nontradables, several studies have tried to reconcile consumption correlations such as those shown for the industrial countries in Table 1 with complete markets and perfect capital mobility.

Backus, Kehoe, and Kydland (1992) and Stockman and Tesar (1990) cbserve that the pairwise correlation coefficients between (Hodrick-Prescott (1980} filtered) industrial-country consumption levels tend to be lower than the corresponding output correlations. This property of the data is quite evident in Table 2: after 1973, p(é,é”) exceeds P(¥,9") only for Denmark, Finland, Sweden, and Switzerland among twenty-three industrial countries. Backus, Kehoe, and Kydland fail to replicate this pattern using a plausibly calibrated two-country intertemporal production model with uncertainty.

Stockman and Tesar introduce nontradable consumption into a similar equilibrium business-cycle model and find that the addition of preference shocks allows a closer approximation to the empirical correlation coefficients for national consumptions and outputs. Devereux, Gregory, and Smith (1992) show that a specific utility nonseparability between consumption and labor Supply allows an equilibrium business-cycle model to replicate the U.S.-: Canadian consumption-correlation coefficient. They do not, however, sub‘ect their model to the tougher test of fitting other moments of the data. Van Wincoop (1992c, table 1) adjusts annual 1970-88 consumption data from tke United Nations System of National Accounts for both nontradability and durability. He finds that for most industrial countries, the correlation

between the growth of adjusted domestic per capita consumption and adjusted ee

4Stulz (1981) addresses these questions in a general setting.

25

world per capita consumption is much higher than in Table 2 above (albeit still imperfect). His calculations do not, however, control for the possibility that correlations are also higher among the growth rates of similarly adjusted per capita domestic outputs.

Lewis (1993) carries out a panel study of the growth of nondurable, tradable consumption using data from forty-eight countries sampled at fiveyear intervals from 1970 to 1985. Remarkably, she finds that, although domestic: output growth is a strong and significant determinant of total consumpt:ion growth in her panel, its effect on nondurable, tradable consumption growth is statistically insignificant; furthermore, domestic output growth explains less than a hundreth of the dependent variable’s variance (as opposed to about two thirds of the variance of total consumption growth). Although imprecisely estimated, the coefficient of output growth in Lewis’s equation for nondurable, tradable consumption remains sizable. In light of possible measurement errors, and her panel methodology’s merging of countries with different degrees of financial openness, a judicious conclusion is that durability and nontradability go part, but probably not all, of the way in explaining why total consumption growth is highly correlated with domestic: output growth. Lewis does not look at the influence on consumption of idiosyncratic factors other than income growth, so her results do not explain why, as in van Wincoop’s (1992c) study, international consumption correlations remain imperfect even after attention is restricted to nondurable tradables.

The message of this body of work seems to be that, after allowing for nontradebles and durables, equilibrium complete-markets models that assume

perfect capital mobility still cannot provide a satisfactory explanation of

26

interuational consumption correlations unless unexplained preference shifts are assumec as in Stockman and Tesar (19903. Taste shocks are not anherencly SMDLBUS.OLe, out. until thev are modeled more fuliy, there is ao way of telling if the heavy explanatory burden they bear in the Stockman-Tesar modei is reasonable."

An alte. iative approach starts by acknowledging that the assumption of complete asset markets is glaringly at odds with the facts. Events such as job loss generally are not completely insurable because of the potential for moral hazard. More generally, labor incomes cannot be privately insured against all contingencies. Some shocks simply cannot be foreseen with sufficient clarity to be provided for in contracts. Thus, even with free and costless international trade in the same range of assets traded domestically, there is no reason to expect high correlations even between the tradable-goods consumptions of different countries.

Empirical studies of U.S. microeconomic data, such as Cochrane (1991), Mace (1991), and Mankiw and Zeldes (1991), confirm that, even within modern industrial economies, there are unexploited opportunities for risk sharing.'® In line with this conclusion, van Wincoop (1992b) finds that the cori-elations among (Hodrick-Prescott filtered) per capita consumption levels in Japanese

prefectures are well explained by a simulation model in which domest:.c

Canova and Ravn (1993), Lewis (1993), and Obstfeld (1994a) all allow for preference shocks in their formal tests of consumption risk-sharing nodels. In tests on quarterly data for nine OECD countries, Canova and Ravn find little evidence against moment restrictions implied by a model based on equation (2) above. They do, however, reject long-run implications of the model.

‘Indeed, Altonji, Hayashi, and Kotlikoff (1992) find such unexpl.oited opportunities even within extended U.S. families. Deaton (1992, p. 3%), who surveys the related microeconomic literature, reminds us that moral hazard problems arise even within families.

27

Japanes2 financial markets are incomplete and subject to limited participation.'’ Work by Baxter and Crucini (1993a) and Kollmann (1993) suggests that general-equilibrium real business cycle models in which countries trade only consumption-indexed bonds can mimic the actual stochastic behavio:: of consumptions and outputs far better than can otherwise similar models ‘shat assume complete asset markets.

These considerations have three implications for the class of models discussed so far in this section. First, the representative national consumer is a hypothetical construct that, although perhaps useful for illustrating the incremental gains from international compared with national risk sharing, gives a misleading picture of how national consumption levels actually are determined. Second, imperfect correlations among industrial-country consumpt:ions are likely to be in large measure the result of generalized asset-market incompleteness rather than of international capital-market segmentation. Third, studies of international consumption-correlatedness that counter/iactually assume complete markets probably cannot throw much light on the international mobility of capital. A more fruitful approach is to consider models admitting alternative financial-market structures (for example, Cole 1988) and, ultimately, models in which market incompleteness arises endogenously (for example, Gertler and Rogoff, 1990; Lucas, 1992).

Comparing Regional and International Risk Sharing If asset: markets are incomplete, is there any way that consumption correlat.ions or related measures can throw light on the extent of

international capital mobility? Atkeson and Bayoumi (1992) propose an

"Ven Wincoop (1992a) shows that such a model also can rationalize crosscountry consumption correlations.

28

imaginative approach to this problem: they use the measured extent o: regional risk sharing within the United States as a benchmark against which the efficiency of international risk sharing among a group of industrial countries can be judged. In principle, this methodology can help determine the extent to which low international consumption correlations are due to internat:.onal asset-trade barriers as opposed to incomplete markets.

The findings, although generally pointing to higher regional than international financial integration, are somewhat ambiguous. Regional). financial transfers within the United States appear to be much large): in absolute value than resource transfers into or out of the main indust:rial countries, suggesting more extensive asset trade within the United States. In contrast, U.S. regional growth in real retail sales (a consumption proxy) is no less correlated with regional ouput growth than is OECD national consumption growth with national output growth.

Atkeson and Bayoumi also find that, in U.S. data, shifts in state capital income are virtually uncorrelated with state capital product but are highly correlated with U.S. capital income. In Europe, national capital incomes, although uncorrelated with national capital products, seem much less correlated than in the United States with total European capital income. Atkeson and Bayoumi interpret this result as indicating better capital-income diversification within the United States than within Europe.

Table 3 provides another regional-~to-international comparison using yearly data assembled by Robert Dekle on per capita consumption and income

(which is interpreted here as an output proxy) in 45 of the 47 Japanese

29

Table 3 Consumption and Output Correlations by Prefecture: Japanese Data, 1975-1988

Prefecture p(é,é) P(¥-¥) p(¢,¥) Prefecture p(é,&) P(¥, 7) pP(é,9) Hokkaido 0.595 0.165 0.339 Kyoto 0.682 0.149 0.778 Aomori -0.096 0.196 0.905 Osaka 0.719 0.053 0.776 Miyagi 0.750 0.555 0.420 Hyogo 0.480 -0.000 0.742 Akita 0.219 0.433 0.367 Nara 0.181 0.766 -O.2i1 Yamagata 0.496 0.303 0.748 Wakayama 0.136 0.105 0.455 Fukushima 0.065 0.386 0.898 Tottori 0.413 0.858 0.492 Ibaraki 0.077 0.205 0.630 Shimane 0.170 0.551 0.717 Tochigi 0.100 0.115 -0.589 Okayama 0.245 0.103 -0.568 Gunma 0.644 0.668 0.444 Hiroshima 0.661 0.075 0.736 Saitama 0.404 0.337 0.696 Yamaguchi 0.777 0.331 -0.201 Chiba 0.547 0.267 0.693 Tokushima 0.313 0.613 0.705 Tokyo 0.238 0.055 0.978 Kagawa 0.610 0.494 0.555 Kanagawa 0.240 -0.015 0.872 Ehime 0.277 0.215 0.577 Yamanashi 0.658 0.513 0.567 Kochi 0.070 0.122 0.115 Nagano 0.252 0.358 -0.474 Fukuoka 0.319 0.123 0.569 Shizuoka 0.297 0.415 0.081 Saga 0.505 0.534 0.913 Toyama 0.098 0.232 -0.713 Nagasaki -0.218 0.254 0.704 Ishikawa 0.723 0.380 0.764 Kumamoto 0.059 0.221 0.907 Gifu 0.258 0.423 -0.313 Oita -0.020 0.096 0.537 Aichi 0.349 -0.004 -0.265 Miyazaki 0.010 0.528 0.824 Mie 0.039 0.211 -0.618 Kagoshima 0.046 0.218 0.982 Fukui 0.012 -0.106 0.849 Okinawa -0.249 0.036 0.949 Shiga 0.625 0.602 -0.142

Note: The numbers p(é,é@’), or p(y,jf’), are simple correlation coefficients between the annual change in the natural logarithm of the prefecture’s real per capita consumption (or output) and the annual change in the natural logarithm of the other forty-four prefectures’ average real per capita consumption (or output). The numbers p(é,/) are correlations between prefecture log consumption per capita and log output per capita changes. Data were supplied by Robert Dekle.

29a

prefectures from 1975 to 1988.'* The column labeled p(é@,é@) shows the correlation of prefectural per capita private consumption growth with mean per capita consumption growth in the other 44 prefectures; these numbers are similar on the whole to those reported for countries in Table 2. Slightly less than half the time, the consumption correlations are below the corresponding income correlations, labeled p(y,y’). The column labeled p(é,¥) shows the correlation between per capita consumption and income growth by prefecture. In about two-thirds of the cases, these numbers are rather high, as are most of the corresponding numbers for national economies in Table 2; but, in other cases, the correlations are relatively low and are sometimes even negative. Although there is thus some limited evidence that risk sharing within Japan may be more efficient than is risk sharing among industrial countries, this is not evident in the intranational consumption correlations.

In contrast to these results for Japan, Crucini (1992) finds in annual data for 1971 to 1990 that consumption growth rates among Canadian prcvinces are generally more highly correlated than are provincial output growtFk. rates or different countries’ consumption growth rates.

A problem in comparing regional risk sharing within nations with. risk sharing among nations when asset markets are incomplete is that a precominance of uninsurable country-specific shocks can create a spurious impression of greater risk sharing within than between countries. Another drawback cf the method is that more goods are nontradable across national borders thar across regional borders, so that, other things being equal, one would naturally

expect interregional consumption correlations to be higher than international

’See Dekle (1993) for a description of these data and an economet:ric analysis of their implications for interregional capital mobility.

30

ones. Finally, government-mediated transfers and spending play a vital role in pooling risks within national borders. It is conceivable that any finding of higher interregional compared with international consumption correlation is entirely an artifact of redistributive domestic fiscal policies. Despite these and other ambiguities, however, refinements of the general approach described above offer the promise of a better understanding of how international and intranational financial linkages differ.

Phe Extent of International Portfolio Diversification

Further evidence on the world capital market’s promotion of risk sharing among countries comes from a direct examination of international portfolio positions. The consensus of studies such as French and Poterba (1990, 1991), Golub (1991), and Tesar and Werner (1992) is that there is a substantial "home bias" in the portfolios of industrial-country investors. French and Poterba and Tesar and Werner argue that conventional models of portfolio choice can expiain these patterns only if domestic investors have a much more optimistic view of the expected return on domestic assets than do foreign investors. Alternatively, imperfect capital mobility simply could make extensive interna:zional diversification prohibitively costly or infeasible. But, in view of the efficiency of international interest-rate arbitrage among industrial countries (Section 2), no one believes that transaction costs or official impediments to foreign investment are universally high enough fully to explain the home bias in equity portfolios. Thus, there is an international

diversiiication puzzle.”

“Dumas (1994) surveys models of international portfolio choice from the perspeci:ive of the international diversification puzzle and other asset-market puzzles. Current trends such as the rapid recent growth of international stock-market mutual funds suggest that the diversification puzzle may well disappear early in the twenty-first century.

31

One widely cited estimate reports that, in December 1989, U.S. investors held 94 percent of their stock-market wealth in home equities; Japanese investors, 98 percent; and U.K. investors, 82 percent (French and Poterrba, 1991). These figures apparently do not control for holdings by "home"-|lbased corporations of assets located abroad, for example, Nissan’s Sunderland, U.K. auto plant. Investors may diversify, moreover, through holdings of assets other than equities, such as direct investments and bonds. French and Poterba (1991) report, for example, that 79 percent of German corporate equity was domestically owned at the end of 1989, which suggests a substantial home bias in German investors’ portfolios. Germany’s December 1991 gross external assets, however, amounted to 72.9 percent of its GDP and its gross external liabilities to 51.4 percent of its GDP--numbers that could be indicative of extensive foreign diversification.” Such diversification might help explain the robust correlation of German with world consumption growth noted above.

The German case may be atypical; U.S. and Japanese investors, for example, probably have not used foreign diversification opportunities as extensively.” Several explanations for this puzzle have been proposed. Stockman and Dellas (1989) argue that the presence of nontraded goods and

services may impart a significant home-asset bias to investors’ portfclio

Data on total German external assets and liabilities come from [ieutsche Bundesbank (1993, p. 45). I have supplemented these numbers with a 191 GDP estimate of $1.58 trillion.

7For the United States, external assets were 34.5 percent of GDP at the end of 1991, and external liabilities were 40.9 percent. The corresponding Japanese figures are 59.2 percent (external assets) and 47.9 percent (external liabilities). Position data come from Deutsche Bundesbank (1993, p. 45). My GDP estimates are $5.68 trillion for the United States and $3.39 trillion for Japan. These figures show considerable growth over the comparable 198°’ figures reported by Brainard and Tobin (1992, p. 536). Their numbers show that:, for the United Kingdom, assets and liabilities already exceeded GNP in 1987.

32

decisions. The empirical importance of home-asset bias due to nontradables remains to be established, however.” Another explanation hinges on the argument that the appropriate criterion for evaluating a country’s gains from international risk pooling is not the impact of global portfolio diversification on the statistical distribution of national equity investment income, bui:, rather, the scope for raising mean consumption and lowering its variance. iind, if this scope is limited, international diversification may be discouraged by even minimal investment barriers such as small transaction costs.

Cole and Obstfeld (1991) use a model calibrated to U.S. and Japanese data to illustrate that at the aggregate or national level, the efficiency gains from risk sharing among industrial countries may be as small as 0.2 percent of GNP per year.” Golub (1991) takes issue with this result, arguing on the basis of 1970-87 data that, despite small aggregate gains, Japanese and U.S. recipients of exclusively corporate income cannot pool risks with human or noncorporate capital and, as a result, would gain substantially from freer asset trade. Thus, strong individual incentives for cross-border diversification might remain. Van Wincoop’s (1992a) calibration model similarly implies that owners of capital can face significantly stronger incentives to diversify than aggregate consumption figures suggest. A useful extension of this line of work would attempt to distinguish empirically between the labor incomes of stockholders and nonstockholders.

Brainard and Tobin (1992) and Baxter and Jermann (1993) argue that

alternative theoretical models of home-asset bias are proposed by Eldor, Pines, anc. Schwartz (1988), by Tesar (1993), and by Feeney and Jones (1994).

3See also Mendoza (1991a) and Obstfeld (1992), who present alternative estimates of small industrial-country gains from asset trade.

33

because human capital is largely nontradable, its owners have a strong incentive to go short in domestic equities and long in foreign equities when the returns to domestic human and physical capital are positively correlated. Whether this deepens the home-bias puzzle in practice requires further research on the international correlations among returns to human and physical capital. Golub (1991), for example, shows that human and physical capital returns (measured by labor income and corporate profits, respective:.y) appear negatively correlated for Japan, and that the optimal portfolio of a Japanese worker can be skewed toward home equities. This inference depends, however, on Golub’s assumption that the national-income account proxies he uses to measure returns to human capital and to equities adequately capture the true statistical relationship between those variables.

Even the magnitudes of the aggregate national gains from risk sharing among industrial countries are in dispute. Van Wincoop (1992c), who examines a larger sample of countries, assumes a lower rate of time preference, and allows for some nondiversifiable consumption risk, finds national gains from risk sharing much larger than those found by Cole and Obstfeld (1991). Obstfeld (1994c) shows that financial integration can bring very large welfare gains if diversification has effects on investment and output growtn rates.” Before the puzzle of low diversification is resolved, more work on understanding both the magnitude and distribution of the gains from international risk sharing is needed.

The importance of transaction costs also is unclear. Cole and Obstfeld

argue that small transaction costs--for example, the extra paperwork needed to

“The model in Obstfeld (1994c) is based on constant expected returns to investment. An alternative model of how diversification affects growth, based on learning-by-doing effects, is proposed by Feeney (1993).

34

obtain a tax credit for asset income withheld by a foreign government--could substantizlly discourage international diversification. Backus, Kehoe, and Kydland (1992) confirm this as a theoretical possibility. They show that introducing small costs of international transactions into their empirically calibrateci model leads to an equilibrium very close to the autarky allocation. This result, however, is based on a representative-agent model that may seriously understate individual, as opposed to aggregate, gains from trade. Tesar and Werner (1992) find that the turnover rate for foreign equity investment:s is higher than that in domestic equity markets and offer this difference as evidence that transaction costs are not important in promoting international equity-market segmentation. Transaction costs other than turnover costs could, however, be important impediments to cross-border investments.

To summarize, the available data on international portfolio positions suggest that many industrial countries are not diversified nearly to the extent that standard models of global portfolio choice would predict. The reasons could range from transactions costs to internationally asymmetric information (Gehrig, 1993) to differential tax treatment of domestic and foreign investors (Gordon and Varian, 1989) to irrational expectations concerning the relative returns on domestic and foreign investments.” Future progress in unraveling the apparent puzzle may come from a more disaggregated analysis of the investing behavior of different income groups. Even at the aggregate level, more detailed information on national balance sheets would

give a be:ter perspective from which to evaluate the risk and return

*Mor:cis Goldstein has suggested that there is also a noticeable regional bias in international investment, a phenomenon consistent with the notion that informational barriers to international investment are important.

35

characteristics of national portfolios.

Such analyses might throw light on the related outstanding puzzle of reconciling convincingly the possibly small aggregate gains from pooling national consumption risks with the apparently large unexploited gains to expected wealth maximizers from international equity diversification. The literatures on stock-market volatility and the equity-premium puzzle show how hard it is to rationalize the behavior of equity returns on the basis of simple optimal-consumption models with representative national consumers (see, respectively, Grossman and Shiller, 1981, and Mehra and Prescott, 1985). Asset prices that appear excessively volatile from the perspective of such models could easily give rise to the divergent estimates of international diversification gains. Partly explaining both the discrepancy in efficiencygain estimates and the asset-pricing puzzles is imperfect domestic risk sharing, as suggested by Mankiw and Zeldes’s (1991) observation that. U.S. stockholders have more variable consumption than have nonstockholders. Even this finding, however, does not enable Mankiw and Zeldes fully to resolve the equity-premium puzzle for the United States. It remains to be seen if generalequilibrium models assuming realistically imperfect asset markets or some form of asset-market segmentation can rationalize both equity-price behavior and the coexistence of small aggregate gains from international risk pooling with large private gains to equity holders.* Such models would, in turn, provide useful vehicles for understanding why limited international equity diversification has persisted.

Gains from Risk Sharing by Developing Countries

Even if it is true that industrial countries would reap only modest gains from

*yan Wincoop (1992a) is a partial step in this direction.

36

further international pooling of risks, there is little doubt that developing . countries could benefit enormously.

Lucas (1987) proposed the thought experiment of eliminating the variability of U.S. consumption around its trend path. For the United States and for most other industrial countries, the aggregate or social benefit this hypothetical event would confer is small, far less than 1 percent of GNP per year in mnost cases. These small numbers are upper bounds on the aggregate gains to industrial countries from international risk sharing (absent dynamic investment effects).

The aggregate cost of consumption variability is, however, substantial for most developing countries. For a representative sample, Table 4 shows the welfare gain per year from eliminating consumption variability, expressed as a percent of annual consumption. The calculations use the Summers-Heston (1991) data on per capita consumption and assume that the natural logarithm of real per capita consumption follows a random walk with trend. Consumers have generalized isoelastic utility functions with annual time-discount factors of 0.95 (Lucas’s number), relative risk-aversion coefficients of 1, and intertemporal-substitution elasticities of 0.25.7

The numbers in Table 4 are based on a greater reduction in consumption variability than would be feasible in reality. But they suggest that for many developing countries, mechanisms to reduce consumption risk, such as increased access to world financial markets or Shiller’s (1993) proposed market in

perpetual claims to national GDPs, could yield a dramatic payoff.

7For details on the formulas used, see Obstfeld (1994b). The assumptions on time preference, risk aversion, and intertemporal substitutability are conservative; more realistic assumptions would raise the costs in Table 4.

37

Table 4 Gains from the Elimination of Consumption Variability in Selected Developing Countries

yr ereane nce

Country Annual Percent Consumption Gain Botswana 4.56 Kenya 4.27 Morocco 1.54 Tanzania 4.53 Zimbabwe 5.31 Bangladesh 3.04 India 0.93 Malaysia 1.17 Thailand 1.07 Turkey 1.52 Barbados 2.69 Mexico 0.54 Argentina 1.94 Brazil 1.80 Chile 2.75 Venezuela 2.22

Note: The calculations assume that the logarithm of per capita consumption follows a random walk with trend and that individuals have generalized isoelastic utility functions with annual time-discount factor 0.95, relative risk-aversion parameter 1, and intertemporal-substitution elasticity 0.25. Data are taken from Summers and Heston (1991). For details on the calculation, see Obstfeld (1994b).

37a

4 The Allocation of Global Investment

A well-functioning world capital market should direct investment toward its most prcductive global uses. Economic efficiency requires that the expected value of investment in any location be the same. The most direct approach to evaluating efficiency would compare capital’s rate of return in different countries, but it is difficult to find internationally comparable measures of the ex ante return to capital.™ This section therefore focuses on two indirect approaches. One indirect approach argues that capital should flow from countries where it is relatively abundant to countries where it is relatively scarce. A second indirect approach is based on an examination of countries’ saving and investment patterns.

Does Capital Flow to Capital~Poor Countries?

In the simplest one-sector growth models, capital mobility ensures that countries sharing a common technology will converge to identical capitaloutput ratios. Figure 8 shows that, for the two years 1973 and 1987, this equality was not even approximately true among the six OECD countries for which Maddison (1991) has constructed comparable capital-stock data. Moreover, there is little discernible tendency for capital-output ratios to converge between 1973 and 1987. A cross-sectional regression of the change in the log Capital-cutput ratio K/Y on the initial log capital-output ratio yields a

small ancl insignificant slope coefficient:

“strictly speaking, one would wish to examine the after-tax marginal rates of return that capital investments in different countries offer to various domestic and foreign investors. Even the states within a national federation may tax capital at different effective rates.

38

Ratio in 1987

Capital-Output Ratios for Some Industrial Countries 1973 versus 1987

0.8 1 1.2 1.4 1.6 1.8 2 Ratio in 1973

FIGURE 8: Industrial-country capital-output

ratios, 1973 and 1987 (from Maddison 1991)

38a

log (K/Y ) i, - log(K/Y) 9, = 0.16 - 0.07 log(K/Y),., 7 R? =0.01. (0.13) (0.47)

Are such persistent international differences in capital-output ratios prima facie evidence of capital market failure? Suppose aggregate output in a country is piroduced through the (possibly country-specific) Cobb-Douglas

production function of capital K and N other productive factors Ly, N Y= (OK)*[] (@,L,)% (3) j=l

The marginal product of capital in this economy is MPK = a/(K/Y). If two countries’ outputs are given by Cobb-Douglas production functions of form (3), then even when those production functions differ in factor productivities (the ©s) and in the array of noncapital inputs, the countries’ MPK ratio will equal the inverse of their relative capital-output ratio provided only that they share a commen value of a, capital’s share in GDP.

This simple result has strong implications. Figure 8 suggests, for example, that, as of 1987, K/Y was about 1.9 for Japan but under 1.3 for the United States. With a common a = 1/3, the value suggested by Mankiw, Romer, and Weil (1992), the marginal product of capital would have been 17.4 percent in Japan, much below its predicted value of more than 25.3 percent in the United States. Under free capital mobility, investment should have been higher in the United States than in Japan; in reality the reverse was true. If one applies this type of argument to compare returns to capital in developed and developing countries (as do King and Rebelo, 1993, and Lucas, 1990), the discrepancies are even greater.

One major pitfall in the preceding reasoning is the assumption of an

aggregate production function of form (3). If there are multiple production

39

activities with different capital requirements, aggregate capital-output ratios can differ widely between economies that pay the same factor rewards. Furthermore, factors could be more substitutable in some activities (at least in the long run) than the Cobb-Douglas form assumes. For example, capital substitutes for land in some Japanese production activities that are carried out in the United States with more land and less capital. The evidence that a is a universal constant is weak. Expected changes in relative prices will influence expected rates of return. Finally, uncertainty is being ignored. If the productivity coefficients 0 are stochastic and imperfectly correlated across countries, we would not expect to observe the same K/Y ratio everywhere: more capital should be placed in countries where the payoff to investment is most highly correlated with the marginal utility of world consumption.”

Examination of countries’ aggregate capital-output ratios canaot, in itself, be informative about opportunities for efficiency~enhancing international investment flows. A more convincing, albeit painstaking, method is to evaluate sectoral rates of return directly, as in Minhas’s (1363) famous monograph. This work, like Harberger’s (1980) later summary of more aggregative studies, suggests that ex post international differences in the return to capital have been relatively moderate in the recent past. Unfortunately, little up-to-date research along these lines is readily available.

The Feldstein-Horioka Approach

As Section 1 described, Feldstein and Horioka (1980) and Feldstein

“Bardhan (1993) explores several deterministic models in which big international wage discrepancies coexist with small international differences in returns to capital.

40

(1983) propcsed as a barometer of capital mobility the size of the association between economies’ savings rates and their investment rates. They reasoned that, in a world of Capital mobility, each country’s savings are free to flow to their most productive uses anywhere in the world; thus, there is no reason for an increase in national Saving necessarily to augment the source country’s domestic capital stock. These papers use regressions of domestic investment rates on national savings rates to measure the fraction of an exogenous increase in national Savings that will remain at home, the "savings retention coefficient," as Feldstein and Bacchetta (1991) call it. The saving-investment puzzle is to explain why this coefficient appears to be high, even in recent data, despite the high international Capital mobility suggested by the evidence on interest-rate links reviewed in Section 2.

Informed policy decisions may depend on whether the saving-investment puzzle really is explained by low capital mobility, or by factors that simultaneous!y drive both saving and investment. For example, under perfect capital mobility, an increase in the government budget deficit of a small economy need not crowd out domestic investment, even if consumers do not behave accorcling to the Ricardian equivalence proposition; instead, foreign savings are available in perfectly elastic supply to finance additional national borrowing. Feldstein and his collaborators have, by contrast, interpreted their saving-investment regressions as implying that any fall in national saving will, over the long run, cause a commensurate fall in domestic investment, as in a closed economy.

The Feldstein~Horioka approach raises two distinct questions. First, is a close association between Saving and investment in fact evidence of low

international capital mobility, as argued in the initial papers by Feldstein

41

and Horioka? Second, do regressions of investment on saving actually measure the investment effect of an exogenous change in the saving rate, foi: example, one caused by fiscal policy? These two questions are inseparably linked: before the investment effect of a change in national saving can be predicted, the precise mechanism underlying the estimated saving~-investment association must be understood. Because of space limitations, however, this chapter will focus on the first question, the relevance of the statistical savinginvestment relationship for assessing international capital mobility.”

Cross-sectional versus time-series estimation. It is helpful t:o distinguish between two possible econometric approaches to estimating savinginvestment relationships. Feldstein and Horioka (1980) implemented a crosssectional estimation strategy. In this approach, each observation consists of a country j’s average investment and saving rates over a given time period; the estimated regression equation based on a cross-sectional sample of N

countries is (I/Y), = aS + BS(S/Y), + U; + (4)

where (I/Y), is country j’s average nominal investment rate out of nominal GNP or GDP over the chosen time period, (S/Y),; is its average saving rat2 over the same period, and u, is a random disturbance.

A second estimation strategy is based on time-series data. In this approach, each observation consists of a given country’s investment and saving rates over some time period t. The estimated regression equation based on a

time-series sample for a single country is

*Obstfeld (1991) analyzes econometric pitfalls of using saving--investment regressions to predict the effects of exogenous shifts in saving.

42

(I/Y), =a® + B™(S/Y), + u, (5)

(or the corresponding equation in first differences) .”!

In a world of completely immobile capital, the error terms in (4) and (5) represerit measurement error, and both estimation strategies yield estimated slope coefficients near 1. More generally, however, the two estimation strategies could yield quite different slope coefficients, even when all countries are integrated into world capital markets to a similar degree, because BP“ in (4) and B™ in (5) measure very different things.

Suppose, for example, that, in the sample of N countries mean saving rates have a high positive cross-sectional association with mean investment rates, but that, for each country, deviations of saving rates from the timeseries mean are uncorrelated with deviations of investment rates from the time-series mean. Suppose also that the cross-sectional observations are country averages over T periods. Then the ordinary least squares (OLS) estimate B© will be high if T and N are sufficiently large, but B”™ will be near zero for each country. If, instead, mean saving rates and investment rates have a zero cross-sectional correlation, but for each country, deviations from its mean saving and investment rates tend to be close, f° will be near zero for T and N sufficiently large but the estimates f™ will be high.

The cross-sectional estimation strategy attempts to capture the relation between long-run saving and investment rates. For this strategy to succeed,

each country’s saving and investment rates must be averaged over a sufficient

‘Feldstein (1983) reports panel estimates that combine the crosssectional and time-series strategies by assuming that p© and B® are equal.

43

interval to eliminate the influence of short-run fluctuations around long-run means. The time-series estimation strategy is meant to uncover the short-run relation between national saving and domestic investment. Both long-run and short-run relationships are pertinent to an assessment of capital mobility. Explanations of the time-series relation between saving and investment will not, however, throw much light on the cross-sectional relationship unless the time period chosen for cross-sectional estimates is so brief that transitory shocks to saving and investment swamp underlying long-run patterns. Conversely, explanations of true long-run patterns may have little power to explain short-run comovements.

Results of cross-sectional estimation. Feldstein and Horioka (1980) estimated equation (4) for a sample of sixteen OECD countries, averaging annual data for subperiods from 1960 to 1974.” Data on gross saving and investment rates” averaged over the entire 1960-74 period led to a

representative OLS result:

(Z/¥), = 0.035 + 0.887(S/Y), + u,; R?=0.91. (0.018) (0.074)

Feldstein and Bacchetta (1991) provide an update; a typical estimate of B& based on a sample of twenty-three OECD countries over the more recent period

from 1974 to 1986 is 0.868 (with a standard error of 0.145), a result quite

*Their country sample was Australia, Austria, Belgium, Canada, Denmark, Finland, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Sweden, the United Kingdom, and the United States.

“Gross, rather than net, rates are more appropriate for this regression. A regression in net rates imposes the assumption that all replacement investment is financed by domestic savings.

44

close to the original findings.* This regression presents a much starker puzzle about: the international capital market than those based on 1960-74 data because it iis generally believed that the world capital market, although relatively shallow and segmented prior to the early 1970s, has become less regulated and has expanded vigorously since then (Marston, 1993b, gives evidence foi: the 1960s). Notwithstanding this evolution, the Feldstein- Bacchetta f:ndings still imply that a 1 percent increase in the national saving rate remains cross-sectionally associated with a nearly equal increase in the domestic investment rate.

A furt:her update is provided in Table 5, which presents the result of estimating (4) for twenty-two OECD countries for subperiods from 1974 to 1990.* Saving and investment rates are gross nominal flows divided by nominal GDP or GNP.

The point estimates for Bp“ in Table 5 are lower than those that Feldstein and Horioka (1980) report and somewhat lower, on the whole, than those that Feldstein and Bacchetta (1991) report. The R? statistics are also below the ones in Feldstein and Horioka (1980). Figure 9 shows a scatter plot for the 1981-90 data, together with the fitted regression line.

The results are suggestive of a decade-to-decade downward trend in Bo: the estimated coefficient for the 1974-80 period, 0.867, has dropped to 0.636 by 1981-90. Such a trend, even if established, would be difficult to interpret

unambiguously. For example, the 1986-90 estimate of p© is higher than that

“The countries are the sixteen listed by Feldstein and Horioka (1980) plus France, Iceland, Norway, Portugal, Spain, Switzerland, and Turkey.

“The countries are the Feldstein-Bacchetta sample minus Turkey, which can be classified as a developing country. Luxembourg traditionally is omitted from this sample; it is such an extreme outlier that its addition reduces the cross-sectional regression coefficient to insignificance.

45

Table 5

~ Cross-Sectional Regressions of Investment Rates on Saving Rates: Period Average Data, 1974-1990

Period pS R?

1974-90 0.715 0.60 (0.131)

1974-80 0.867 0.56 (0.170)

1981-90 0.636 0.64 (0.108)

1981-85 0.567 0.43 (0.147)

1986-90 0.636 0.69

(0.094)

Note: Estimates of equation (4) in text. Standard errors appear in parentheses below estimates of slope coefficient B©. The sample of twenty-two countries consists of Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom and the United States.

45a

Saving and Investment Rates for the Industrial Countries 1981-90

0.35

0.3

australia

Z 0.25 nz w > 5 § & > & 02

0.15

0.1 0.1 0.15 0.2 0.25 0.3 0.35 Saving/GNP & actual —~ fitted FIGURE 9:

Average saving and investment rates for 22 industrial countries, 1981-90

45b

for the 1981-85 period, yet would not be taken as evidence of a decreasing degree of international capital mobility. Another reason for caution is that the coefficient differences in Table 5 are not statistically significant.

The basic finding is that the positive.cross-sectional association between OECD saving and investment rates is economically and statistically significant, although far from perfect and possibly declining over time. Although the cross-sectional results are less striking than those for the 1960-74 period, they may present more of a puzzle given the current level of industrial-country residents’ participation in international capital markets (Goldstein et al., 1993, document this activity).

Results for a wider sample including developing countries are not reported, because there is less of a Saving-investment puzzle as far as those countries are concerned. Most of those countries even now control capital flows and in some periods have faced binding external credit constraints. Notwithstanding these tangible impediments to capital flow, the cross-~ sectional association of saving and investment rates is often found to be lower for the developing countries than for the OECD countries over the period from 1960 to the early 1980s, when the debt crisis began (Fieleke, 1982; Dooley, Frankel, and Mathieson, 1987;- and Summers, 1988).

Results of time-series estimation. Table 6 examines the time-series properties of annual saving and investment rates from 1974 to 1990 for the twenty-two countries that made up the cross-sectional sample, plus Luxembourg. "Levels" estimates of B™ come from OLS estimation of equation (5), with a time trend included in the regression. "Differences" estimates come fron the

regression

46

Table 6 Time-Series Regressions of Investment Rates on Saving Rates: Annual Data, 1974-1990

_ EEE Tl eee

Country p*® (levels) B™ (differences)

:

Australia 0.792 0.857 Austria 0.825 0.732 Belgium 0.637 0.749 Canada 1.097 0.963 Denmark 0.727 0.657 Finland 1.803 1.172 France 0.909 1.101 Germany 0.327 0.561 Greece 0.845 0.892 Iceland -0.450 -0.654 Ireland -0.037 0.208 Italy 0.214 1.154 Japan 1.161 1.100 Luxembourg -0.135 0.042 Netherlands 0.381 0.457 New Zealand 1.154 0.787 Norway -0.614 -0.515 Portugal 0.736 0.718 Spain 1.104 0.246 Sweden 0.717 0.574 Switzerland 1.221 1.547 United Kingdcem 0.113 1.002 United Statee 0.848 1.090 Note: Estimates of levels are based on the OLS regression (Z/Y), = a® + BRcsyy), + yPe +

uj; estimates of differences are based on the OLS regression A(I/Y), = a®™ + BA(S/Y), + u,.

46a

Table 7

Time-Series Correlation Coefficients between Saving and Investment Rates: Annual Data, 1974-1990

ee etrerenasnneseevesnnnntenmeneneenr sane

Country p™ (levels) 6” (differences) ee Australia 0.834 0.742 Austria 0.746 0.575 Belgium 0.848 0.773 Canada 0.745 0.823 Denmark 0.783 0.662 Finland 0.846 0.682 France 0.851 0.710 Germany 0.401 0.610 Greece 0.836 0.750 Iceland -0.333 -0.333 Ireland -0.031 0.157 Italy 0.150 0.560 Japan 0.837 0.795 Luxembourg -0.247 0.071 Netherlands 0.505 0.518 New Zealand 0.517 0.562 Norway -0.659 -0.474 Portugal 0.591 0.584 Spain 0.711 0.193 Sweden 0.785 0.514 Switzerland 0.784 0.736 United Kingdom 0.092 0.668 United States 0.773 0.895

Note: Fatimates of Doosan Ee

Note: Estimates of levels are simple correlation coefficients between (I/Y), and (S/Y),, where both variables are linearly detrended. Estimates of differences are correlation coefficients between A(I/Y), and A(S/Y),.

46b

A(I/Y), =a™ + BPA(S/Y), + u, .

Table 7 reports the corresponding simple correlation coefficients between linearly detrended and differenced saving and investment rates.

There is a wide dispersion of outcomes, a reflection not only of different degrees of financial openness, but also of different country sizes and the different shocks that have buffeted these diverse economies. For most countries, the saving and investment time series are positively related, and the relationship is typically strong. Australia, New Zealand, and Portugal all show positive time-series saving-investment associations despite having run sizable current-account deficits over parts of the sample period. (Portugal’s 1982 deficit was 13.5 percent of GDP.) Norway, which also ran a deficit, shows a strongly negative relationship. These findings underscore the point that annual time-series correlations contain little information about the relation between saving and investment over long periods.*

Even under perfect capital mobility, positive regression coefficients such as those reported in Table 7 are not hard to explain. If labor is internationally immobile, for example, positive shocks to investment productivity can cause both investment and saving to rise (Obstfeld 1986; Finn 1990; Tesar, 1991; Ghosh, 1994). If the usual outcome of such a shock is a current-account deficit, and, if productivity shocks are the dominant form of disturbance, then it would not be surprising to find an estimate of B™ above 1, a result found for several countries in Table 6 but difficult to explain if

capital is internationally immobile. A positive time-series correlation

*Observe that the choice between levels and differences can matter, at least in this finite sample (for example, for the United Kingdom).

47

between saving and investment is reinforced if global as well as local shocks to investment and Saving are important (as found by Glick and Rogoff, 1993) .””

Unlike the time series results, which can be rationalized in several plausible ways, the cross-sectional finding that countries with hicher longterm saving rates also have higher long-term investment rates is mcre difficult to explain in a world of capital mobility. The balance of this section therefore focuses on alternative interpretations of the cross-— sectional saving-investment pattern as it persisted through the 1980s. Explanations for the Cross-Sectional Saving-Investment Relationship Many researchers have taken the high estimates of B® in (4) as evidence that national savings for the most part are still retained at home and are not channeled toward their most efficient global uses by the world capital market. Others have tried to approach the saving-investment puzzle by identifying economic forces that underlie both Saving and investment and cause Long-term averages of these two variables to move together. A wide variety of contributory mechanisms has been proposed.

Demographic factors. Characteristics of a nation’s labor force can simultaneously affect national saving and the profitability of domesitic investment. Labor-force growth provides one example: higher growth can raise national saving by increasing the ratio of young savers to old dissavers. At the same time, higher growth raises the investment needed to keep the labor force equipped with Capital (Black 1982; Obstfeld 1986). Higher productivity

growth concentrated among prime-age workers would likewise raise trend saving

aan

"Baxter and Crucini (1993a,b), Cardia (1992), Mendoza (1991a, 1991b), ana Stockman and Tesar (1990) explore simulation models in which free international asset trade igs consistent with high time-series correlations between saving and investment.

48

as well as trend investment.

Summers (1988) and Feldstein and Bacchetta (1991) dismiss the hypothesis that growth, either in the labor force or in factor productivity, is the primary factor generating the cross-sectional saving-investment relationship. They show that the addition of growth variables to the cross-sectional regression does not reduce the apparent influence of saving on investment. Notwithst:anding these regressions, it remains quite plausible that labor-force developments are a part of the story, more important in some countries than in others. “Yesar (1991) presents evidence along these lines, showing that the fraction of the population between ages 15 and 64 is positively related to both sav:ng and investment rates. In a more recent contribution, Taylor (1993) uses the Summers-Heston data to estimate versions of the Feldstein-Horioka vegression that control for measures of domestic relative prices, the age structure of the population, and the interaction of the age structure with the growth of domestic output. He finds that in a number of country samples the cross-sectional saving-investment association disappears. The role of growth clearly cdleserves further detailed study.

Other potential links between household intertemporal allocation decisions and investment remain to be investigated. For example, are there systemati.c links among fertility rates, Saving, expenditures on schooling, and the profitability of domestic investment?

Real interest rates. Even if capital is perfectly mobile and uncovered interest parity holds true, national real interest rates need not be equal. Frankel (1986, 1993) claims that this point resolves the Feldstein-Horioka puzzle. The apparent puzzle arises, he argues, because increases in national

saving depress the local real interest rate, spurring investment and inducing

49

a statistical correlation between saving and investment rates.

Although this mechanism may help us understand time-series correlations between saving and investment rates, its bearing on the longer-run =zrosssectional patterns is less obvious. Under perfect capital mobility and uncovered interest parity, the real-interest differential between two countries equals the expected percentage change in their currencies’ real exchange rate. If real-interest effects are to explain the cross-sectional regression results, countries with high saving and investment rates must have low real interest rates and so their currencies must be continually appreciating in real terms against foreign currencies.

Cardia (1992) describes a simulation model that is based on Frankel’s suggested mechanism but that nonetheless may have some explanatory power for the cross-sectional Feldstein-Horioka pattern. In her model, adjustment to Shocks can be drawn out over decades because of capital-installation costs and an overlapping-generations population structure. Although Cardia does not report cross-sectional simulations, the long-lived effects of the disturbances she considers probably would contribute to a strong cross-sectional association between long Saving- and investment-rate averages.

As Balassa’s (1964). work implies, models with different sectoral productivity growth rates can exhibit permanently trending real exchange rates. This suggests another potential mechanism causing high-saving, highinvestment countries also to be countries with low real interest rates.

Imagine a small Open economy producing traded and nontraded goods using Capital, which is internationally mobile, and labor, which is not. Assume that initially all countries are identical, with growing labor forces. Consider the

effect of a permanent increase in traded-goods productivity growth in one

50

economy.

The currency of this economy will begin to appreciate in real terms, its real interest rate will fall, and its investment rate will rise. Saving, which depends on the real interest rate, also may change. If the average domestic intertemporal substitution elasticity is below 1, as several empirical studies suggest, the fall in the real interest rate can cause saving to rise. Saving and investment may therefore show a positive cross-sectional correlation, seemingly driven by cross-country real-interest-rate differences but really driven by clifferences in traded-goods productivity growth.*

No one has yet established a robust cross-sectional relationship among Saving, investment, the real interest rate, and the real exchange rate’s expected path. Mechanisms such as the one described thus remain speculative.

Hysteresis of factor supplies. Results presented above (Figure 8) show that OECD countries are characterized by wide and persistent differences in capital-out:put ratios. This pattern suggests another possible explanation for the saving--investment puzzle.

European countries entered the postwar era burdened by external payments controls and limited access to foreign resources. For some time, therefore, countries had to finance most of their capital accumulation through domestic savings. High-saving countries accumulated large capital stocks and specialized in capital-intensive industries, and low-saving countries produced

a more labor-intensive product mix.

%In general, when an economy has several sectors of differing capital intensity, some of which produce nontraded goods, there is no longer a presumption that the economy’s consumption side and its production side (including investment) can be analyzed separately, even under capital mobility. This point is made through various examples by Murphy (1986), Engel and Kletzei: (1989), and Wong (1990).

51

The substantial liberalization of capital movements starting in the 1970s need not have disturbed this production pattern greatly. In the presence of labor-force growth, however, high-capital countries required high investment rates to maintain their established industries, whereas low-capital countries could get by with lower investment rates. Because the high~-capital countries were also those with high saving rates, a high cross-sectional correspondence between saving and investment rates was the result. On this view, the historical accident of Capital immobility during the first art of the postwar period had an effect on the distribution of national investment rates that persisted even after capital mobility returned.

If the preceding interpretation is valid, countries with higher saving and investment rates should have higher shares of capital income in GDP. Mankiw, Romer, and Weil (1992) argue, however, that this is not the case and that, in fact, there is little international variation in capital’s GDP share.” Their argument, based on limited data from the 1960s and 1970s, contradicts Kaldor’s (1961, p. 178) fifth "stylized fact" of economic growth of "a high correlation between the share of profits in income and the share of investment in output." More research on this point would be useful.

Corporate financing frictions. The need for firms facing imperfect domestic capital markets to finance investment out of corporate savincs has been suggested as another explanation of the Feldstein-Horioka puzzle. But is a tight link between corporate Saving and investment enough to produce: a tight link between national saving and investment? A dollar rise in corporate saving

may raise domestic investment if firms are borrowing-constrained, but it will eee

*This pattern would be consistent with a world in which national outputs are produced according to equation (3), with a the same in all countries, and capital is internationally immobile.

52

raise national saving only if shareholders fail to pierce the corporate veil and adjust their own total saving downward by a dollar. The largest corporations, moreover, probably do not face binding finance constraints. The general hypothesis is that strict domestic segmentation of financial markets might generate a country-by-country saving-investment association. Empirical documentation for this mechanism has not yet been produced.

A related hypothesis concerns the possibility that domestic and foreign residents value domestic equities differently, as might (but need not) be the case in the absence of efficient consumption risk sharing among countries (Dooley, F:rankel, and Mathieson, 1987, examine a polar case in which claims to domestic physical capital are nontradable). In this situation, domestic saving and investinent could be positively correlated, even for a small country, despite pe::fect international arbitrage in bonds. A strong positive correlation is no necessity, however, because there remains the possibility in principle of substantial bond-intermediated foreign financing of investment. Equity-market segmentation along national lines underlies the international diversification puzzle; but can the phenomenon help explain the crosssectional saving-investment relationship? Different plausible models yield different answers. One obvious empirical approach would be to look for a negative ciross-sectional correlation between the cost of capital and the saving rate in industrial countries.”

Government policies. Systematic current-account targeting by governments

would, if successful, tend to produce a strong cross-sectional association of

“There is some limited evidence of such a relationship in the past; see McCauley and Zimmer (1989). However, it is hard to disentangle the effect of saving from the effect of tax provisions that simultaneously affect saving and the cost of capital. Obviously, such tax effects could be another influence on the cross-sectional pattern of saving and investment rates.

53

saving and investment even with high capital mobility (Fieleke 1982; summers 1988). Fiscal and monetary policy, as well as capital controls, have all been used to limit the sizes of current-~account imbalances. There is some evidence that government policies in a number of countries have aimed to curtail external imbalances (Artis and Bayoumi, 1989), but it is difficult to judge how well these policies succeeded. It is also possible that government: policies aimed at domestic stabilization or international reserve management have effects similar to current~-account targeting.

The economy’s intertemporal budget constraint. An open economy faces an intertemporal budget constraint relating the difference between its saving and investment, the current account, to the change in its net external assets. Under some economic conditions this constraint alone implies that saving and investment ratios averaged over sufficiently long periods must be close despite capital mobility (Obstfeld, 1986; Sinn, 1992; Vikoren, 1991).

To appreciate this point, let A, denote a given country’s nominal net foreign assets at the end of period t and recall the current-account identity’s implication that A, - A,, = S, - JI,.“' Suppose that the data are average saving and investment rates over T periods. Let a, = A,/Y, be the ratio of external assets to income and g, = (Y, - Y.1)/Y,, the growth rate of

nominal income. Then the current~account identity implies that the difference

“This relation will not hold exactly in the data because Saving as measured by national income and product accounts does not include capital gains or losses on foreign assets (Obstfeld, 1986).

54

between the averaged saving and investment rates is”

15 S,- I, _ 1 A, - A, TS Y, To Y, 1 Yr. Yo 6 = ger + (2 es] rites [yz a (6) 1 = lar 7~ a) + Gray, + + + - + Goa, + 9140) °

In principle, the foregoing identity alone places no constraints on the average difference between saving and investment rates. Suppose, however, that there is a steady-state ratio of net foreign assets to income from which the economy dces not greatly diverge between the start and end of the sample period. Then, if nominal income growth is moderate, equation (6) implies that the averaged difference between saving and investment rates may well be small.

Mature economies may have attained a stationary distribution of the foreign-assets-to-GNP ratio; the intertemporal trade gains that arise between mature eccnomies will generally be transitory and their distribution symmetrical.“ This conjecture may help explain why, even in the late 1980s, a fairly high cross-sectional saving-investment relation persisted for the industrial countries. The conjecture also explains why, before the debt crisis of the 1980s, developing countries displayed lower cross-sectional saving-

investment correlations than did the industrial countries. Developing

“The income growth rates below are nominal rather than real rates because the natior.al~income and product-account concept of saving does not correct income for the inflationary erosion of the real values of nominal assets.

“An exception is Norway, which borrowed abroad so heavily during the 1970s to cevelop its oil production that, by 1978, its foreign-debt-to-GDP ratio stocd near 60 percent (Vikgren, 1991). Norway repaid this debt quickly. By 1985, the country’s net foreign debt stood at around 12 percent of GDP, its 1970 level. The U.S. current-account deficit, driven by government deficits and demographic shifts, is another exception.

55

countries with significant unexploited investment opportunities have external debts well below their steady-state levels. This perspective suggests that, ultimately, the cross-sectional saving-investment correlation within a group of countries with open capital markets depends on the extent of each ore’s long-term intertemporal trade gains with other countries. Attempts to éssess these gains (as in Ghosh, 1994, and Glick and Rogoff, 1993) are critical for understanding how puzzling the saving-investment puzzle really is. Comparisons with the Gold Standard and with Regional Data

An indirect way to judge whether the Feldstein-Horioka puzzle reflects true capital immobility or some subset of the alternative factors listed above is to examine the strength of the cross-sectional saving-investment association in settings of presumed high capital mobility. Data from the gold-stanclard period and regional data have both been used for this purpose.

The saving-investment relation under the gold standard. Table 8 reports results for three data samples. The first consists of Australia, Canada, Denmark, France, Germany, Italy, Norway, Sweden, the United Kingdom, arid the United States with data averaged over the period from 1880 to 1913. The second sample adds Japan, using data averaged over 1885 to 1899 and 1900 to 1913. The third sample, based on 1926-38 data, subtracts France but adds Finland, which gained independence from Russia in 1917. I first discuss the pre-1914 results, which fall under the classical gold standard (Jones and Obstfeld, 1994, give details on data construction).

For 1880 to 1913, the estimated regression coefficient f° is almost Significant (with a one-tailed test) and not very different from the estimates in Table 5 based on data from the 1980s (the R* is, however, much lower in

Table 8). For 1885 to 1899, the estimate B° is about the same but is

56

Table 8 Cress-Sectional Regressions of Investment Rates on Saving Rates During the Gold Standard and Interwar Period: Period Average Data

Period pS R

C.576 0.27 (0.335)

L8BS-99 0.568 0.41 (0.228)

1900-13 0.774 0.26 (6.436)

ES 3a u.959 6.94

ates of equation (4) in text. Standard errors appear in parentheses below oe 5° gieope coefficient 8°. The 1880-1913 sample consists of Australia, Canada, vance, Germany, Italy, Norway, Sweden, the United Kingdom, and the United eampiles for 1885 tc 1899 and 1500 to 1913 add Japan. The sample for 1926 to “tracts France and adds Finland.

56a

significant. For 1900 to 1913 (with data pictured in Figure 10), the coefficient rises to 0.77 but loses significance.“

To the extent that the classical gold standard was a period of high international financial integration, the pre-1914 findings in Table 8 and Figure 10 suggest that the recent long-run behavior of saving and invest:ment rates is not inconsistent with substantial capital mobility.

True, the dispersion of saving and investment rates during the god standard is greater than among industrial countries over the 1980s; and among the largest economies we now see nothing like the surpluses the U.K. persistently ran. Three factors should be considered, however, in assessing capital mobility under the classical gold standard and comparing it with current conditions. First, as Nurkse (1954) emphasized, international capital movements were abetted by complementary large-scale labor movements fron Europe into regions of recent (white) settlement. Pre-1914 levels of international migration have not been approached in the recent postwar era. Second, the inclusion of Australia and Canada means that developing- and industrial-country data are being pooled, a procedure that would loosen the saving-investment association in modern data. Finally, Britain’s close cultural and political ties with some borrowers certainly facilitated its large-scale foreign lending. As is evident from Figure 10, Canada and the

United Kingdom are behind the poor fit of the regression for 1900 to 197.3.

“Bayoumi (1990) finds no cross-sectional saving-investment association for a smaller eight-country sample over any subperiod of 1880 to 1913. Eichengreen (1990) amends Bayoumi’s data and adds the United States. The results in Table 8 are very similar to Eichengreen’s, despite my use of different data for some countries and an expanded set of countries.

“See Razin and Sadka (1993) for a recent discussion of international labor mobility.

57

FIGURE 10: Average saving and investment rates for 11 countries under the classical gold standard, 1900-13

Saving and Investment Averages 1900-1913 (as percent GNP /GDP)

0.08 0.08 0.1 0.120.140.160.18 0.2 0.220.240.26

s/Y

57a

Table 8 also reports a regression for the interwar pericd following the (short-lived) reinstatement of the international gold standard, 1926 to 1938; the data are displayed in Figure 11. The results stand in the sharpest possible contrast to those for the classical gold standard and show a stronger saving-investment association even than the Feldstein-Horioka 1960-74 results. Eichengreen (1990) discusses possible reasons for this contrast, which are complex but seem related to a genuine post-World War I decline in capital mobility. One factor behind this decline was the rise of the political Left. This development made international investors less secure in their property rights than they were before 1914. It also focused the attention of policymakers on domestic economic problems at the expense of laissez-faire principles of international economic relations.

Governments practiced less pervasive management of their econories during the classical gold-standard era than they did later. Do the results discussed here therefore show that the hypothesis of current-account targeting is not needed to explain the current cross-sectional saving-investmert relation? Not at all. Even under the gold standard, some governments may have curtailed current-account imbalances as a side effect of actions taken to maintain gold convertibility, or in pursuit of foreign-policy aims.

Regional saving-investment links. The use of regional saving ard investment data is a potentially fruitful way to throw light on the siavinginvestment puzzle.“ Bayoumi and Rose (1993) construct saving and investment

data for eleven British regions for 1971 to 1985; they find no significant

“Murphy (1984) applied an analogous idea to the 143 largest industrial corporations from the 1981 Fortune 500. He found a significant cross-sectional relation between corporate saving and investment. It would be interesting to know if this relationship has held up in view of financial-market developments since the early 1980s.

58

FIGURE 11: Average saving and investment rates for 11 countries in the interwar period, 1926-38

Saving and Investment Averages 1926-1938 (as percent GNP/GDP) 0.22

I/Y

0.08

0.06

0.04 0.040.060.08 0.1 0.120.140.160.18 0.2 0.22

s/Y

58a

positive cross-sectional relation between saving and investment rates. Bayoumi and Sterne (1993) find a similar result for Canadian provinces. Sinn (1992), who looks at both 1953 and 1957 data for the forty-eight U.S. continental states and Alaska, finds a negative cross-sectional relation between saving and investment rates. Data for 1975 to 1988 on average saving and investment rates for the forty-five Japanese prefectures listed in Table 3 are graphed in Figure 12. Again, no positive relationship is apparent.

The data used in these calculations aren’t always ideal. For exemple, Bayoumi and Rose have data for only part of regional expenditure and investment. More seriously, Bayoumi and his coauthors define saving agi regional GDP less a regional consumption measure, not as GNP less that. measure. Thus, these measures of saving fail to include in income not only net interest and dividend payments from outside the region, but also net transfers from the domestic central government and others. The much greater dispersion of saving as compared to investment rates in Figure 12 raises suspicions that measurement errors are a problem in the Japanese saving data shown there, despite their definition as prefecture GNP less consumption.

There are, moreover, differences between regions and countries i:hat might weaken the saving-investment link. The comparative ease with which labor can migrate between regions could alter the response of regional saving and investment to disturbances. (This is especially possible in Japan, where commuting between prefectures is significant.) Furthermore, regions within countries tend to be more specialized in their production activities than are countries themselves. Thus, some of the shocks that can make national saving and domestic investment move together may not induce similar comovements in

regional saving and investment.

59

Gross investment rate

FIGURE 12: Average saving and investment rates for 45 Japanese prefectures, 1975-88

1 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 Gross saving rate

59a

The strength of factors such as these is unknown at present. Unt:.1 more work is done and better data assembled, the regional saving-investment regressions provide the most persuasive evidence that national boundaries or macroeconomic policies contributed to limiting industrial-country currentaccount imbalances through the 1980s.

Because regional current accounts are not objects of government policy, the regional results leave current-account targeting as one of the prime suspects generating the positive cross-sectional saving-investment relationship that has persisted in international data. The results are also consistent with the view that capital is still not as mobile between, as

within, countries.

5 Conclusion

The main conundrum in thinking about international capital mobility is to reconcile measures of mobility that superficially contradict each other. How can one square the generally smooth international interest-rate arbitrage documented in Section 2 with the low international consumption correlations and home portfolio bias discussed in Section 3 or the still-sizable crcsssectional coherence between saving and investment documented in Section 4? In this chapter, I have reviewed a number of economic models and data limitations that potentially can contribute to a reconciliation. Despite years of research, however, economists still have not reached the semblance of a consensus on which factors are most relevant. Much work remains to be done; one can hope that the rapid evolution of world capital markets, if not braked

by renewed regulation, will furnish more clues as well as data.

After this lengthy and arduous trek through the literature, I owe the

60

reader more, however, than just a plea for more of the same. So, here are my tentative conclusions.

How mobile is capital in the world economy? As far as industrial countries are concerned, capital mobility appears substantial when judged by past experience, such as that of the gold-standard era. Although the experience of the developing countries is diverse and the market access of many of them is currently in flux, it is clear that much of the developing world still stands outside the nexus of industrial-country financial markets.

Capital mobility appears noticeably lower between industrial economies than within them, although inter~economy capital mobility certainly has increased over time. The threat of government intervention in cross-border capital movements has not disappeared. Indeed, in the wake of the August 1993 ERM collapse, European Commission President Jacques Delors signaled support for concerted EC measures to limit capital mobility ("Return of Capital Controls Raised by Delors," Financial Times, September 16, 1993). Financial flows apparently are less extensive between than within countries. International portfolio diversification appears inexplicably limited for some major countries. And long-run saving and investment rates remain positively associated in international cross sections to an extent greater than is true in the (usually imperfect) regional data that are available. This last phenomenon could reflect central-government policies that have the effect of limiting national current-account imbalances.

It is doubtful that capital will ever be fully as mobile between nations as it can be within them. The mere existence of national governments sovereign within their borders means that no investor can think about domestic and

foreign assets in quite the same way. What is at issue, then, is the extent to

61

which actual conditions approximate free capital mobility. Among industrial countries, the approximation has become better and better in recent years, but

clearly scope for greater financial integration remains.

62

References

Altonji, Joseph G., Fumio Hayashi, and Laurence J. Kotlikoff, "Is the Extended Fami:.y Altruistically Linked? Direct Tests Using Micro Data," American Economic Review 82 (December 1992), pp. 1177-1198.

Artis, Michael J., and Tamim A. Bayoumi, "Saving, Investment, Financial Integration, and the Balance of Payments," International Monetary Fund Working Paper No. 89/102, Washington, D.C., International Monetary Fund, December 1989.

Atkeson, Andrew, and Tamim Bayoumi, "Do Private Capital Markets Insure Regional Risk? Evidence from the United States and Europe," University of Chicago and International Monetary Fund, June 1992, processed.

Backus, David K., Patrick J. Kehoe, and Finn E. Kydland, "International Real Business Cycles," Journal of Political Economy 100 (August 1992), pp. 745-775.

Balassa, Bela, "The Purchasing-power Parity Doctrine: A Reappraisal," Journal of Political Economy 72 (December 1964), pp. 584-596.

Bardhan, Pranab, "Disparity in Wages but Not in Returns to Capital between Rich and Poor Countries," Working Paper C93-017, Center for International and Development Economics Research, Berkeley, Calif., University of California at Berkeley, July 1993.

Baxter, Marianne, and Mario J. Crucini, "Business Cycles and the Asset Structure of Foreign Trade," University of Rochester and Ohio State University, June 1993a, processed.

------ , “Explaining Saving/Investment Correlations," American Economic Review 83 (June 1993b), pp. 416-436.

Baxter, Marianne, and Urban J. Jermann, "The International Diversification

63

Puzzle is Worse Than You Think," Working Paper No. 350, Rochester, N.Y., Rochester Center for Economic Research, May 1993.

Bayoumi, Tamim A., "Saving-Investment Correlations," International Monetary Fund Staff Papers 37 (June 1990), pp. 360-387.

Bayoumi, Tamim A., and Andrew K. Rose, "Domestic Saving and Intra-National Capital Flows," European Economic Review 37 (August 1993), pp. 1197- 1202.

Bayoumi, Tamim A., and Gabriel Sterne, "Regional Trading Blocs, Mobile Capital, and Exchange Rate Coordination," London, Bank of England, January 1993, processed.

Black, Stanley W., "Discussion," in Saving and Government Policy, Conference Series 25, Boston, Federal Reserve Bank of Boston, 1982, pp. 153-161.

Brainard, William C., and James Tobin, "On the Internationalization of Portfolios," Oxford Economic Papers 44 (April 1992), pp. 533-565.

Calvo, Guillermo A., Leonardo Leiderman, and Carmen M. Reinhart, "Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors," International Monetary Fund Staff Papers 40 (March 1993), pp. 108-151.

Campbell, John Y., and N. Gregory Mankiw, "The Response of Consumption to Income: A Cross-Country Investigation," European Economic Review 35 (May 1991), pp. 723-767.

Canova, Fabio, and Morten Overgaard Ravn, "International Consumption Risk Sharing," San Domenico di Fiesole, Italy, European University Institute, March 1993, processed.

Cardia, Emanuela, "Crowding Out in Open Economies: Results from a Simulation

Study," Canadian Journal of Economics 25 (August 1992), pp. 708-728.

64

Carroll, Christopher D., and Lawrence H. Summers, "Consumption Growth Parallels Income Growth: Some New Evidence," in B. Douglas Bernheim and John B. Shoven, eds., National Saving and Economic Performance, Chicago, University of Chicago Press, 1991, pp. 305-343.

Cochrane, John H., "A Simple Test of Consumption Insurance," Journal of Political Economy 99 (October 1991), pp. 957-76.

Cole, Harold L., "Financial Structure and International Trade," International Economic Review 29 (May 1988), pp. 237-259.

Cole, Harold L., and Maurice Obstfeld, "Commodity Trade and International Risk Sharing? How Much do Financial Markets Matter?" Journal of Monetary Economics 28 (August 1991), pp. 3-24.

Committee of Governors of the Central Banks of the Member States of the European Economic Community, Annual Report 1992, Basle, April 1993.

Crucini, Mario J., "International Risk Sharing: A Simple Comparative Test," Columbus, Oh., Ohio State University, August 1992, processed.

Deaton, Angus, Understanding Consumption, Oxford, Clarendon Press, 1992.

Dekle, Robert, "Saving-Investment Correlations and Capital Mobility: On the Evidence from Japanese Regional Data," Boston, Mass., Boston University, October 1993, processed.

Deutsche Bundesbank, "The Trend in Germany’s External Assets and Investment Income," Monthly Report of the Deutsche Bundesbank 45 (January 1993), pp. 43-66.

Devereux, Michael B., Allan W. Gregory, and Gregor W. Smith, "Realistic Cross- Country Consumption Correlations in a Two-Country, Equilibrium, Business-Cycle Model," Journal of International Money and Finance 11

(February 1992), pp. 3-16.

65

Dooley, Michael P., Jeffrey A. Frankel, and Donald J. Mathieson, "International Capital Mobility: What Do Saving-Investment Correlations Tell Us?," International Monetary Fund Staff Papers 34 (September 1987), pp. 503-530.

Dumas, Bernard, "Partial~Equilibrium vs General-Equilibrium Models of International Capital Market Equilibrium," in Frederick van der Ploeg, ed., Handbook of International Macroeconomics, Oxford, Blackwell, 1994.

Eichengreen, Barry, "Trends and Cycles in Foreign Lending," in Horst Siebert, ed., Capital Flows in the World Economy, Tubingen, Mohr, 1990, pp. 3-28.

Eldor, Rafael, David Pines, and Abba Schwartz, "Home Asset Preference and Productivity Shocks," Journal of International Economics 25 (August 1988), pp. 165-176.

Engel, Charles, and Kenneth Kletzer, "Saving and Investment in an Open Economy with Non-Traded Goods," International Economic Review 30 (November 1989), pp. 735-752.

Feeney, JoAnne, "International Financial Markets and Learning-by-Doing in a Small Economy," Boulder, Col., University of Colorado at Bouldei:, July 1993, processed.

Feeney, JoAnne, and Ronald W. Jones, "Risk Aversion and International Markets: Does Asset Trade Smooth Real Income?" Review of International Economics 2 (February 1994), pp. 13-26.

Feldstein, Martin, "Domestic Saving and International Capital Movement:s in the Long Run and the Short Run," European Economic Review 21 (March/April 1983), pp. 129-51.

Feldstein, Martin, and Philippe Bacchetta, "National Saving and International

Investment," in B. Douglas Bernheim and John B. Shoven, eds., Netional

66

Saving and Economic Performance, Chicago, University of Chicago Press, 19911.

Feldstein, Martin, and Charles Horioka, "Domestic Saving and International Capital Flows," Economic Journal (London) 90 (June 1980), pp. 314-329.

Fieleke, Norman S., "National Saving and International Investment," in Saving and Government Policy, Conference Series 25, Boston, Federal Reserve Bank of Boston, 1982, pp. 138-157.,

Finn, Mary G., "On Savings and Investment Dynamics in a Small Open Economy," Journal of International Economics 29 (August 1990), pp. 1-21.

Frankel, Jeffrey A., “International Capital Mobility and Crowding-Out in the U.S. Economy: Imperfect Integration of Financial Markets or of Goods Markets?" in Rik W. Hafer, ed., How Open is the U.S. Economy?, Lexington, Mass., Heath, 1986.

Sleaietetaae , “Quantifying International Capital Mobility in the 1980s," in On Exchange Rates, Cambridge, Mass., MIT Press, 1993, pp. 41-69.

French, Kenneth R., and James M. Poterba, "Japanese and U.S. Cross-border Common Stock Investments," Journal of the Japanese and International Economies 4 (December 1990), pp. 476-493.

------ , "Investor Diversification and International Equity Markets," American Economic Review 81 (May 1991), pp. 222-226.

Gehrig, Thomas, "An Information Based Explanation of the Domestic Bias in International Equity Investment," Scandinavian Journal of Economics 95 (March 1993), pp. 97-109.

Gertler, Mark, and Kenneth Rogoff, "North-South Lending and Endogenous Domestic Capital-Market Inefficiencies," Journal of Monetary Economics

26 (October 1990), pp. 245-266.

67

Ghosh, Atish R., "Capital Mobility amongst the Major Industrial Countries: Too Little or Too Much?" Economic Journal 104 (1994), in press.

Giavazzi, Francesco, and Marco Pagano, "Capital Controls and the European Monetary System," in Capital Controls and Foreign Exchange Legislation, Milan, Euromobiliare, June 1985, pp. 19-38.

Glick, Reuven, and Michael Hutchison, "Financial Liberalization in the Pacific Basin: Implications for Real Interest Rate Linkages," Journal of the Japanese and International Economies 4 (March 1990), pp. 36-48.,

Glick, Reuven, and Kenneth Rogoff, "Global versus Country-Specific Productivity Shocks and the Current Account," International Finance Discussion Papers 443, Washington, D.C., Board of Governors of the Federal Reserve System, April 1993.

Goldstein, Morris, David Folkerts-Landau, Peter Garber, Liliana Rojas-Suarez, and Michael Spencer, International Capital Markets, Part I. Exchange Rate Management and International Capital Flows, Washington D.c., International Monetary Fund, April 1993.

Golub, Stephen S., "International Diversification of Social and Private Risk: The U.S. and Japan," Swarthmore, Penna., Swarthmore College, November 1991, processed.

Gordon, Roger H., and Hal R. Varian, "Taxation of Asset Income in the Presence of a World Securities Market," Journal of International Economics 26 (May 1989), pp. 205-226.

Grossman, Sanford J., and Robert J. Shiller, "The Determinants of the Variability of Stock Market Prices," American Economic Review 71 (May 1981), pp. 222-227.

Harberger, Arnold C., "Vignettes on the World Capital Market," American

68

Economic Review 70 (May 1980), pp. 331-337.

Hodrick, Robert J., and Edward C. Prescott, "Post-War U.S. Business Cycles: An Empirical Investigation," Pittsburgh, Penna., Carnegie-Mellon University, November 1980, processed.

Jones, Matthew T., and Maurice Obstfeld, "Saving and Investment under the Gold Standard," University of California at Berkeley, 1994, processed. Kaldor, Nicholas, "Capital Accumulation and Economic Growth," in Friedrich A. Lutz and Douglas C. Hague, eds., The Theory of Capital, New York, St.

Martin’s, 1961, pp. 177-222.,

King, Robert G., and Sergio T. Rebelo, "Transitional Dynamics and Economic Growth in the Neoclassical Model," American Economic Review 83 (September 1993), pp. 908-931.

Kollmann, Robert, "Consumptions, Real Exchange Rates and the Structure of International Asset Markets," Université de Montréal, March 1992, processed.

-—--—— , "Fiscal Policy, Technology Shocks and the US Trade Balance Deficit," Université de Montréal, April 1993, processed.

Lewis, Karen K., "What Can Explain the Apparent Lack of International Consumption Risk Sharing?," Philadelphia, University of Pennsylvania, July 1993, processed.

Lucas, Robert E., Jr., Models of Business Cycles, Oxford, Blackwell, 1987.

------ , "Why Doesn’t Capital Flow from Rich to Poor Countries?" American Economic Review 80 (May 1990), pp. 92-96.

cleat , "On Efficiency and Distribution," Economic Journal 102 (March 1992), pp. 233-247.

McCauley, Robert N., and Steven A. Zimmer, "Explaining International

69

Differences in the Cost of Capital," Federal Reserve Bank of New York Quarterly Review 14 (Summer 1989), pp. 7-28.

Mace, Barbara J., "Full Insurance in the Presence of Aggregate Uncertainty," Journal of Political Economy 99 (October 1991), pp. 928-956.

Maddison, Angus, Dynamic Forces in Capitalist Development, Oxford, Oxford University Press, 1991.

Mankiw, N. Gregory, David Romer, and David N. Weil, "A Contribution to the Empirics of Economic Growth," Quarterly Journal of Economics 107 (May 1992), pp. 407-437

Mankiw, N. Gregory, and Stephen P. Zeldes, "The Consumption of Stockholders and Nonstockholders," Journal of Financial Economics 29 (1991), pp. 97- 112

Marston, Richard C., "Determinants of Short-Term Real Interest Differentials between Japan and the United States," Bank of Japan Monetary and Economic Studies 11 (July 1993a), pp. 33-61.

------ , "Interest Differentials under Bretton Woods and the Post-Bretton Woods Float: The Effects of Capital Controls and Exchange Risk," in Michael D. Bordo and Barry Eichengreen, eds., A Retrospective on the Bretton Woods System, Chicago, University of Chicago Press, 1993b, pp. 515-538.

Mathieson, Donald J., and Liliana Rojas-Suarez, Liberalization of the Capital Account: Experiences and Issues, Occasional Paper 103, Washington, D.C., International Monetary Fund, March 1993.

Mehra, Rajnish, and Edward Cc. Prescott, "The Equity Premium: A Puzzle," Journal of Monetary Economics 15 (March 1985), pp. 145-161.

Mendoza, Enrique G., "Capital Controls and the Gains from Trade in a Business

Cycle Model of a Small Open Economy," International Monetary Fund Staff

70

Papers 38 (September 1991la), pp. 480-505 ------ , "Real Business Cycles in a Small Open Economy," American Economic Review 81 (September 1991b), pp. 797-818.

Minhas, Bagicha Singh, An International Comparison of Factor Costs and Factor

Use, Amsterdam, North-Holland, 1963 Murphy, Robert G., "Capital Mobility and the Relationship between Saving and

Investment Rates in OECD Countries," Journal of International Money and

Finance 3 (December 1984), pp. 327-342.

------ , "Productivity Shocks, Non-Traded Goods and Optimal Capital

Accumulation," European Economic Review 30 (1986), pp. 1081-1095. Nurkse, Ragnar, "International Investment Today in the Light of Nineteenth-

Century Experience," Economic Journal 64 (December 1954), pp. 134-150 Obstfelcd, Maurice, "Capital Mobility in the World Economy: Theory and

. Measurement," Carnegie-Rochester Conference Series on Public Policy 24

(Spring 1986), pp. 55-103 ------ , "How Integrated Are World Capital Markets? Some New Tests," in

Guillermo A. Calvo, Ronald Findlay, Pentti Kouri, and Jorge Braga de

Macedo, eds., Debt, Stabilization and Development: Essays in Memory of

Carlos Diaz-Alejandro, Oxford, Blackwell, 1989, pp. 134-155.

o-oo , "Comment," in B. Douglas Bernheim and John B. Shoven, eds., National Saving and Economic Performance, Chicago, University of Chicago Press, 1991, pp. 261-270

"International Risk Sharing and Capital Mobility: Another Look,"

Journal of International Money and Finance 11 (February 1992), pp. 115-

121

, "Are Industrial-Country Consumption Risks Globally Diversified?," in

71

Leonardo Leiderman and Assaf Razin, eds., Capital Mobility: The Impact on Consumption, Investment, and Growth, Cambridge, Cambridge University Press, 1994a.

healer , “Evaluating Risky Consumption Paths: The Role of Intertemporal Substitutability," European Economic Review 38 (1994b), in press.

------ , “Risk-Taking, Global Diversification, and Growth," American Economic Review 84 (1994c), in press.

Popper, Helen, "Long-Term Covered Interest Parity: Evidence from Currency Swaps," Journal of International Money and Finance 12 (August 1993), pp. 439-448.

Razin, Assaf, and Efraim Sadka, “The Interactions between International Migration and International Trade," Research Memorandum 316, Vienna, Institute for Advanced Studies, February 1993.

Rubinstein, Mark, “An Aggregation Theorem for Securities Markets," Journal of

| Financial Economics 1 (September 1974), pp. 225-244.

Shiller, Robert J., "Aggregate Income Risks and Hedging Mechanisms," Cowles Foundation Discussion Paper No. 1048, New Haven, Conn., Yale University, June 1993.

Sinn, Stefan, "Saving-Investment Correlations and Capital Mobility: On the Evidence from Annual Data," Economic Journal 102 (September 1992), pp. 1162-1170.

Stockman, Alan C., and Harris Dellas, "International Portfolio Nondiversification and Exchange Rate Variability," Journal of International Economics 26 (May 1989), pp. 271-289.

Stockman, Alan C., and Linda L. Tesar, "Tastes and Technology in a Two-Country

Model of the Business Cycle: Explaining International Comovements, "

72

National Bureau of Economic Research Working Paper No. 3355, Cambridge, Mass., National Bureau of Economic Research, December 1990.

Stulz, René, "A Model of International Asset Pricing,". Journal. of Financial Economics 9 (1981), pp. 383-406.

Summers, Lawrence H., "Tax Policy and International Competitiveness," in Jacob A. Frenkel, ed., International, Aspects of Fiscal Policies, Chicago, University of Chicago Press, 1988,. pp. . 349-375

. Summers, Robert, and Alan Heston, "The Penn.World Table (Mark. 5): An Expanded. , Set of International Comparisons, 1950-1988," Quarterly Jaurnal,.of Economics 106 (May.1991), pp. 327-368.

Svensson, Lars E. O., "Trade in Risky Assets," American Economic Review +78 (June 1988), pp. 375-394.

Taylor, Alan M., "Domestic Saving and International Capital Flows Reconsidered," Evanston, Ill., Northwestern University, November 1993, processed.

Tesar, Linda L., "Savings, Investment, and International Capital Flows," Journal of International Economics 31 (August 1991), pp. 55-78.

------ , "International Risk-Sharing and Nontraded Goods," Journal of International Economics, 35 (August 1993), pp 69-89.

Tesar, Linda L., and Ingrid M. Werner, "Home Bias and the Globalization of Securities Markets," National Bureau of Economic Research Working Paper No. 4218, Cambridge, Mass., National Bureau of Economic Research, November 1992.

Ueda, Kazuo, "A Comparative Perspective on Japanese Monetary Policy: Short-Run Monetary Control and the Transmission Mechanism," in Kenneth J.

Singleton, ed., Japanese Monetary Policy, Chicago, University of Chicago

73

Press, 1993, pp. 7-29.

van Wincoop, Eric, "International Risksharing,” Milan, Italy, Innocenzo Gasparini Institute for Economic Research, 1992a, processed.

enw---, “Regional Risksharing,” Milan, Italy, Innocenzo Gasparini Institute for Economic Research, 1992b, processed.

s~---~, "Welfare Gains from International Risksharing,” Milan, Italy, Innocenzo Gasparini Institute for Economic Research, 1992c, processed.

Vikeren, Birger, “The Saving-Investment Correlation in the Short Run ancl in the Long Run,” Oslo, Norges Bank, 1991, processed.

Wong, David ¥., “What Do Saving-Investment Relationships Tell Us about Capital Mobility?" Journal of International Money and Finance 9 (March 19%0),

pp. 60-74.

74

IFDP Number

469

468

467

466

465

464 463

462

461 460

International Finance Discussion Papers

—_ Ke) Ke)

International Capital Mobility in the 1990s

The Effect of Changes in Reserve Requirements on Investment and GNP

International Economic Implications of the End of the Soviet Union

International Dimension of European Monetary Union:

Implications For The Dollar

European Monetary Arrangements: Implications for the Dollar, Exchange Rate Variability and Credibility

Fiscal Policy Coordination and Flexibility Under European Monetary Union: Implications for Macroeconomic Stabilization

The Federal Funds Rate and the Implementation of Monetary Policy: Estimating the Federal Reserve’s Reaction Function

Understanding the Empirical Literature on Purchasing Power Parity: The Post-Bretton Woods Era

Inflation, Inflation Risk, and Stock Returns

Are Apparent Productive Spillovers a Figment of Specification Error?

When do long-run identifying restrictions give reliable results?

1993

Fluctuating Confidence and Stock-Market Returns

Dollarization in Argentina

Author(s)

Maurice Obstfeld

Prakash Loungani

Mark Rush

William L. Helkie David H. Howard Jaime Marquez

Karen H. Johnson

Hali J. Edison Linda S. Kole

Jay H. Bryson

Allan D. Brunner

Hali J. Edison Joseph E. Gagnon William R. Melick John Ammer

Susanto Basu John S. Fernald

Jon Faust Eric M. Leeper

Alexander David

Steven B. Kamin Neil R. Ericsson

Please address requests for copies to International Finance Discussion Papers, Division of International Finance, Stop 24, Board of Governors of the Federal Reserve System, Washington, D.C. 20551.

75

IFDP Number

459 458

457

456

455

454

453

452

451

450 449

448

447

446

International Finance Discussion Papers

Titles

1993

Union Behavior, Industry Rents, and Optimal Policies

A Comparison of Some Basic Monetary Policy Regimes:

Implications of Different Degrees of Instrument Adjustment and Wage Persistence

Cointegration, Seasonality, Encompassing, and the Demand for Money in the United Kingdom Exchange Rates, Prices, and External Adjustment

in the United States and Japan

Political and Economic Consequences of Alternative Privatization Strategies

Is There a World Real Interest Rate?

Macroeconomic Stabilization Through Monetary and Fiscal Policy Coordination Implications for Monetary Union

Long-term Banking Relationships in General Equilibrium

The Role of Fiscal Policy in an Incomplete Markets Framework

Internal Funds and the Investment Function

Measuring International Economic Linkage with Stock Data

Macroeconomic Risk and Asset Pricing: Estimating the APT with Observable Factors

Near observational equivalence and unit root processes: formal concepts and implications

Market Share and Exchange Rate Pass-Through in World Automobile Trade

76

Author(s)

Phillip Swagel Dale W. Henderson Warwick ©. McKibbin

Neil R. Ericsson David F. Hendry Hong-Anh Tran

Peter Hooper Jaime Marquez

Catherine L. Mann Stefanie Lenway Derek Utter

Joseph E. Gagnon Mark D. Unferth

Jay H. Bryson Michael S. Gibson Charles P. Thomas Guy V.G. Stevens John Ammer Jianping Mei

John Amraer

Jon Faust

Robert C. Feenstra Joseph E. Gagnon Michael M. Knetter