I. Introduction

The international environment of the 1980s was characterized by larger current account imbalances among major countries than at any point since the early post-war period. In 1987, the United States ran a current account deficit equal to about 3 1/2 percent of its Gross National Product (GNP), Japan recorded a surplus of a similar relative size, and the Federal Republic of Germany had a current account surplus that exceeded 4 percent of its GNP. Moreover, sizeable current account positions prevailed throughout most of the decade, leading to massive changes in net foreign assets and liabilities of these countries. The United States, which was the largest net creditor country during most of the post-war period up until the early 1980s, was at the end of 1990 the world’s largest net debtor, having net foreign liabilities of $423 billion—larger than the net liability position of the 10 most heavily indebted developing countries combined. Over this period, Japan and Germany accumulated sizeable net foreign claims; at the end of 1990 Japan was a net creditor to the extent of $328 billion, while Germany had net foreign assets of $265 billion at the end of 1989. ^2/

Relative to their output levels such claims positions are not unprecedented—for instance, the net foreign liability position of the United States was in 1990 about 8 percent of GNP, while Canada’s has long exceeded 20 percent of GNP—but because of the size of the countries concerned, their importance for international credit flows, and the key role of their currencies for the international monetary system, such payments imbalances are potentially cause for concern. An important issue is whether forces exist that bring about smooth adjustment of net foreign assets to equilibrium levels, or whether instead there are tendencies toward increasing asset and liability positions which may result in instability in international credit and foreign exchange markets.

The causes of the imbalances are not the subject of this paper; the extent to which they are related to the relative stance of fiscal policies in the United States, Japan, and Germany since the early 1980s has been discussed extensively elsewhere. ^3/ Instead, the analysis will consider stabilizing movements of domestic absorption that result from private and public sector behavior.

It is useful in considering the question to distinguish between two conceptually different but interrelated issues: first, whether a long-run equilibrium exists for net foreign assets (possibly as a function of other variables), and second, what short- or medium-run feedback mechanisms exist to bring about the long-run equilibrium if the latter exists. Long-run factors may include demographics and public debt accumulation, for instance. In order to assess whether a long-run equilibrium exists, we both test whether deviations from alternative long-run equilibria reverse themselves, and examine whether short-run feedback mechanisms sustaining the long-run equilibrium can be identified within a dynamic model. The existence of a stable long-run equilibrium and of short-run equilibrating feedback can in fact be viewed as facets of the same phenomenon (Hendry, 1986; Engle and Granger, 1987). Since changes in net foreign assets equal the current account (in the absence of valuation changes), and the current account balance equals output minus domestic absorption, some expenditure component should respond to deviations of net foreign assets from equilibrium. We test whether these feedbacks seem to be present in consumption, investment, and government spending.

Our analysis is related to a number of strands in the literature. First, since the article by Feldstein and Horioka (1980), many authors have considered why national saving and investment tend to be highly correlated. We consider not only the level of the current flows, but particularly the extent to which they cause asset stock accumulation. ^4/ Even if current accounts are relatively small, they may still bring about excessive accumulations of net claims or indebtedness; we are interested in forces that would prevent this happening. The existence of stabilizing feedbacks of asset stocks on saving and investment may help explain the empirical regularity identified by Feldstein and Horioka.

This paper is also related to the literature on the transfer problem (see, among others, Johnson, 1956; Frenkel and Razin, 1987). That literature considers whether transfers of wealth between countries have effects on expenditure that feed back onto the current account balance, allowing the transfer to be effected without terms of trade changes. Since accumulations of net foreign claims correspond to transfers of wealth, depending on spending propensities, stabilization of foreign claims may or may not require real exchange rate changes.

The role of the exchange rate as a mechanism of adjustment is not considered explicitly in this paper. To the extent that changes in exchange rates alter current account positions, they operate by changing domestic saving and investment (or absorption relative to output) because of the identity linking the variables. Thus our estimates of the response of absorption to net foreign assets include, among other channels, exchange rate effects. A different question relates to the effects of exchange rate changes on the valuation of net foreign asset stocks. Valuation effects result from aggregation of assets and liabilities in different currencies at varying exchange rates, and require for their calculation a detailed breakdown of those assets and liabilities. Published data on net foreign assets typically take account of valuation effects to some extent (this is the case for the three countries we consider; see Appendix for data sources), so they do not equal the accumulation of current account flows. However, there is no presumption that exchange rate movements automatically stabilize net foreign asset positions through such valuation effects.

Finally, in the absence of Ricardian Equivalence (Barro, 1974), the existence of a long-run equilibrium for net foreign assets is related to the existence of a long-run equilibrium for the stock of government debt (both as a ratio to GNP). Net foreign liabilities play the same role for the economy’s intertemporal budget constraint as public debt for the government’s budget constraint. ^5/ There have been a number of studies examining the empirical implications of the government’s intertemporal budget constraint for the conduct of budgetary policy (Hamilton and Flavin, 1986; and Kremers, 1989), and studying the sustainability of current account positions (Wickens and Uctum, 1990; Trehan and Walsh, 1991; Ahmed and Rogers, 1992). It seems to us that the proper question is not whether the economy ultimately satisfies its intertemporal budget constraint, but rather whether there are feedback mechanisms that operate smoothly to ensure that disruptive adjustment can be avoided.

The plan of the paper is as follows. The long-run determination of net foreign asset positions is discussed in Section II. This is followed by a brief description of the econometric methods in Section III. The empirical results are in Section IV, and Section V concludes.

II. Long-Run Influences on Net Foreign Assets

It is useful to begin by considering the pattern of net foreign claims in a two-country version of the familiar Yaari-Blanchard-Weil model (Yaari, 1965; Blanchard, 1985; Weil, 1985). If birth rates are positive, then, as shown in Weil (1985) and Buiter (1988), Ricardian equivalence does not hold, and the choice between financing government expenditure through taxation or issuing bonds has real effects. As shown by Blanchard (1985), higher government debt in a small open economy is associated with lower net foreign assets.

This model does not take into account the life-cycle nature of saving (Ando and Modigliani, 1965), which, as Blanchard (1985, p. 235) points out, is difficult to handle tractably. However, demographic variables that reflect the age-structure of the population seem to be important determinants of the cross-country variations of saving rates (Modigliani, 1970; Graham, 1987; Masson and Tryon, 1990), and hence should affect net foreign asset positions.

Consider a simple overlapping generations model in which individuals work and save in the first period of their lives, and retire and dissave in the second period; there are no bequests and the wage rate is constant. In this world, the saving rate will depend on how many workers there are relative to retirees. If the latter are numerically more important (for instance, due to a decline in the birth rate), then the economy’s saving rate may even be negative. However, the implications for asset stocks are quite different: if there are more retirees this period, then, since they own the assets (having accumulated them in the previous period), the economy’s stock of assets will be larger, ceteris paribus. In this world, asset accumulation only occurs because individuals want to transfer wealth to their retirement years. It should therefore be the case that an economy with an older population has higher net foreign assets in steady state than an identical economy with a younger population (the two economies could have the same population growth, but the former a lower birth rate and mortality rate), since the former economy involves more transfers of wealth between time periods.

Of course, demographic transitions take many decades and we are unlikely to observe economies in steady state with equal population growth rates. Therefore, net foreign assets need not be a constant ratio to GNP, and demographic variables should be important for long-run movements of this ratio. In our empirical work, we examine the effects of the relative ratios of both the young (under 15 years) and the old (65 years and over) to the working-age population, where in each case we compare the home country’s age structure to that in other countries. In contrast to the aged dependency ratio, the effect of an increase of the youth dependency ratio is not so clear cut since providing for their current consumption may decrease household saving, but providing for their future needs (e.g., college education) may increase saving.

III. Econometric Methods

The idea underlying cointegration is to study the behavior of variables moving apart in the short run, but which are brought back to an equilibrium relationship by government policy or market forces, or both, in the long run. A variable is defined to be integrated of order one (denoted by I(1)) if it must be differenced once to produce a stationary series (denoted by I(0)), that is, a series with constant mean and constant, finite variance. A set of variables, each I(1), is said to be cointegrated if a linear combination of them is I(0) (i.e., if they “move together in the long run”). ^6/ That linear combination is characterized by a cointegrating vector.

The empirical analysis in the next section starts by testing whether the ratio of net foreign assets to GNP is stationary. In the event that non-stationarity of this ratio cannot be rejected, it will subsequently be assessed whether the ratio is cointegrated with other non-stationary variables, in particular, government debt and demographic variables.

Given the small available data sample (the postwar period), it is important to follow an approach to testing that has high power to reject non-stationarity. We therefore use the maximum-likelihood approach of Johansen (1988) to test for cointegration rather than the Engle-Granger single-equation procedure. We also estimate dynamic error correction models, using the methodology of Hendry (1987) to go from general specifications to more parsimonious ones; as suggested in Phillips and Loretan (1991), we use non-linear least squares to impose the restrictions implied by the cointegrating vectors.

IV. Empirical Results

1. The long run

The post-war patterns of the ratios of net foreign assets to GNP for the United States, Japan, and Germany are shown in Figure 1. ^7/ The graphical evidence suggests no sustained tendency on the part of any of these ratios to return to a constant long-run equilibrium. The net foreign asset positions of Japan and Germany, relative to GNP, have fluctuated along an upward trend, while that of the United States has fallen dramatically in the 1980s after following a more stable pattern in the previous decades.

The graphical impression is confirmed by the fact that, for this sample, non-stationarity of the ratio cannot be rejected for any of the three countries (Table 1). However, the non-stationarity of its first difference can be rejected, indicating that the ratio is I(1). Hence, if a stationary long-run equilibrium exists, it must be a function of at least one other non-stationary variable.

In order to explore this possibility, we examine the cointegration of F/Y (the ratio of net foreign assets to GNP) with the government debt/GNP ratio and demographic variables. The discussion of Section II suggested that, in equilibrium, higher net foreign assets (scaled by GNP) should be associated with lower government debt at home and higher government debt abroad. Two alternative specifications were tried for the debt variable: the domestic debt/GNP ratio, B/Y, and a relative debt variable B/Y-B*/Y*, with B*/Y* based on data for the six remaining G-7 countries. In practice the domestic debt ratio seemed to work better, possibly because of limited cross-country comparability of debt data; the relative debt variable is therefore not reported. The two demographic variables included in the analysis are the domestic ratio of the population aged under 15 years to that from 15 to 64 years, minus the corresponding ratio for the remaining G-7 countries (RDEM1), and an analogous variable for the ratio of the population aged 65 years and over to that from 15 to 64 (RDEM2).

Table 1.

Tests of Order of Integration ^1/

^1/A constant and a time trend were included in the regressions, and the lagged change in the variable if there was evidence of residual serial correlation.

^2/Truncated as necessary to account for lags.

^3/The calculation of p-values in TSP is based on MacKinnon (1991).

Table 1.

Tests of Order of Integration ^1/

Variable		Sample 2/	ADF Statistic	(p - value) 3/
United States
	F/Y	1950-90	−1.44	(0.85)
	Δ(F/Y)		−3.65	(0.04)
	B/Y	1950-90	−1.49	(0.84)
	Δ(B/Y)		−7.73	(0.00)
	RDEM1	1950-90	−3.06	(0.13)
	ΔRDEM1		−1.07	(0.93)
	RDEM2	1950-90	−2.89	(0.18)
	ΔRDEM2		−1.78	(0.73)
Japan
	F/Y	1950-90	−2.21	(0.51)
	Δ(F/Y)		−8.65	(0.00)
	B/Y	1954-90	−2.27	(0.46)
	Δ(B/Y)		−1.31	(0.88)
	RDEM1	1950-90	−3.07	(0.13)
	ΔRDEM1		−0.90	(0.95)
	RDEM2	1950-90	−2.06	(0.58)
	ΔRDEM2		−1.66	(0.77)
Germany
	F/Y	1950-89	−0.64	(0.97)
	Δ(F/Y)		−4.51	(0.00)
	B/Y	1954-89	−2.45	(0.37)
	Δ(B/Y)		−3.14	(0.11)
	RDEM1	1950-90	−2.97	(0.15)
	ΔRDEMl		−1.90	(0.67)
	RDEM2	1950-90	−1.95	(0.65)
	ΔRDEM2		−1.96	(0.64)

^1/A constant and a time trend were included in the regressions, and the lagged change in the variable if there was evidence of residual serial correlation.

^2/Truncated as necessary to account for lags.

^3/The calculation of p-values in TSP is based on MacKinnon (1991).

View Table

Table 1.

Tests of Order of Integration ^1/

Variable		Sample 2/	ADF Statistic	(p - value) 3/
United States
	F/Y	1950-90	−1.44	(0.85)
	Δ(F/Y)		−3.65	(0.04)
	B/Y	1950-90	−1.49	(0.84)
	Δ(B/Y)		−7.73	(0.00)
	RDEM1	1950-90	−3.06	(0.13)
	ΔRDEM1		−1.07	(0.93)
	RDEM2	1950-90	−2.89	(0.18)
	ΔRDEM2		−1.78	(0.73)
Japan
	F/Y	1950-90	−2.21	(0.51)
	Δ(F/Y)		−8.65	(0.00)
	B/Y	1954-90	−2.27	(0.46)
	Δ(B/Y)		−1.31	(0.88)
	RDEM1	1950-90	−3.07	(0.13)
	ΔRDEM1		−0.90	(0.95)
	RDEM2	1950-90	−2.06	(0.58)
	ΔRDEM2		−1.66	(0.77)
Germany
	F/Y	1950-89	−0.64	(0.97)
	Δ(F/Y)		−4.51	(0.00)
	B/Y	1954-89	−2.45	(0.37)
	Δ(B/Y)		−3.14	(0.11)
	RDEM1	1950-90	−2.97	(0.15)
	ΔRDEMl		−1.90	(0.67)
	RDEM2	1950-90	−1.95	(0.65)
	ΔRDEM2		−1.96	(0.64)

^1/A constant and a time trend were included in the regressions, and the lagged change in the variable if there was evidence of residual serial correlation.

^2/Truncated as necessary to account for lags.

^3/The calculation of p-values in TSP is based on MacKinnon (1991).

View Full Size

Net Foreign Assets to GNP Ratios: Actual and Equilibrium Values

Citation: IMF Working Papers 1993, 033; 10.5089/9781451979244.001.A001

• Download Figure
• Download figure as PowerPoint slide

The first step of the empirical analysis is to ascertain that the variables to be used in the cointegration regressions are indeed I(1). The non-stationarity of B/Y cannot be rejected but that of Δ(B/Y) can be with fairly high confidence (Table 1), confirming that B/Y is I(1) for all countries except Japan. However, there may be grounds for questioning whether B/Y is in fact I(1); see Kremers (1989). In any case, one should be cautious of inference with such relatively short samples. For the sample under investigation, the tests suggest that the two demographic variables may be integrated of different order. While the relative youth dependency ratio (RDEMl) seems to be stationary for all three countries (at least at a 15 percent significance level), the relative elderly dependency ratio (RDEM2) is not. In fact, the data suggest that RDEM2 may not even be I(1). However, demographic ratios suffer from the fact that the underlying census data are not available on a yearly basis, and hence involve some interpolation. We therefore do not place too great a weight on tests of their apparent order of integration, and we include RDEM2 in the cointegrating equations for all three countries.

The tests for cointegration of the set of variables comprising F/Y, B/Y, and RDEM2 are presented in Table 2. For the United States and Japan, there is a clear rejection of the hypothesis of no cointegrating vectors; moreover, that vector, reported in Table 2, is unique. It implies that a higher domestic debt ratio lowers, and an older population raises, net foreign assets in the long run, as suggested in Section II.

For Germany, in contrast, the picture is less clear: these three variables do not seem to be cointegrated, and, moreover, debt has the wrong sign. We proceeded to include other variables in the cointegrating vector for Germany, since it appears that the debt ratio, which rises over the sample period, may be proxying for them. We separately included the real oil price, the terms of trade, and the government spending to GNP ratio; however, there was still no evidence of cointegration. An alternative specification containing both demographic variables is reported in Table 2. RDEMl is included, despite appearing to be I(0), partly because of our concern about the reliability of order of integration tests when applied to these interpolated variables, and partly because it seems logical that the two variables should be of the same order of integration since they depend on the same mortality and fertility factors. The inclusion of RDEMl allows rejection of the hypothesis of no cointegration, and also reverses the sign of B/Y. However, the coefficient values are unreasonably large for the two demographic variables—and sensitive to sample period—no doubt because of collinearity between them. Therefore, in our subsequent work we use the simpler specification which includes only RDEM2 for all three countries, despite the test result for Germany suggesting lack of cointegration for this specification.

Taken together, this evidence—though not yet very firm—nevertheless suggests that, for the three countries under investigation, a stationary long-run equilibrium relationship for the F/Y ratio may exist, and that it is a function of demographic factors as well as the public debt/GNP ratio. In order to help an understanding of the implications of these equations for long-run equilibrium, we compare their actual and fitted values in Figure 1. It is notable that trend movements of net foreign assets seem to follow their calculated long-term equilibrium path, but there are substantial short-run deviations. For instance, in the early 1950s, actual F/Y was in the United States considerably above its equilibrium, and in Japan and Germany, below it—perhaps a legacy of the war.

Table 2.

Cointegration Tests for Net Foreign Assets, Debt, and Dependency Ratios

^*Significant at 5 percent level. See Johansen and Juselius (1990), Table A.3.

Table 2.

Cointegration Tests for Net Foreign Assets, Debt, and Dependency Ratios

Tests of Number of Cointegrating Vectors					Largest Eigenvalue	Corresponding Eigenvector
	r = 0	r = 1	r = 2	r = 3
United States (1950-90)
	51.2*	19.4	2.9	—	0.56	F/Y = 0.32 -0.77 B/Y + 0.71 RDEM2
Japan (1954-90)
	37.2*	18.4	8.6	—	0.43	F/Y = 0.11 -0.50 B/Y + 1.76 RDEM2
Germany (1954-89)
	25.6	13.8	6.2	—	0.29	F/Y = -0.18 + 0.67 B/Y + 1.36 RDEM2
	54.6*	28.4	14.5	5.2	0.54	F/Y = -4.03 -0.31 B/Y -20.6 RDEM1 + 59.9 RDEM2

^*Significant at 5 percent level. See Johansen and Juselius (1990), Table A.3.

View Table

Table 2.

Cointegration Tests for Net Foreign Assets, Debt, and Dependency Ratios

Tests of Number of Cointegrating Vectors					Largest Eigenvalue	Corresponding Eigenvector
	r = 0	r = 1	r = 2	r = 3
United States (1950-90)
	51.2*	19.4	2.9	—	0.56	F/Y = 0.32 -0.77 B/Y + 0.71 RDEM2
Japan (1954-90)
	37.2*	18.4	8.6	—	0.43	F/Y = 0.11 -0.50 B/Y + 1.76 RDEM2
Germany (1954-89)
	25.6	13.8	6.2	—	0.29	F/Y = -0.18 + 0.67 B/Y + 1.36 RDEM2
	54.6*	28.4	14.5	5.2	0.54	F/Y = -4.03 -0.31 B/Y -20.6 RDEM1 + 59.9 RDEM2

^*Significant at 5 percent level. See Johansen and Juselius (1990), Table A.3.

It appears that the U.S. net foreign asset equilibrium, conditional on the paths of B/Y and RDEM2 (and assuming that any possible I(0) influences are zero), sharply declined in the 1980s. This primarily reflects the change in stance of U.S. fiscal policy. The long-run equilibrium of the Japanese net foreign asset position has shown an upward movement since 1985, reflecting both the aging of the population and fiscal consolidation. In Germany, there has been a gradual, but steady rise in the estimated net foreign asset equilibrium, where the debt ratio is no doubt capturing the effects of other variables (including RDEM1) that contribute to the positive trend.

2. The dynamics

In this sub-section, dynamic error correction models are estimated for components of absorption in order to shed some light on the dynamic forces underlying the long-run equilibrium, i.e., the feedback mechanisms through absorption. First note that in the absence of valuation changes, the first difference of net foreign assets (F) can be written:

$\begin{array}{cc}\left(1\right)& \mathrm{\Delta }F=Y-A\end{array}$

where Y is nominal GNP and A is domestic absorption, which is equal to the sum of consumption (C), investment (I), and government spending (G). Therefore,

$\begin{array}{cc}\left(2\right)& \mathrm{\Delta }\left(F/Y\right)=1-A/Y-\left(F_{-1}/Y_{-1}\right)\mathrm{\Delta }Y/Y\end{array}$

The existence of cointegration implies an error correction model for F/Y, and if nominal income growth ΔY/Y is given in the long-run by other factors (e.g., money supply growth), then this feedback mechanism should operate through one or several components of A. ^8/

Standard models suggest several mechanisms by which net foreign assets may affect absorption. 1) Since it is a component of wealth, higher F/Y should raise consumption, thereby increasing imports and lowering Δ(F/Y). 2) If there is a risk premium on interest rates on foreign borrowing that depends on the net foreign liability position, then higher F/Y should lower domestic interest rates and stimulate investment (and perhaps also consumption). 3) The government may vary its fiscal policy instruments to prevent the current account or net foreign assets from diverging too much from some desired level (Summers, 1988).

These three feedback mechanisms operate through different components of absorption; thus, it is of interest to discover whether deviations of F/Y from its long-run equilibrium affect consumption, investment, or government spending.

For each of the three countries, we estimated a system of dynamic equations for the annual change in the three components of domestic absorption scaled by GNP, including the lagged deviations from the respective long-run cointegrating relationships as representations of disequilibrium feedback (these deviations will be denoted by u). ^9/ The cointegrating vectors are reestimated here, and the appropriate cross-equation restrictions are imposed. ^10/ Initially, the same general standard equation was run for each country, including all the variables listed in the left margin of Table 3. In order to gain precision on this relatively limited sample, the general equations were subsequently simplified to a more specific lag structure for each country. The step from “general to specific” was guided by the diagnostic tests, though in each case the error correction term and the constant were retained. ^11/

The dynamic analysis confirms the existence of a long-run relationship in which the net foreign asset ratio is negatively related to the government debt ratio and positively related to the aged dependency ratio. Table 3 presents evidence of positive error correction feedback from foreign asset disequilibrium onto some expenditure component in each of the three countries. Feedback onto consumption is positive for two of the three countries, the United States and Japan, though t-ratios are only about 1.5. For investment, feedback is positive for the United States and Germany, and significant for the former. For government spending, the feedback is positive and significant for Japan, while for the other two countries coefficients are negative. Thus, the results suggest some support for each of the three linkages for stabilizing the net foreign asset ratio, depending on the country. Though the equations suggest considerable diversity across countries, other tentative conclusions also emerge. There is evidence of significant stabilizing feedback of lagged government debt on government spending in the three countries, and of contemporaneous crowding out of investment by increases in government debt in the United States and Germany, for which Δ(B/Y) has a negative sign in the investment equation.

Table 3.

Error Correction Models for Consumption, Investment, and Government Spending

(Standard errors in parentheses)

^*Indicates rejection of the null hypothesis of absence of serial correlation, homoscedasticity, or normality at 5 percent significance level.

^**Indicates rejection of null at 1 percent significance level.

^1/Constant was included but is not reported.

^2/Statistics are the explained sum of squares and standard error of the regression, and tests for autoregressive conditional heteroscedasticity, residual auto correlation for lags 1 to n, and normality, respectively. The ARCH, AR, and NORM test statistics are distributed as χ2. See Spanos (1986) for a description of the tests, including appropriate degrees of freedom.

Table 3.

Error Correction Models for Consumption, Investment, and Government Spending

(Standard errors in parentheses)

		United States (1952-90)			Japan (1957-90)			Germany (1957-89)
Cointegrating		F/Y = -0.55 B/Y + 2.64 RDEM2			F/Y = -0.11 B/Y + 5.91 RDEM2			F/Y = -0.58 B/Y + 3.17 RDEM2
Vector		(0.11)		(1.02)	(0.22)		(2.28)	(0.59)		(1.68)
Coefficients 1/		Δ(C/Y)	Δ(I/Y)	Δ(G/Y)	Δ(C/Y)	Δ(I/Y)	Δ(G/Y)	Δ(C/Y)	Δ(I/Y)	Δ(G/Y)
	û-1	0.024 (0.018)	0.074 (0.024)	−0.028 (0.022)	0.050 (0.038)	−0.046 (0.041)	−0.037 (0.010)	−0.020 (0.018)	0.080 (0.053)	−0.024 (0.017)
	Δ(B/Y)	0.296 (0.049)	−0.170 (0.051)	0.197 (0.058)				0.513 (0.057)	−0.162 (0.103)	0.215 (0.045)
	Δ(B/Y)-1		0.164 (0.036)	−0.260 (0.044)			−0.041 (0.016)	−0.235 (0.054)		−0.276 (0.042)
	Δ(F/Y)	0.095 (0.056)	−0.076 (0.055)			−0.346 (0.114)			−0.400 (0.130)
	Δ(F/Y)-1		−0.072 (0.053)		−0.345 (0.157)	0.678 (0.169)	−0.134 (0.038)	0.104 (0.068)
	ΔRDEM2		−1.284 (0.574)						0.851 (0.626)
Lagged dependent variable		−0.205 (0.083)	0.230 (0.095)	0.272 (0.080)	0.207 (0.092)	0.340 (0.086)			0.144 (0.147)
Statistics 2/
	R2	0.478	0.551	0.607	0.249	0.460	0.435	0.764	0.529	0.695
	$\hat{\sigma }$	0.0053	0.0047	0.0058	0.0099	0.0105	0.0024	0.0039	0.0069	0.0031
	ARCH (1)	0.57	1.12	0.27	0.99	1.13	1.61	0.29	0.11	0.02
	AR (1)	2.23	4.83*	4.97*	0.62	0.56	0.09	1.72	1.66	0.84
	AR (1 to 4)	1.94	9.62*	13.2*	8.92	4.99	2.60	13.2*	3.40	1.68
	NORM	0.90	0.32	67.8**	0.67	0.44	1.04	0.58	0.02	1.39

^*Indicates rejection of the null hypothesis of absence of serial correlation, homoscedasticity, or normality at 5 percent significance level.

^**Indicates rejection of null at 1 percent significance level.

^1/Constant was included but is not reported.

^2/Statistics are the explained sum of squares and standard error of the regression, and tests for autoregressive conditional heteroscedasticity, residual auto correlation for lags 1 to n, and normality, respectively. The ARCH, AR, and NORM test statistics are distributed as χ2. See Spanos (1986) for a description of the tests, including appropriate degrees of freedom.

View Table

Table 3.

Error Correction Models for Consumption, Investment, and Government Spending

(Standard errors in parentheses)

		United States (1952-90)			Japan (1957-90)			Germany (1957-89)
Cointegrating		F/Y = -0.55 B/Y + 2.64 RDEM2			F/Y = -0.11 B/Y + 5.91 RDEM2			F/Y = -0.58 B/Y + 3.17 RDEM2
Vector		(0.11)		(1.02)	(0.22)		(2.28)	(0.59)		(1.68)
Coefficients 1/		Δ(C/Y)	Δ(I/Y)	Δ(G/Y)	Δ(C/Y)	Δ(I/Y)	Δ(G/Y)	Δ(C/Y)	Δ(I/Y)	Δ(G/Y)
	û-1	0.024 (0.018)	0.074 (0.024)	−0.028 (0.022)	0.050 (0.038)	−0.046 (0.041)	−0.037 (0.010)	−0.020 (0.018)	0.080 (0.053)	−0.024 (0.017)
	Δ(B/Y)	0.296 (0.049)	−0.170 (0.051)	0.197 (0.058)				0.513 (0.057)	−0.162 (0.103)	0.215 (0.045)
	Δ(B/Y)-1		0.164 (0.036)	−0.260 (0.044)			−0.041 (0.016)	−0.235 (0.054)		−0.276 (0.042)
	Δ(F/Y)	0.095 (0.056)	−0.076 (0.055)			−0.346 (0.114)			−0.400 (0.130)
	Δ(F/Y)-1		−0.072 (0.053)		−0.345 (0.157)	0.678 (0.169)	−0.134 (0.038)	0.104 (0.068)
	ΔRDEM2		−1.284 (0.574)						0.851 (0.626)
Lagged dependent variable		−0.205 (0.083)	0.230 (0.095)	0.272 (0.080)	0.207 (0.092)	0.340 (0.086)			0.144 (0.147)
Statistics 2/
	R2	0.478	0.551	0.607	0.249	0.460	0.435	0.764	0.529	0.695
	$\hat{\sigma }$	0.0053	0.0047	0.0058	0.0099	0.0105	0.0024	0.0039	0.0069	0.0031
	ARCH (1)	0.57	1.12	0.27	0.99	1.13	1.61	0.29	0.11	0.02
	AR (1)	2.23	4.83*	4.97*	0.62	0.56	0.09	1.72	1.66	0.84
	AR (1 to 4)	1.94	9.62*	13.2*	8.92	4.99	2.60	13.2*	3.40	1.68
	NORM	0.90	0.32	67.8**	0.67	0.44	1.04	0.58	0.02	1.39

^*Indicates rejection of the null hypothesis of absence of serial correlation, homoscedasticity, or normality at 5 percent significance level.

^**Indicates rejection of null at 1 percent significance level.

^1/Constant was included but is not reported.

^2/Statistics are the explained sum of squares and standard error of the regression, and tests for autoregressive conditional heteroscedasticity, residual auto correlation for lags 1 to n, and normality, respectively. The ARCH, AR, and NORM test statistics are distributed as χ2. See Spanos (1986) for a description of the tests, including appropriate degrees of freedom.

V. Conclusion

This study has sought to confirm the presence of stabilizing mechanisms for the net foreign asset positions of the three largest industrial countries in the post-World War II period. Persistent and large current account imbalances have led to quite dramatic changes in the net international positions of the United States, Japan, and Germany. However, various feedback effects from foreign asset stocks may act as stabilizing mechanisms to prevent a continued increase of these net foreign asset or liability positions, and to ensure an eventual return to long-run equilibrium.

Cointegration tests were used to investigate empirically the nature of the long-run equilibrium relationship for net foreign assets of the United States, Japan and Germany using postwar data. These tests suggest the existence of a long-run relationship between the net foreign asset/GNP ratio, the public debt/GNP ratio, and dependency ratios relative to other G-7 countries.

Developments in the 1980s seem to have differed between the United States on the one hand and Japan and Germany in an important respect: the conditional long-run net foreign asset equilibrium of the United States moved sharply downward, reflecting a rapid accumulation of public debt. In contrast, the Japanese and German authorities implemented policies to consolidate the public finances; the net foreign assets of these countries increased strongly.

The evidence of these long-run relationships is supported by error correction models of components of domestic absorption. These results suggest, however, that stabilizing feedback operates through different linkages in different countries: primarily through private investment in the United States and Germany and through government spending in Japan, though there is also weaker evidence of a consumption feedback in the United States and Japan.