Domestic Saving and International Capital Flows

The paper asks whether the saving generated in a country stays there to be invested or flows abroad to equalise yields, and the authors motivate it through two consequences that have nothing to do with the balance of payments: the national return to extra saving, and the incidence of capital taxation (§1). In a closed economy the nation captures the pretax domestic marginal product of capital on additional saving – “it is this pretax marginal product of capital that should influence national saving policy” – because the state collects the tax revenue as well as the investor receiving the after-tax yield. Under perfect mobility, incremental saving either leaves the country or displaces foreign capital that would have come in, so the nation gets only the investor’s net-of-tax return. Since “after-tax real yields are only about 50 percent of the pretax yield,” the practical stakes are large: for the United States “a pre-tax yield of 10 to 12 percent may be a convincing reason for more saving while a post-tax yield of 5 to 6 percent may be much less compelling.” The tax-incidence point is parallel: closed-economy results would be “radically altered” if capital were free to leave, since a large part of the burden of a capital income tax “could be shifted to domestic labor and to foreign capital owners.”

Because the two hypotheses make sharply different predictions for one slope coefficient, and because that coefficient is algebraically a statement about capital flows (§2). The authors estimate the ratio of gross domestic investment to GDP on the ratio of gross domestic saving to GDP across countries. “With perfect world capital mobility, there should be no relation between domestic saving and domestic investment: saving in each country responds to the worldwide opportunities for investment while investment in that country is financed by the worldwide pool of capital.” The predicted coefficient is not merely “small”: for a country infinitesimally small relative to the world it is zero, for a larger country it is “of the order of magnitude of its share of total world capital,” and “would thus vary among the OECD countries but would average less than 0.10.” The equivalence that makes the test interpretable is that gross domestic investment minus gross domestic saving is the net inflow of foreign investment, so regressing that inflow ratio on the saving ratio gives a coefficient of beta minus one; “testing the hypothesis that beta equals one is therefore equivalent to testing the hypothesis that the international capital flows do not depend on domestic savings rates.”

Period-average ratios for OECD countries over 1960-74, with the main cross-section regressions run on 16 countries (§2-3). Twenty-one OECD countries have data for the full fifteen years (Iceland, Portugal, Turkey and Yugoslavia do not); five of those – France, Luxembourg, Norway, Spain and Switzerland – are dropped from the main regressions because they switched national-income accounting method during the period, and the authors report that re-running some regressions on all 21 gave “very similar” coefficients. The variation being exploited is substantial and persistent: gross saving averaged 0.250 of GDP across the 21 countries with a standard deviation of 0.045, ranging from 0.372 in Japan to 0.184 in the United Kingdom; investment averaged 0.254 with a standard deviation of 0.041; and the cross-country saving-rate rankings are highly stable across five-year subperiods (correlations of 0.974, 0.931 and 0.895 between pairs of them). The average absolute gap between saving and investment was 1.3 percent of GDP, or 5.2 percent of gross saving.

0.887 with a standard error of 0.074 using gross flows over 1960-74, and 0.938 (0.091) using net flows: “Neither coefficient is significantly different from one while both are obviously incompatible with the hypothesis that the true value of beta is zero” (§3, Table 2). The five-year subperiod estimates on gross flows – 0.909 (0.060) for 1960-64, 0.872 (0.101) for 1965-69, 0.871 (0.092) for 1970-74 – are “similar to the overall coefficient,” with R-squared between 0.83 and 0.94. The authors state the reading plainly: “the evidence strongly contradicts the hypothesis of perfect world capital mobility and indicates that most of any incremental saving tends to remain in the country in which the saving is done.”

Gross, for two reasons – conceptual and econometric – and instrumenting lowers the net-flow estimate without disturbing the conclusion (§2-3). Conceptually, “it is the gross flow of savings that is, in principle, free to move from country to country in response to yield differentials.” Econometrically, depreciation accounting is “very imperfect, especially when there is significant inflation,” so measurement error in depreciation “would cause a spurious correlation between net saving and investment” – which is exactly the direction in which the net coefficients come out higher. To handle that, the authors instrument net saving with the gross saving rate, which “involves no estimate of depreciation but is likely to be highly correlated with true net saving.” All the instrumental-variable estimates are lower than OLS, “but the difference is never as large as 0.1”: for the full fifteen years the net coefficient falls from 0.938 (0.091) to 0.867 (0.102).

No, on any of the three margins the authors test (§3). A quadratic generalisation puts the coefficient on the squared saving ratio at 0.432 with a standard error of 1.146 for the full period – “always statistically insignificant and positive” – so “there is clearly no indication of nonlinearity.” Allowing the slope to vary with openness, measured by exports plus imports per dollar of GDP, gives an interaction coefficient that is “negative as expected but very small and not significantly different from zero” (-0.033, standard error 0.071 for the fifteen-year period). Using the logarithm of GDP as a size measure instead, the interaction terms came out negative – the opposite sign to the authors’ prior that large, more self-contained economies should retain more saving – and although they “differed from zero in a statistically significant way,” the estimates are “very small,” with the level coefficient at 0.999 (0.075). The authors’ summary: “while the link between domestic saving and investment may vary among countries, we found no evidence that it varied in relation to either the size of the economy or the importance of international trade.”

They test the most obvious candidate, concede the general point, and then shift the burden of proof (§2-3). The life-cycle theory makes population growth the leading exogenous determinant of aggregate saving, and it could plausibly raise investment too – but adding mean annual population growth “had almost no effect on the estimated value of beta,” with the growth variable’s own coefficient “very small and statistically quite insignificant.” The authors do not claim to have exhausted the possibilities: “There may of course be other variables that independently influence both saving and investment; although we do not pursue this question here, further analysis would clearly be desirable.” Their argument is instead that any such story is itself inconsistent with the hypothesis under test – a high coefficient “might reflect the fact that domestic saving and domestic investment are both stimulated by a high rate of return but that this interpretation is inconsistent with the hypothesis of perfect world capital mobility; with perfect capital mobility, the domestic saving rate does not depend on the domestic investment opportunities” – so the finding “would place on the defenders of that hypothesis the burden of identifying such common causal factors.”

For total and total private investment the three coefficients cannot be shown to differ; only at the level of corporate investment does the paper find a suggestion that corporate saving matters more (§4, Table 3). For the nine countries where the OECD splits saving into household, corporate and government components (Australia, Belgium, Canada, Finland, Germany, Japan, Netherlands, Sweden and the United Kingdom, 1961-74), total gross investment has a coefficient of 0.957 on aggregate saving; the household (1.17) and government (1.12) coefficients are slightly higher and the corporate coefficient (0.55) substantially lower, but “with the small sample of only 9 countries, these differences are not statistically significant” – the F-statistic for equality is 4.5 against a 5 percent critical value of 5.8. For net investment the coefficients are closer still. Gross corporate investment is the one case that looks different, appearing “most sensitive to gross corporate saving and substantially less sensitive to household and government saving,” which the authors read as possibly reflecting “institutional rigidities or portfolio preferences within national economies” or, alternatively, corporations choosing to save more where corporate investment is greater – and they note that “in either case, the evidence is inconsistent with the hypothesis of perfect world capital mobility.”

They show a weaker short-run link – an average coefficient of 0.64 across 21 countries – which the authors explicitly refuse to compare with the cross-section estimate (§5). Relating the change in the average investment ratio between 1960-69 and 1970-74 to the corresponding change in the saving ratio gives a coefficient of 0.724 (standard error 0.158, R-squared 0.52), “somewhat lower than [the] estimated long-run value” but not substantially so, which the authors read as implying that “domestic investment rates adjust within a few years to changes in saving rates.” Country-by-country annual regressions for 1960-74 average 0.64, with “substantial variation among the individual countries”: in 6 of the 21 countries a two-thirds confidence interval contains zero, while in 9 others it includes unity, and the United States, Germany and France are all close to one and not significantly different from it. The interpretive caution is the authors’ own: these coefficients “reflect short-run responses and are not comparable to the long-run equilibrium responses indicated by the cross-section coefficients,” with the gap attributable to lags in physical investment and to institutions holding transitory saving in liquid form.

By distinguishing the kinds of capital that move from the kinds that respond to national saving rates (§1, §6). They concede at the outset that “it is clear from the yields on short-term securities in the Eurocurrency market and the forward prices of those currencies that liquid financial capital moves very rapidly to arbitrage such short-term international yield differentials,” and argue only that such arbitrage involves “a small part of the total world capital stock” held in liquid form. For long-term capital they give three reasons for scepticism: risk aversion, since “the risks of investing in different countries and currencies are not perfectly correlated” so investors need not equalise expected yields, and these considerations “become increasingly important for longer-term and less liquid investments”; official restrictions on capital export, plus fear of future controls or adverse tax changes by host countries; and institutional rigidities, such as US saving institutions legally required to invest in local mortgages, insurance companies matching dollar liabilities with dollar assets, and “prudent man” fiduciary rules. They add an indirect empirical argument: with existing tax rules, net-return maximisation implies investors should specialise in particular countries, and “the absence of such specialisation is an indication that portfolio considerations or restrictions on capital movement prevent capital from flowing to maximize each investor’s net return.” Direct investment, finally, “appears to be associated with implementing marketing strategies, exploiting production knowledge, or overcoming trade restrictions rather than with an undifferentiated pursuit of profit opportunities” – which is “probably the major reason why individual countries are both importers and exporters of capital,” and why such flows “may exist even if they are not responsive to changes in domestic taxation or relative capital supplies.”

A comparative judgement between two views, not a structural estimate of a mobility parameter (§6). “The statistical evidence presented here on the relation between domestic investment and saving implies that the truth lies closer to the second view than to the first. International differences in domestic savings rates among major industrial countries have resulted in almost equal corresponding differences in domestic investment rates.” They then draw the three consequences they had set up in §1: the results shed light on “the true nature of the world capital market and the character of existing long-term capital movements”; they confirm “that it is appropriate, at least as an approximation, to study income distribution in general and tax incidence in particular with models that ignore international capital mobility”; and they imply “that the national return on domestic saving is approximately equal to the pretax domestic marginal product of capital since such saving does increase the domestic capital stock rather than either flowing abroad or replacing foreign investment at home.”