The author wishes to thank M. Abramovitz, S. Engerman, G. Hawke, H. Klein, D. Mathieson, C. O'Grada, P. Temin and the Conference on “The International Economy and British Growth, 1860–1913,” Harvard University, September 1973, where an earlier version of this paper was presented. The financial support of the Council for Research in the Social Sciences of Columbia University is acknowledged gratefully.

¹ Feinstein, C. H., National Income Expenditure and Output of the United Kingdom, 1855–1865 (Cambridge: Cambridge University Press, 1972), pp. T8Google Scholar, T12, T37–8, T106. Neither proportion manifests a statistically significant trend.

² An issue which remains unsettled is whether the risk-included return on the aggregate of overseas U.K. assets was lower than the risk-included return on domestic assets. Considerable evidence suggests that U.K.-held overseas assets were heavily weighted toward the lower end or the risk spectrum and hence their risk-included return could well have been lower than that obtained at home. For a recent discussion of the evidence see Wm. Kennedy, “Foreign Investment, Trade and Growth in the U.K., 1870–1913,” unpub. ms. prepared for the Conference on “The International Economy and British Growth, 1860–1913,” Harvard University, September, 1973.Google Scholar

³ Williamson, J. G., American Growth and the Balance of Payments, 1820–1913 (Chapel Hill, N.C.: University of North Carolina Press, 1964), Ch. 4Google Scholar; Ford, A. G., “British Investment in Argentina and Long Swings, 1880–1913,” Journal of Economic History, XXXI (1971), 650–663CrossRefGoogle Scholar; Richardson, H. W., “British Emigration and Overseas Investment, 1870–1914,” Economic History Review, 2d Series, XXV (1972) 99–113.Google Scholar

⁴ Feis, H., Europe: the World's Banker, 1870–1914 (New York: Norton, 1965), p. 23.Google Scholar

⁵ In both level and fluctuation the yield on U.K.-held U.S. investment was closely related to that on Latin American assets and, to a lesser extent, those on western and eastern European assets. Edelstein, M., “The Rate of Return on U.K. Home and Foreign Investment, 1870–1913,” unpublished Ph.D. dissertation, University of Pennsylvania, 1970, pp. 198–202.Google Scholar

⁶ Imlah, A. H., Economic Elements in the Pax Britannica (Cambridge: Harvard University Press, 1958), pp. 61–81.CrossRefGoogle Scholar

⁷ See for example, Sharpe, W. F., Portfolio Theory and Capital Markets (New York: McGraw-Hill, 1970).Google Scholar

⁸ Arrow, K. J., Essays in the Theory of Risk-Bearing (Chicago: Markham, 1971), Ch. 3.Google Scholar

⁹ Miller and Whitman tried the level of national income as a proxy for cyclical risk patterns but its sign varied across assets in our regressions and was often statistically insignificant. From this inconsistency and other evidence, it appears that national income is also a bad proxy for things other than risk and was therefore dropped from consideration. Miller, N. C. and Whitman, M. v. N., “A Mean-Variance Analysis of U.S. Long-term Portfolio Foreign Investment,” Quarterly Journal of Economics, LXXXIV (1970), 175–196.CrossRefGoogle Scholar

¹⁰ See Kindleberger, C. P., Economic Growth in Finance and Britain, 1850–1950 (New York: Simon and Schuster, 1969), pp. 61–67Google Scholar, for an excellent summary of the debate.

¹¹ Committee on Finance and Industry, Report (London: H.M.S.O., 1931), p. 171.Google Scholar

¹² Edelstein, M., “Rigidity and Bias in the British Capital Market, 1870–1913,” in McCloskey, D. N. (ed.), Essays in a Mature Economy: Britain Since 1840 (London: Methuen, 1971), pp. 85–94.Google Scholar

¹³ Ulmer, Melville J., Capital in Transportation, Communications, and Public Utilities: Its Formation and Financing (Princeton: Princeton University Press-NBER, 1960), p. 502.Google Scholar

¹⁴ Kmenta, J. and Williamson, J. G., “The Determinants of Investment Behavior: United States Railways, 1872–1941,” Review of Economics and Statistics, XLVIII (1966), 172–181CrossRefGoogle Scholar; Neal, L., “Investment Behavior by American Railroads: 1897–1914,” Review of Economics and Statistics, LI (1969), 126–135.CrossRefGoogle Scholar

¹⁵ To further test this conclusion, the exogenous variables from the U.S. railway investment functions of Kmenta and Williamson, and Neal were introduced separately into the reduced form equation in the yield on U.S. railway bonds in the London market. This equation was derived from the portfolio demand and asset supply relations shortly to be presented in the text. In general, the exogenous variables from the U.S. investment demand equation were not significantly related to the London yield variable.

¹⁶ The partial correlation of N.Y. and London yields on first-class U.S. railway bonds was .98, 1871–1913; for changes in their levels, the partial correlation coefficient was .77.

¹⁷ While the interpretation of the estimated U.S. wealth elasticity, is fairly unambiguous for most years of the 1870–1913 period, it is likely that in the six or seven years just before World War I the wealth variable is, to some important degree, an instrument for both the general U.S. demand for railway assets and the surging influence of railway corporate savings. Unfortunately, due to the presence of multicollinearity it was impossible to give separate econometric expression to these two influences.

¹⁸ Hall, A. R. (ed.), The Export of Capital from Britain, 1870–1913 (London: Methuen, 1968), pp. 13Google Scholar, 36.

¹⁹ Full notes on the construction of each data series may be found in die Data Appendix.

²⁰ W2 may possibly be affected by double-counting. Short-term commercial loans were extended to holders of long term portfolio investments and the totals for domestic and overseas portfolio investments include the equity of home and foreign banks. To test for possible bias arising from this factor, a third U.K. wealth variable, W3, was constructed which consisted of domestic and overseas portfolio investments alone. Regressions run with W3 as the U.K. wealth variable yielded coefficients which were not significantly different from those run with W2.

²¹ A crude approximation of domestic U.S. holdings may be obtained by subtracting U.K. holdings of U.S. assets, A, from the series on the total outstanding debt of the U.S. railway found in U.S. Bureau of the Census, Historical Statistics of the United States (Washington: G.P.O., 1960).Google Scholar

²² In the first attempts to estimate the structural equations, indices of equity rates of return which incorporated capital gains were utilized as estimators of the expected rate of return on home and overseas assets, both by themselves and alongside the estimators of debenture yields noted in the text. (See M. Edelstein, “The Rate of Return,” Appendix III, for estimators of equity return incorporating capital gains.) These early tests suggested that the debenture yield series was the best indicator of year-to-year movements in the expected return to financial assets. While the regression coefficients on the equity return variables were statistically insignificant, this was rarely the case with the coefficients on the debenture yield variables. The low explanatory value of equity rates of return incorporating capital gains is consistent with studies of recent financial behavior, a result which is, generally attributed to the inability of this equity return estimator to effectively capture investor views of corporate retained earnings and the tendency of dividend yields to move inversely with expected changes in earnings. See, for example, C. Wright, “Savings and the Rate of Interest,” in Harberger, A. C. and Bailey, M. J., The Taxation of Income from Capital (Washington: Brookings Institute, 1969), pp. 286–292.Google Scholar

²³ See Malinvaud, E., Statistical Methods of Econometrics (Chicago: Rand Mc-Nally, 1966), pp. 473–496Google Scholar, for an excellent discussion of this model of lagged behavior and its econometric problems. The specific technique employed to estimate emthese equations is that of Fair, R. C., “The Estimation of Simultaneous Equation Models with Lagged Endogenous Variables and Serially Correlated Errors,” Econotnetnca, XXXVIII (1970), 507–515.Google Scholar

²⁴ While the problem of multicollinearity allows only the most tentative statements on differences in lagged behavior, it seems useful to note that the lags in the excess supply equations tended to be somewhat longer than those in the demand relations and that lags in both relationships were longer in the years 1874–1894, than during 1895–1913. The latter movement suggests that both asset demanders and suppliers underwent a familiarization or learning process as the period progressed. Since the relative risk on American assets held in the U.K. probably fell across the period, it may also be that stock adjustment was more rapid because these assets were becoming more liquid. It would, however, be quite difficult to distinguish these two hypotheses for risk is, in part, a subjective judgment; thus familiarity may have led to lowered risk premiums and lowered risk premiums to greater familiarity.

²⁵ While the central focus of our model of U.K. investment in the U.S. is the long run or secular behavior of the market's participants,- it is useful to ask how well the model captures their actions subsequent to the period's financial panics. Given the sudden downturn of expectations on these occasions, it is reasonable to hypothesize that in the year immediately following a financial panic, the estimated demand and excess supply equations would tend to overestimate the level of U.S. securities held in the U.K. The financial panics of 1875, 1878, 1882, 1887, 1890, 1895 and 1907 either engulfed the London market or were so important to the New York market that it can be assumed that the American segment of the London market was substantially disturbed; see Morgenstern, O., International Financial Transactions and Business Cycles (Princeton: Princeton University Press-NBER, 1959), pp. 546–548.Google Scholar Inspection of the sign on the residuals from Table 1's equations in the year immediately following these panic years (nearly all of the panics occurred late in the calendar year) confirms the overestimation hypothesis, but it is important to note that the magnitude of the average residual in the post-panic year is quite small relative to the standard error of the regression. To the extent that our model effectively captures the secular tendencies of market participants, the latter finding suggests that the post-panic year did not involve large amounts of aberrant behavior and it seems reasonable to attribute this to the relative maturity of the London capital market. The relatively largest residuals tended to occur at major peaks and troughs of A, which as Morgenstern carefully notes were often not years of financial panics.

²⁶ Rates of price change in the U.S. and the U.K. were roughly similar and fairly small on an annual basis for both 1874–1894 and 1895–1913. It therefore seems plausible to assume expected rates of price change in the two countries were based on similar and lengthy past price histories and thus may be safely ignored.

²⁷ C. Feinstein, National Income, T96, T103.

²⁸ M. Edelstein, “Rigidity and Bias”; see also Payne, P. L., “The Emergence of the Large Scale Company in Great Britain, 1870–1914,” Economic History Review, 2nd Series, (1967), 519–542.CrossRefGoogle Scholar

²⁹ While there is some debate over the precise extent of the shift to the use of external long term financing instruments, there is little doubt concerning the fact of the shift. See Hall, A. R., “A Note on the English Capital Market as a Source of Funds for Home Investment before 1914,” Economica, XXIV (1957), 59–66CrossRefGoogle Scholar; A. K. Cairncross, “The Capital Market before 1914,” Ibid., XXV (1958), 142–146; A. R. Hall, “The English Capital Market before 1914—A Reply,” Ibid., 339–343.

³⁰ The average yield on first class U.S. railway ana U.K. industrial debentures listed on the London stock exchange, 1892–1913, were 3.94 percent and 4.06 percent, respectively; M. Edelstein, “The Rate of Return,” p. 190.

³¹ The effects of e/g+e and g/g+e are directly analogous to the involvement of the price elasticities in a model of a commodity demand and supply. The amount the equilibrium quantity increases when, for example, income increases and the demand curve shifts, depends on how much more elastic the supply curve is than the demand curve. If the own-price elasticities of the demand and supply curves are equal, the rise in the equilibrium quantity bought and sold is only half the relative rise in incomes. If g is very large relative to e, the relative rise in the equilibrium quantity is very close to the relative income rise.

³² Williamson's study of movement in the U.S. balance of payments is the only previous econometric work which deals directly with the push vs. pull problem in the U.S. context. A “reduced form” methodology is employed. In two separate regressions, one with U.K. and U.S. domestic investment as the independent variables and the other with U.K. and U.S. domestic stock prices, the U.K. variables accounted for most of the variance of net U.S. capital imports. Unfortunately, even though Williamson thought U.S. and U.K. financial market behavior was critical to the U.K.-U.S. flow, his use of purely domestic U.K. variables carries the implicit assumption that non-U.K., non-U.S. accumulations and returns were irrelevant to the U.K. decision to invest in the U.S. As noted in the text, this is a highly tenuous assumption. J; G. Williamson, American Growth, Ch. 4.

³³ M. Edelstein, “Rigidity and Bias,” pp. 85–95.

³⁴ M. J. Ulmer, Capital in Transportation, pp. 256–7.

³⁵ It is sometimes suggested that the industrial and geographic distribution of British wealth was sub-optimal during the years 1870–1913, because of certain dynamic institutional or technological inefficiencies connected with Britain's largely private processes of savings and investment. Since Britain was the major international long term lender during these years, it would be interesting to assess Britain's power to affect market interest rates and clearings, and to calculate the foregone income potentially available to compensate the nation for these suspected inefficiencies. Given that the own-interest elasticities of the U.S. excess supply and U.K. asset demand relations, g and e, respectively, were less than unity and roughly equal, 1895–1913, it follows that, all else held constant, a stamp tax on American assets held in Britain probably would have raised the interest income (tax included) paid by the American borrowers and that the burdens of the tax would have been shared equally by the Americans and private British lenders. Of course, “all else” most certainly would not have remained constant. It seems probable that a large portion of these direct gains would have disappeared due to, first, U.S. shifts to alternative European and domestic U.S. funding sources, and second, the effects of a likely diminished rate of U.S. railway capital formation on the price of British imports from the U.S. It is, of course, somewhat artificial to hypothesize a tax on only U.S. holdings when Britain's market power probably was greater elsewhere abroad. However, an attempt to assess the effects of a stamp tax on all overseas holdings would require an aggregative general equilibrium model specifying, among other relationships, Britain's aggregate savings behavior. A forthcoming study will explore this topic.