Editor's Note: This paper was presented at a joint session of the Economic History Association with the American Economic Association and the American Finance Association in San Francisco on December 29, 1974.

The authors are indebted to Professor Arcadius Kahan, Joseph Persky and Allen Sinai who made many helpful suggestions and Professor Richard Kosobud who made his Japanese data available. Computer time was supplied by the University of Illinois at Chicago Circle Computer Center. The authors are responsible for any errors or omissions.

¹ Gerschenkron, Alexander, Economic Backwardness in Historical Perspective (Cambridge: Harvard University Press, 1962).Google Scholar

² Neuburger, Hugh and Stokes, Houston H., “German Banks and German Growth, 1883–1913: An Empirical View,” Journal of Economic History, XXXIV (September 1974), 710–731CrossRefGoogle Scholar, hereafter called Neuburger-Stokes.

³ On this question see: Tsuru, Shigeto, Essays on Economic Development (Tokyo: Kinokuniya Bookstore Co., Ltd., 1968), p. 112.Google Scholar

⁴ See Sakurai, Kinichiro, Financial Aspects of Economic Development of Japan, 1868–1958 (Tokyo: The Science Council of Japan Division of Economics, Commerce and Business Administration Economic Series, #34, 1964), p. 63.Google Scholar Also see Lockwood, William W., The Economic Development of Japan (Princeton: Princeton University Press, 1968).Google Scholar

⁵ Ibid., p. 66.

⁶ The implications of this arrangement are clearly explained by Sakurai, Financial Aspects, p. 67, who comments that “although most of the [Zaibatsu bank] funds were used effectively from the point of economic development, the dominant motive in allocation decisions was more often the private interests of the Zaibatsu as a whole from the aspect of a holding company than the interest of the banks.”

⁷ Yamamura, Kozo, “Japan, 1868–1930: A Revised View,” in Banking and Economic Development, edited by Cameron, Rondo (New York: Oxford University Press, 1972).Google Scholar

⁸ Patrick, Hugh T., “Japan, 1868–1914,” in Banking in the Early Stages of Industrialization, edited by Cameron, Rondo (New York: Oxford University Press, 1967).Google Scholar

⁹ Ibid., pp. 283–284.

¹⁰ Gerschenkron, Economic Backwardness in Historical Perspective.

¹¹ Eistert, E., Die Beeinflussung des Wirtschaftswachstums in Deutschland von 1883 bis 1913 dutch das Bankensystem (Berlin: Duncker and Humblot, 1970), p. 91.Google Scholar

¹² See Binson, T. A., Zaibatsu Dissolution (Berkeley: University of California Press, 1954).Google Scholar

¹³ See Bank of Japan, Money and Banking in Japan (1964). Legal restrictions did prevent a bank from holding more than ten percent of the shares of a corporation, but these restrictions did not bar the resurgence of the Zaibatsu.

¹⁴ Schiffer, Hubert F., The Modem Japanese Banking System (New York: University Publishers, Inc., 1962), pp. 139–140.Google Scholar

¹⁵ Ibid.., p. 149.

¹⁶ For the unconstrained case the functional form actually estimated was In Y = In A + λ₁V₁ + λ₂V₂ + … + α₁ In + α₂ In K + In u for the constrained case

where L and K are labor and capital, Y is real output and V₁ is time and V₂ … V_n are various time periods of our shift parameter.

¹⁷ For a good description of the BLUS procedure see Principles of Econometrics by Theil, Henri (New York: John Wiley and Sons, 1971), Chapter 5.Google Scholar

¹⁸ This test involved dividing all the GLS residuals by the square root of the residual variance and testing whether the resulting vector was normally distributed.

¹⁹ The Kmenta test (see Neuburger-Stokes for a more extended discussion of the use of this test) has been used to test whether Cobb-Douglas was the correct specification. The result for the period 1952–1968 was

R²= .998; DW = 1.91; SEE where t-statistics are under the coefficients Since the coefficient of the squared term of the regression is not significant and is equal to 1/2 ργδ[l — δ] where ρ is a substitution parameter, γ is a scale parameter and δ a distribution coefficient and the coefficient or In (K/L) was significant and is defined as γδ, this implies that the elasticity of substitution σ (which is equal to 1/ (1 + ρ)) is unity. Since the elasticity of substitution is unity Cobb-Douglas is an appropriate specification. An extensive discussion of the data is given below.

²⁰ For a more extensive discussion of the data see Allen Sinai and Houston H. Stokes, “Real Money Balances and Production: The Japanese Case, 1952–1968,” which forms the basis of the appendix discussion of Kosobud's data and from which we quote: “There were some minor differences between he Christensen-Jorgenson and Kosobud approaches to the data. These were due primarily to a lack of adequate data. For example, excise and sales taxes were not subtracted from output in Kosobud's study. He did not add production subsidies to output but included as part of output government expenditures on transportation, power and communication. There was not as detailed a disaggregation of capital as in Christensen-Jorgenson. Some of the estimates, e.g., of rental prices and wages, required uncertain statistics. However, Kosobud appears to have followed Christensen-Jorgenson as closely as possible and his data are probably superior to any other for the post-war Japanese economy.”

²¹ A detailed discussion of data sources is found in the Appendix. Although we were able to get data on gross national enterprise national product, labor and capital services for the period 1952–1968 we were unable to obtain data for our index of the rental payments for the import to technology except for the period 1955–1967. It is for this reason that we have restricted our estimation to the period 1955–1967. If equation (1) and (2) were estimated without PT, then a significant positive coefficient for (MF/TO ) _t—1 would have an ambiguous interpretation since it would be difficult to distinguish between upward shifts of the production function due to the import of new technology and upward shifts of the production function due to changes in the financial structure or the way in which loans are made. If we had been able to adjust fully our input indexes to reflect the import of new technology, we would not have had to use PT explicitly in the production function. Such an adjustment was not possible. If PT and (MF/TO ) _t—1 can both be shown to enter the production function, then a more precise interpretation of the coefficient λ₃ is possible.

²² Kosobud also reported data for labor, capital and output (see Kosobud, page 43 manuscript) which he describes as “conventional” indexes. We have used this data source to estimate a constrained equation including PT, In (K/L) and (MF/TO)_t—1. The results for this equation were

R² = .996; SEE = .01944; DW = 2.0588; MVN = 1.358; F_h = 4.195; F_hProb = .90 It is important to note that (MF/TO)_t—1 entered the equation using a completely different data source. Although the “conventional” data have not been corrected for utilization and quality change and as a consequence are not as suitable as the “divisia” data that were used in the equations reported in Table 2, the results showing (MF/TO)_t—1 to be significant are additional evidence that our findings in Table 2 were not due to some pecularity of the “divisia” data that was used. When the “conventional” data were used in an equation of the form of (6) we found that time reduced the significance of In (K/L) but that (MF/TO)_t—1 continued to remain significant (coefficient 1.814, t-stat 4.027). When the “conventional data” were used in equations of the form of (3) and (4) (MF/TO)_t—1 remained positively significant (coefficient 2.167, t-stat 6.319 and 2.754, t-stat 3.831 respectively) with all variables significant and the right sign in all equations except for the fact that time was not significant in (4). We can conclude from these findings that the sign and significance of the coefficient of (MF/TO)_t—1 appears to be invariant to either various specifications of the functional form estimated or the data source used.