This paper investigates the household demand for four financial assets: marketable bonds, time and savings deposits at commercial banks, life insurance reserves, and savings accounts at other financial institutions-credit unions, savings and loan associations, and mutual savings banks. The focus of the analysis is on the substitution relationships among liquid assets, and between these assets and marketable securities. The explanatory variables used in the study are: income, wealth, and the yields on various assets included in household portfolios.

A CCORDING to the expectations hypoth.t esis, the forward rates of interest implicit in the yield curve provide unbiased estimates of the market's expectations of future spot rates. Recently, the empirical validity of the expectations hypothesis has been tested by examining the conformity of movements in forward rates to the properties implied by the efficient market model.' latter model contends that, subject to the constraints imposed by the costs of obtaining and using information, expectations incorporate all publicly available information and prices (forward interest rates) always reflect these expectations. In this paper, the utilization of the efficient market model as a theoretical construct for testing explanations of the yield curve is extended in two directions. First, the long dormant question of the predictive accuracy of forward rates is reopened. Interest in this topic ebbed following the publication of Meiselman's work which presented a test of the expectations hypothesis that did not depend upon forecast accuracy for its statistical power. However, it appears that the analysis of forecast errors contains valuable information regarding the validity of the expectations hypothesis and the efficient market model.2 second point of departure from the existing literature is the performance of semistrong form tests of market efficiency. purpose of these tests is to determine whether the market employs all publicly available information in particular, estimates of income and liquidity variables in forming expectations of future rates. By comparison, previous investigations of the efficiency of the debt market have been limited to weak form tests which merely seek to determine whether the market correctly uses past interest rate data in forming expectations.3 Received for publication June 27, 1973. Revision accepted for publication June 10, 1974. * authors are indebted to Michael Adler, John Ciccolo, Jr., Stanley Diller, Lawrence Fisher, Kevin Hurley, Levis Kochin, Charles Nelson, Frank Schott, and Christopher Sims for helpful comments and criticism; and to Patricia Chick and Rona Stein for their excellent research assistance. Earlier versions of the paper were presented at the meetings of the American Statistical Association, Montreal, Canada, August 1972 and at seminars at Columbia University, Harvard University, the University of Chicago, and the University of Minnesota. Comments on those occasions also contributed to the paper. views expressed are those of the authors and do not necessarily reflect those of the individuals mentioned above, or the Federal Reserve Bank of New York. authors, of course, bear sole responsibility for all remaining errors. 1 See, for example, Richard Roll, Behavior of Interest Rates (New York: Basic Books, 1970), and Thomas J. Sargent, Rational Expectations and the Term Structure of Interest Journal of Money, Credit and Banking, IV, February 1972, pp. 74-97. Within the context of the expectations hypothesis, the efficient market model implies that forward rates of interest follow a martingale sequence. One important property of such sequences is that forward rates applicable to a particular point in time are not expected to change. original analytical work on the efficient market model is due to Samuelson and Mandelbrot. Paul A. Samuelson, Proof that Properly Anticipated Prices Fluctuate Randomly, Industrial Management Review, VI (Spring 1965), pp. 41-49 and Benoit Mandelbrot, Forecasts of Future Prices, Unbiased Markets and Martingale Models, Journal of Business, XXXIX (Special Supplement, January 1966), pp. 242-255. 2 Meiselman argued that the hypothesis would be corroborated if it were shown that expectations were revised systematically. David Meiselman, Term Structure of Interest Rates (Englewood Cliffs, New Jersey: PrenticeHall, 1962). Prior to the publication of this study the expectations hypothesis had been cast into disrepute because empirical researchers found that forward rates were inaccurate predictors of future spot rates. Among those who rejected the expectations hypothesis because forward rates had little predictive value were Hickman and Culbertson. By comparison, Macaulay-who found only limited evidence of predictive accuracy -argued that while successful forecasting might be rare, forecasting might still be attempted. See, W. Braddock Hickman, The Term Structure of Interest Rates: An Exploratory Analysis (New York: National Bureau of Economic Research, 1942), mimeo; John W. Culbertson, The Term Structure of Interest Quarterly Journal of Economics, November 1957, pp. 485-517; and Frederick R. Macaulay, Movement of Interest Rates, Bond Yields and Stock Prices in the United States Since 1859 (New York: National Bureau of Economic Resarch, 1938). 3 implications of this limitation in Sargent's work are discussed by Robert Shiller, Rational Expectations and the Term Structure of Interest Rates: A Comment, Journal of Money, Credit and Banking, V, August 1973, pp. 856-860.

the probability of an equal percentage change occurring is far smaller for R. In terms of equally probable changes, income has a greater impact on interest rates than unborrowed reserves. The partial correlation coefficients present less clear evidence as to the relative contributions of R. and Y. The partials are greater for R. than for Y in the equations for 1953-1960, but greater for Y than for R,2 in the 1961-1965 period. For the overall period, the partials are greater for RX when the equations are specified in the H-S form for 19531960 and for Y when the equations are specified in the H-S form for 1961-1965. Thus, the partial correlation coefficients also provide little evidence in support of H-S's conclusion that RX has a greater impact on interest rates than Y. , i the 1961-1965 period. For the over-

Michael J. Hamburger, Levis A. Kochin, Money and Stock Prices: The Channels of Influences, The Journal of Finance, Vol. 27, No. 2, Papers and Proceedings of the Thirtieth Annual Meeting of the American Finance Association, New Orleans, Louisiana, December 27-29, 1971 (May, 1972), pp. 231-249

Michael J. Hamburger, Gerald L. Thompson, Roman L. Weil, Jr., Computation of Expansion Rates for the Generalized von Neumann Model of an Expanding Economy, Econometrica, Vol. 35, No. 3/4 (Jul. - Oct., 1967), pp. 542-547