ABSTRACT Beliefs of professional forecasters are benchmarked against those of a Bayesian econometrician who is learning about the unknown dynamics of the bond risk factors. Consistent with rational Bayesian learning, the forecast errors of individual professionals and are comparably predictable over the business cycle. The secular and cyclical patterns of professionals' forecasts relative to those of are explored in depth. Inconsistent with many models with belief dispersion, the relationship between professionals' yield disagreement and their matched disagreements about macroeconomic fundamentals is very weak.

Kenneth J. Singleton, Adjustment Costs and Capital Asset Pricing: Discussion, The Journal of Finance, Vol. 40, No. 3, Papers and Proceedings of the Forty-Third Annual Meeting American Finance Association, Dallas, Texas, December 28-30, 1984 (Jul., 1985), pp. 705-709

Variance bounds are derived for general present-value relations involving the expected future values of any finite number of variables. The estimators of these bounds and the variance being bounded are then shown to have a joint distribution converging to that of the multivariate normal, with moments which can be consistently estimated from the data. As a special case of these results, it is shown that expectations models of the term structure imply upper and lower bounds on the variance of the long-term rate. These bounds are used to test a rational expectations model of long-term U.S. Treasury bond yields.

The term structure relations implied by a model in which preferences are non-separable functions of the service flows from two goods are investigated. The parameters characterizing preferences are estimated and restrictions on the co-movements of consumptions and Treasury bill returns are examined. Both the durability of goods and the non-separability of preferences are important factors in explaining the time paths of individual returns, but there is substantial evidence against the cross-sectional restrictions implied by our model. Differences between sample mean returns are too large relative to the sample covariances of the return differences and the marginal utility of consumption.