This paper examines the pricing of exchange-traded long-term corporate bond portfolios. Observable instruments measuring the term structure of interest rates, levels of bond and stock prices, and a January dummy are found to predict excess returns on corporate bonds. An intertemporal asset pricing model with changing expectations and unobservable factors is then estimated for the predictable excess returns using Hansen's Generalized Method of Moments. The results show that a multibeta linear time-varying model of con? ditional expected returns with constant betas can successfully value corporate bonds. Spe? cifically, the tests indicate the presence of two time-varying hedge portfolios. The data, however, support a single latent variable specification when all January observations are excluded. This result suggests the existence of a strong January seasonal in one of the latent variables.

Despite nonstationarities in the factor betas and factor prices of the Chen, Roll, Ross (1986) multifactor model, investors are rewarded for bearing risks associated with the change in expected inflation and industrial production in non-January months; however, variations in these factors have opposite influences on stock prices. These findings may partially explain why several recent studies fail to detect a significant non-January risk premium in the stock market, but this evidence is only suggestive since theoretical and statistical difficulties prevent precise interpretations of specific pricing relations in the Chen, Roll, Ross model.

Results of tests contrasting tax-loss selling with intertemporal information variation as explanations of the January seasonal in stock returns are reported. Closed-end fund shares display the typical size-related January seasonal while their net asset values do not. Interpreting the net asset value return as a proxy for information about underlying assets, this result indicates information variation is not a necessary condition for the January effect in stocks. The share returns at the turn of the year are negatively related to their mean preceding year returns and positively related to the standard deviations of their preceding year returns. These results are consistent with tax-loss selling.

[Results of tests contrasting tax-loss selling with intertemporal information variation as explanations of the January seasonal in stock returns are reported. Closed-end fund shares display the typical size-related January seasonal while their net asset values do not. Interpreting the net asset value return as a proxy for information about underlying assets, this result indicates information variation is not a necessary condition for the January effect in stocks. The share returns at the turn of the year are negatively related to their mean preceding year returns and positively related to the standard deviations of their preceding year returns. These results are consistent with tax-loss selling.]