[downloadable]

Summary

• I construct a real-time consistent return forecasting factor of Cochrane and Piazzesi (2005,2008) and compare it to the original as well as variants that weight more heavily recent data. I do this to unlock a behavioral interpretation of foward interest rates - these are key financial market prices that other market participants use to form beliefs about future interest rate changes.
• Under this vantage point, I consider the importance of news effects on forecasts of bond excess returns. I also explicitly consider the central bank given that they are key macroeconomic actors that use their authority to alter current and future short-term interest rates as well as influence long-term rates.
• Because forecasters have imperfect information about future interest rates (and macroeconomic conditions), they form beliefs about them. Further, because of the inherent concerns about mis-specification, they re-estimate their model intermittently. Their beliefs and how they change over time are encoded into forward interest rates. Because of the inherent fragility of the expectations formation process under imperfect information constraints, recursive forecasts behave more cautiously; they react swiftly to large changes in realized excess returns but incrementally to marginal news.
• Further, I analyze the changes in expected bond excess returns over time in relation to hypothetical forecasts that implicitly assume an econometrician has access to all historically available information. These differences uncover the key concern that adaptive forecasts are anticipating, future short-term interest rate levels.
• I tie my analysis to how central banks conduct policy and set interest rates using the anticipation effect. I find empirical support that expectations about where long-term interest rates will eventually settle have been anchored. I also use my analysis to explain the inversion of the yield curve in 2004-2005 as an event consistent with an expected decline of future short-term and interest rate levels by market participants.
• More pertinent to my future work, I use this imperfect information apparatus to uncouple the direct link between bond risk prices and expected excess returns. I use the different return forecasting factors to understand the stochastic properties of currency excess returns.

Abstract

• I construct alternative forecast rules in relation to the return forecasting factor of Cochrane and Piazzesi (2005, 2008) to determine their relative performance. I compare forecasts assuming all historical data is available to recursively made ones that are revised with the arrival of news. Differences in the two forecast rules systematically move with realized bond excess returns. This anticipation effect also forecasts future short-rates and interest rate levels in government bond markets of other industrialized economies. I show that lower long-term rates relative to short-rates in 2004-2005 is consistent with an expected decline of interest rates by market participants. Finally, I discuss macroeconomic factors in relation to my analysis.

JEL Classification: C3, E3, E4, E5, E6, G1, O4

Keywords: adaptive learning; model misspecification; parameter uncertainty; return predictability