Date of Degree
Finance and Financial Management | Macroeconomics
Yield curve, term premium, New Keynesian
This dissertation studies the modeling of U.S. Treasury (UST) yield curve term premia under the New Keynesian (NK) framework. Loosely speaking, term premium is the difference between a government bond’s yield for a specific tenor and the average of the expected short rates up to that tenor. The dissertation is divided into three chapters. The first chapter proposes a New Keynesianism-based macro-finance model estimated by a one-step full information maximum likelihood (FIML) method. The second chapter shows that the one-step FIML method may produce estimation biases, which result in biased expected short rates and term premia. The chapter then presents an alternative estimation strategy. The third chapter addresses the policy rate’s zero lower bond (ZLB) constraint in the NK model by including a shadow rate concept.
The first chapter fills a gap in the macro-finance term premium modeling literature by building a two-way feedback loop between the economy and the yield curve in a micro-founded way. In doing so, the chapter incorporates a latent Financial Risk Index (FRI) in the IS curve and Taylor rule of an NK model with consumption habit formation. Using the Affine Term Structure (ATS) finance theory, it fits the model to macroeconomic and yield data to obtain time-varying term premia. The chapter also replaces the FRI with the UST three-month vs. 10-year yield slope in the NK system to form Model 2, which offers the central bank and financial market participants an observable market variable to monitor and to communicate with and which thus builds the two-way feedback loop. The two models are both estimated by a one-step FIML method, in which the reduced-form vector autoregression of order 1, or VAR(1), coefficients and the structural NK parameters are estimated simultaneously.
The second chapter reveals that the one-step FIML method employed in the first chapter (and in the class of NK term premium models) may produce negative bias to the reduced-form VAR(1) coefficients, which in turn result in a too stable estimated 10-year average expected short rate series and 10-year term premium whose variations track those of the 10-year yield too closely. The chapter presents a two-step estimation strategy. The first step estimates the reduced-form VAR(1) model using ordinary least squares (OLS) and adjusts the negative small-sample estimation bias to the coefficients. The second step recovers structural NK parameters. The chapter proposes a structural restriction to the IS curve and thus improves the model fit to the data. The new estimation method produces a more cyclical and structural 10-year average expected short rate and a more counter-cyclical 10-year term premium than the first chapter. The method also restores consistency between the NK system and the reduced-form VAR system.
The third chapter addresses the ZLB issue by bringing in the Wu-Zhang Shadow Rate New Keynesian (SR-NK) model (Wu and Zhang 2016) into the first chapter’s macro-finance term premium modeling approach. The chapter points out that a connection between the Wu-Zhang SR-NK model and the yield curve cannot possibly be established and that there is a tenor mismatch between the short rate and the Wu-Xia shadow rate (Wu and Xia 2016). The chapter proposes a new SR-NK model that inherits the NK model of the first chapter with the short rate replaced by the latent shadow rate. The new model assigns the shadow rate and the FRI different roles of yield curve level and slope drivers. It also dedicates the shadow rate to capture the effect of the Federal Reserve (Fed)’s forward guidance and the FRI to capture the effect of quantitative easing (QE). Thus, my SR-NK model addresses the two issues of the Wu-Zhang SR-NK model. The chapter then proposes a simple way to replace the latent shadow rate by adjusting the short rate during the ZLB period using the variations of the one- and two-year yields to construct Model 2. The adjusted short rate is shown to reach negative levels similar to the negative rates adopted by other central banks. This adjustment constructs a ZLB constraint-free NK model without latent variables. It avoids the imputed latent variables’ sensitivity to parameter values and preserves the two-way feedback loop between the central bank and market participants.
Fu, Weiguo, "Essays on New Keynesian Term Premium Model with Financial Risks" (2019). CUNY Academic Works.