« Back to Results

Time Series and Empirical Macroeconomics and Finance

Paper Session

Friday, Jan. 3, 2020 12:30 PM - 2:15 PM (PDT)

Manchester Grand Hyatt, Old Town B
Hosted By: Chinese Economists Society
  • Chair: Jun Ma, Northeastern University

Reconstructing the Yield Curve

Yan Liu
,
Texas A&M University
Jing Cynthia Wu
,
University of Notre Dame and NBER

Abstract

The constant-maturity zero-coupon Treasury yield curve is one of the most studied datasets. We construct a new dataset with a non-parametric method. Our curve is globally smooth while still capturing important local variation. We show our dataset preserves information in the raw data and has much smaller pricing errors than existing benchmarks. We also provide how much information is in the raw data to complement our dataset.

Common Shocks in Stocks and Bonds

Anna Cieslak
,
Duke University
Hao Pang
,
Duke University

Abstract

Using joint variation in the stock market and the Treasury yield curve, we provide a new approach to identify shocks to investors’ expectations of monetary policy and economic growth as well as pure risk-premium shocks. Tracing out the effects of those shocks day-by-day, we explain the puzzling fact why stocks but not bonds earn high returns on Federal Open Market Committee (FOMC) announcement days and over the FOMC cycle. About 70% of the average positive stock returns earned over the FOMC cycle stems from the declining premiums, with the remaining 25% explained by accommodating monetary news, and only a small fraction by positive growth news. As bonds hedge growth risk in stocks, the signs of their responses to news are ambiguous: FOMC-induced reductions in the value of the bond insurance premium nearly completely offset any gains, making overall bond returns economically small. Since the mid-1980s, conventional monetary news accounts for about 40% of the variation in the two-year yield, but less than 10% and 20% of the variation in the ten-year yield and the aggregate stock market, respectively. The results suggest that the Fed has a significant effect on long-term assets through its ability to affect the risk premiums. (JEL: E43, E44, G12, G14)

GHH Preferences on Households' Portfolio Choices: Theoretical Implications and Empirical Evidence

Zongwu Cai
,
University of Kansas
Haiyong Liu
,
East Carolina University
Xuan Liu
,
East Carolina University

Abstract

This paper explores theoretical implications and empirical evidence of GHH preferences [Greenwood et al. (1988)] over portfolio choices. First, we analytically solve a parsimonious life-cycle portfolio choice model with the GHH preferences and endogenous labor-leisure choice and obtain a closed-form solution. Second, our analytical solution identifies four effects due to the GHH preferences (through endogenous labor-leisure choices) on risky shares; and it shows that two net effects hinge on the value of one key structural parameter. Third, we empirically test main theoretical predictions with the Panel Study of Income Dynamics data. Overall, the estimation results provide empirical evidence in support of GHH preferences. Thus, our analysis provides a definitive answer to one of the most fundamental, yet highly contentious, questions in quantitative macroeconomic analysis: the choice of utility functions in a representative agent model.

A New Approach to Multivariate Beveridge-Nelson Decomposition: The Case of Omitted or Unobservable Granger-Causing Variables

Chang-Jin Kim
,
University of Washington
Jun Ma
,
Northeastern University
Charles R. Nelson
,
University of Washington

Abstract

This paper shows that the typical VAR-based multivariate Beveridge-Nelson (1981) decomposition as studied in Evans and Reichlin (1994) may produce spuriously persistent cycles if either of the two implicit assumptions is violated: no irrelevant but highly persistent variable is included or the predictable component of the decomposed variable is perfectly correlated with the lags of the variables that Grange-cause the decomposed variable. In line with Pastor and Stambaugh (2009), we propose to model the predictable component as a latent variable that follows an ARMA process. Our framework nests the linear VAR as a special case and thus permits more general type of Granger-causality. We show that our proposed methodology produces more correct cycles regardless whether the two implicit assumptions underlying Evans and Reichlin’s theorem are violated or not.
JEL Classifications
  • C1 - Econometric and Statistical Methods and Methodology: General
  • G1 - General Financial Markets