Treasury return predictability and investor sentiment

• Published date: 01 December 2023
• Author: Chen Gu,Xu Guo,Ruwan Adikaram,Kam C. Chan,Jing Lu
• Date: 01 December 2023
• DOI: http://doi.org/10.1111/jfir.12342

Received: 1 September 2022

|

Accepted: 22 May 2023

DOI: 10.1111/jfir.12342

ORIGINAL ARTICLE

Treasury return predictability and investor

sentiment

Chen Gu

1

|Xu Guo

2

|Ruwan Adikaram

3

|

Kam C. Chan

1

|Jing Lu

4

1

Research Center of Finance, Shanghai

Business School, Shanghai, China

2

Department of Risk Management and

Insurance, College of Economics, Shenzhen

University, Shenzhen, China

3

Department of Accounting and Finance,

University of Minnesota Duluth, Duluth,

Minnesota, USA

4

School of Economics and Business,

Chongqing University, Chongqing, China

Correspondence

Xu Guo, Department of Risk Management

and Insurance, College of Economics, 1310

Mingde Building, Shenzhen University,

Shenzhen, China.

Email: xuguo@szu.edu.cn

Funding information

National Natural Science Foundation of

China, Grant/Award Numbers: 71973018,

72201174, 91846108; Ministry of Education

of Humanities and Social Science Project,

Grant/Award Number: 22YJC790031

Abstract

We document that the Treasurymarket investor sentiment

(TSENT) of institutional investors is a powerful predictor of

bond risk premia. Specifically, TSENT positively predicts

Treasury bond excess returns in and out of sample. The

forecasting gains of TSENT are incremental to those in

conventional bond return predictors: Fama–Bliss forward

spreads, Cochrane–Piazzesi forward rate factor, and

Ludvigson–Ng macro factor, as well as equity market

sentiment proxies such as the investor sentiment index and

the partial least squares sentiment index. Asset allocation

analysis indicates the forecasting power of TSENT is

economically valuable to investors. Finally, we show that

the time‐series bond risk premia predictability associated

with TSENT relatesto itspredictive powerfor macro-

economic performance, such as payroll employment,

unemployment rate, and industrial production.

JEL CLASSIFICATION

G11, G12, G17

1|INTRODUCTION

Treasury bonds play a central role in investors' asset allocation. Their risk and return dynamics have received

considerable attention from financial researchers. An extensive literature demonstrates the predictability of

Treasury bonds. For instance, predictors based on forward spreads (Fama & Bliss, 1987), yield spreads (Campbell &

Shiller, 1991), a linear combination of forward rates (Cochrane & Piazzesi, 2005), and macro factors (Ludvigson &

J Financ Res. 2023;46:905–924. wileyonlinelibrary.com/journal/JFIR

|

905

© 2023 The Southern Finance Association and the Southwestern Finance Association.

Ng, 2009) possess significant in‐sample predictive power for Treasury bond excess returns with maturities of

2–5 years. Gargano et al. (2019) findboth statistically andeconomically significant Treasury bondreturn

predictability after accounting for estimation error and model uncertainty. In general, ample studies highlight the

role of financial and macro factors in predicting bond premia.

Behavioral predictors, such as marketwide investor sentiment, also hold predictive power for aggregate assets

returns (e.g., Baker & Wurgler, 2007; Baker et al., 2012; Doukas & Milonas, 2004; Huang et al., 2015). Recent

research focuses mainly on investor sentiment effects in the stock market and posits that investor sentiment affects

financial markets by driving asset prices away from fundamentals (De Long et al., 1990; Shefrin & Statman, 1994).

The subsequent price reversals lead to a negative relation between investor sentiment and future returns (Brown &

Cliff, 2004; Garcia, 2013; Jiang et al., 2019; Schmeling, 2009). However, several recent studies document that

marketwide investor sentiment not only contains irrational factors but also includes information about

fundamentals and possesses some positive predictive power about stock returns (e.g., Bollen et al., 2011; Gao

et al., 2021).

Although these two concepts are prominent in their academic fields, there is little research that explores the

role of investor sentiment in predicting bond premia in a unified framework. Consequently, several research

questions remain unanswered: Can the investor sentiment that pertains to theTreasury market forecast the future

excess returns of government bonds? Is the forecasting power of Treasury sentiment beyond that contained in

conventional sovereignbond returnpredictors? Areforecasting gainsfrom the Treasurysentiment economically

valuable to investors? What drives the predictive power of Treasury sentiment?

We contribute to the literature by addressing all of these questions. Using the Treasury market sentiment

(TSENT) of institutional investors provided by SENTIX (a leading behavioral finance research institution in market

sentiment data),we find that TSENTpositively predicts1‐month‐ahead Treasury bond excess returns in sample.

Specifically, a 1 SD increase in TSENT leads to 9.1, 12.2, 14.3, and 15.6 basis point (bps) increases for 2‐,3‐,4‐, and

5‐year Treasury returns in the subsequent month, respectively. We also find that TSENT's explanatory power

decreases with bond maturity. The univariate predictive model R2decreases from 4.721% to 1.531% as bond

maturity increases from 2 to 5 years. The results are qualitatively similar when using excess bond returns based on

the Gürkaynak et al. (2007) method or Fama Treasury bond portfolio returns from the Center for Research in

Security Prices(CRSP) database.The positive relationbetween TSENT andfutureTreasuryreturns isrobust after

adopting bootstrapping exercises and the weighted least squares method of Yohai (1987).

The positive predictive power of TSENT indicates that TSENT contains valuableinformation aboutthe

future price movement of Treasury bonds, which has not been incorporated into current prices. This finding is

different from that documented in most of the stock market sentiment literature, which argues that investor

sentiment is irrational, and its negative impact on future stock index returns is due to stock price reversal after

high sentiment. The contradicting implications may arise for two reasons. First, there exists segmentation

between the pricing of debt and equity (Geanakoplos, 2010;Simsek,2013) and sentiment may have distinct

implications. For instance, López‐Salido et al. (2017)pointoutthatcreditmarketsentimentisfundamentally

different from stock market sentiment. They show that only high credit market sentiment forecasts

macroeconomic slowdown. Therefore, unlike stock market sentiment that reflects irrational behaviors among

investors, itis possible that TSENT containsfundamental‐related information. Second, we obtain our Treasury

sentiment measure from sophisticated institutional investors, whereas most equity studies use sentiment

measures from retail investors. It is well recognized that institutional investors have an information advantage

(Gao et al., 2021; Hendershott et al., 2015) because they are more likely to have a cohesive level of

professional knowledge, rich resources at their disposal, and an abundant amount of time available to perform

intensive analysis.

However, the forecasting power of Treasury sentiment could be due to the incorporation of other relevant

information knownto predict Treasury bondexcess returns. Toaddress thisconcern, we compareour Treasury

sentiment with a list of conventional bond return predictors, such as Fama and Bliss (1987) forward spreads (FB),

906

|

JOURNAL OF FINANCIAL RESEARCH