The US zero-coupon yield spread as a predictor of excess daily stock market volatility

Abstract

Slope of the yield curve has often been cited as an indicator of economic activity. Based on this premise, we extend the study to examine whether one can use the US zero-coupon yield spread to predict excess stock market volatility of three international stock markets – the United States, the United Kingdom and Hong Kong. By using daily trading data and changes in the US yield spread, our study entails four spread maturity spectrums, three stock markets, 7-trading day forecast horizons and four probit models. In the static models, we find evidence to support a US spread–volatility relationship in all three stock markets in the medium spread maturity up to 3 days ahead. In the dynamic models, although the predictive power of yield spread weakens slightly, the lagged dependent variable plays an important role.

Keywords:

• probit model
• predictive power
• zero-coupon yield spread
• stock market volatility

JEL Classification:

• C22
• E44
• F30
• G15

Acknowledgement

I would like to thank an anonymous referee for very helpful comments. All errors and omissions remain the responsibility of the author.

Notes

¹  A good survey of these term spread and economic activities can be found in Wheelock and Whohar (2009).

²  With reference to Zhou (1996) and Fama and Schwert (1977), we believe that by exploring daily observations in various stock markets, it serves as an extension of their contributions to the study of interest rate movements and stock market fluctuations.

³  Difference between 3-month LIBOR and 3-month Treasury bills.

⁴  Also known as the Gordon growth model, which was originally published in Gordon and Shapiro (1956).

⁵  Of course, we exclude the possibility of using the indexed bonds that are issued by the UK government to raise the demand for government bonds.

⁶  Spot yield comes from zero-coupon bonds with redemption price fixed and known at time of issue.

⁷  For maturities beyond 1 year, spot yields can be calculated from coupon paying Treasury securities, see McCulloch (1990).

⁸  The data were originally provided by Gurkaynak et al. (2006).

⁹  This stock market volatility measure is used in Schwert (2002). The 253/20 coefficient is applied to annualize our calculated daily volatilities. The magnitude of our stock volatilities is comparable to those used in the majority of Schwert (1989, 1990, 1998, 2002) papers on stock market volatility.

¹⁰  The use of second moments of returns to study stock return volatility dates back to the work of Officer (1973), Merton (1980), French et al. (1987) and Schwert (1989). More recent applications can be found in work by Andersen et al. (2003, 2005) and others.

¹¹  USTED is the difference between the US 3-month interbank rate and the 3-month Treasury bill secondary market rate, and UKTED is the difference between the UK 3-month interbank rate and the 3-month Treasury bill yield, whereas HKTED measures the difference between the Hong Kong 3-month interbank rate and the 3-month Hong Kong exchange fund bill yield.

¹²  In both static and dynamic models, we begin our yield spread study by using the level of yield spread as our independent variable. Unfortunately, regression results fail to obtain statistically significant estimated coefficient and produce extremely low pseudo-R². That is, yield spread measured at this level provides no predictive information about excess stock market volatility. Hence, in this article, we focus our attention on changes of yield spread.

¹³  A statistically significant negative estimated coefficient can be interpreted as excess volatility being less likely to occur.

¹⁴  See Chou (1988), Engle and Patton (2001).

¹⁵  3M6M (4, 5, 6 days ahead) and 3M2Y (7 days ahead) (S1 and S2).

¹⁶  Percentage in terms of the full sample size.

¹⁷  To maintain our focus on the predictive power of the changes in spread, only results from the single-variable estimations of S1 and D1 are discussed.