Why does the correlation between stock and bond returns vary over time?

Abstract

This article examines the impact of inflation and economic growth expectations and perceived stock market uncertainty on the time-varying correlation between stock and bond returns. The results indicate that stock and bond prices move in the same direction during periods of high inflation expectations, while epochs of negative stock–bond return correlation seem to coincide with subdued inflation expectations. Furthermore, consistent with the ‘flight-to-quality’ phenomenon, the results suggest that periods of elevated stock market uncertainty lead to a decoupling between stock and bond prices. Finally, it is found that the stock–bond return correlation is virtually unaffected by economic growth expectations.

Acknowledgements

We are grateful to an anonymous referee, Seppo Pynnönen, Paul Söderlind, Frank K. Reilly, Manfred Kremer, Robert Driver, Stanley Martin, Kenichi Ueda, seminar participants at the European Central Bank and conference participants at the 2005 Midwest Finance Association Meeting, the 2005 Southwestern Finance Association Meeting and the 9th International Conference on Macroeconomic Analysis and International Finance for helpful discussions and comments. The views expressed in this article are those of the authors and should not be interpreted as reflecting the views of the European Central Bank.

Notes

¹ Stock prices should, in theory, equal the discounted sum of all expected future dividends. The discount rate consists of two components, a premium that investors demand for holding risky assets and the risk-free rate, which is usually approximated by the yield on government bonds. Thus, if expected future dividends and the equity risk premium remain unchanged, higher risk-free rate will cause downward pressure on stock prices and thereby result in a positive correlation between bond and stock returns. Empirical evidence on the common discount rate effect is provided in Gulley and Sultan (2003).

² Due to data availability, we have excluded Japan from the analysis. However, although Japan is among the three largest stock markets in terms of market capitalization and trading activity, the bond markets in the US, UK and Germany are substantially larger in terms of the amounts of outstanding debt.

³ The analysis was also conducted using 2-year government bond price indices. However, since the stock–bond return correlations are virtually similar regardless of the maturity of the bonds, we only report results based on 10-year bonds.

⁴ The sample periods for the UK and Germany begin in January 1992 due to data availability.

⁵ For additional details on implied volatility indices, see e.g. Fleming et al . (1995), Blair et al . (2001) and Graham et al . (2003).

⁶ DCC modelling has previously been used, e.g. by Bautista (2003), Cappiello et al . (2006), Lee (2006) and Manera et al . (2006).

⁷ Virtually similar correlation estimates were obtained when 2-year government bond indices were used instead of 10-year bonds.

⁸ Investment strategies and asset allocation decisions that assume a constant relationship between stock and bond returns are obviously inappropriate and may lead to considerable losses during sudden changes in correlation structures. Moreover, commonly used risk monitoring techniques may result in spurious conclusions if the dynamics of stock–bond correlations are neglected.

⁹ The link between economic activity and the real interest rate dates back to Fisher (1907), who showed that the real interest rate is determined by the ratio of optimal future consumption to optimal current consumption. This ratio, including the discount factor adjustment, is the marginal rate of inter-temporal substitution reflecting agents’ preferences and the presence of the discount factor ensures that the real rate of interest exceeds real consumption growth in the long run.

¹⁰ Given that the choice of the GDP variable may affect the results regarding the relationship between economic growth expectations and stock–bond correlations, we examined the robustness of our regression results by using the slope of the term structure as a proxy for economic growth expectations. For the US and UK, the choice of the economic growth variable does not make any difference, as the coefficient estimates for the slope of the term structure are statistically insignificant. For Germany, however, we find evidence for a positive relationship between stock–bond correlations and the slope of the term structure.