69th International Atlantic Economic Conference

March 24 - 27, 2010 | Prague, Czech Republic

Empirical Analysis of Dynamic Correlations Between Stock and Bond Returns

Saturday, 27 March 2010: 15:10
Thomas Chiang, Ph.D. , Finance, Drexel University, Philadelphia, PA
Jiandong Li, Ph.D. , Chinese Academy of Finance and Development, Central University of Finance and Economics (CUFE), Beijing, China
Empirical Analysis of the Determinants of Dynamic Correlations Between Stock and Bond Returns Over Time 
The relation between the stock and  bond markets is an important factor in asset allocation and risk management. The experience in the late 1990s suggests that the wealth effect and optimistic expectations may become dominant factors that encourage investors to hold both types of assets simultaneously.  The evidence indicates the correlations of these two asset returns are positive.  For instance, empirical studies by Keim and Stambaugh (1986), Campbell and Ammer (1993), and Kwan (1996) provide some supportive evidence. The literature also suggests a negative correlation between returns on the two assets. The phenomena of “flight to quality” (Hartmann et al., 2001; Hakkio and Keeton, 2009) and “flight from quality” lead to a negative correlation between stock returns and bond returns. Thus, the existing literature shows no consensus on the direction of the correlation. This paper investigates the correlation of returns between the U.S. stock and bond markets based on two prominent index funds provided by Vanguard: the Vanguard Total Bond Market Index Fund (VBMFX) and the Vanguard Total Stock Market Index Fund (VTSMX).

By employing both rolling correlation model and dynamic correlation coefficient models (the BEKK model and the asymmetric dynamic conditional correlation [ADCC] model (Engle, 2002)), we construct different measures of the time-varying correlation coefficients. Testing on the monthly data from 1996 through 2008, our finding suggests an average negative stock-bond correlation, although it is very negligible. The conditional correlation coefficients from the BEKK and ADCC methods are similar to the rolling window method using 22-trading-day observations. The Pearson’s correlation coefficients are greater than 0.90 among these three coefficient series.

The correlation coefficients between stock and bond markets depend on a few key measures of bond and stock uncertainty in addition to oil price volatility. The evidence concludes that stock-bond return correlations are negatively correlated with credit spread, default risk, implied volatility (fear), and oil price volatility, reflecting both a “flight to quality” and a “flight to liquidity.”