Return and volatility spillovers between Dubai financial market and Abu Dhabi Stock Exchange in the UAE

Abstract

This article investigates return and volatility spillover effects between Dubai Financial Market (DFM) and Abu Dhabi Stock Exchange (ADSE) using two methodologies: A simple asymmetric Vector Autoregressive-Baba, Engle, Kraft, Kroner (VAR-BEKK) framework introduced by Kroner and Ng (1998), and an asymmetric version of the Dynamic Conditional Correlation (DCC) model proposed by Engle (2002). We find that return and volatility transmission mechanisms between DFM and ADSE in the UAE are asymmetric. In particular, there are significant spillover effects in both returns and volatility from DFM to ADSE. The DFM is playing the dominant role, and the feedback effect from ADSE to DFM is relatively weak, albeit significant. These results are consistent with an exchange market in which information is first incorporated into the DFM before being impounded into the ADSE.

Keywords:

• UAE stock markets
• volatility spillover
• dynamic conditional correlation
• VAR-BEKK framework

JEL Classification::

• G10
• O53

Notes

¹ See, for example, Cifarelli and Paladino (2000), Sheng and Tu (2000), Hashimi and Xingyun (2001), Jang and Sul (2002), Ratanapakorna and Sharma (2002), Yang et al. (2002), Higgs and Worthington (2004), Click and Plummer (2005), Kim (2005), Tsouma (2007), Lucey and Voronkova (2008), Ahlgren and Antell (2010), Marcal et al. (2011), Singh et al. (2010), Yilmaz (2010) and Tan et al. (2012).

² Studies like Bracker et al. (1999), Pretorius (2002), Johnson and Soenen (2003) and Nam et al. (2008) have focused on the factors that are likely to cause pronounced spillovers.

³ However, one may argue that using daily prices will lead to synchronization problem due to different trading hours of the markets. This problem may lead to a downward bias in the estimated correlations (Martens and Poon, 2001). This issue should not impact the nature of return dynamics and volatility spillover among the UAE markets, since these markets are regarded as having same synchronous trading. Nevertheless, we have also performed the analysis using weekly data and we got similar results (results are not reported in this article and available from the authors upon request).

⁴ Here, r_{i , t} for i = (1, 2) is the stock return index in each market at time t. The conditional distribution of ε_t is assumed to follow Student-t distribution with E_{t −1}(ε_t ) = 0, E_{t −1}(ε_t ε_t ) = H_t , and the degrees of freedom ν, where 2 < v < ∞. The variance–covariance matrix of unexpected returns is denoted as H_t  = [h_{i ,  j , t} ], and it is 2×2 square matrix.

⁵ Own asymmetry refers to the stylized fact that conditional volatility is higher after a negative return shock than after a positive one.

⁶ None of the multivariate GARCH models is able to capture the documented stylized fact that the correlation between markets tends to increase after volatility increases.

⁷ The table is arranged as follows: the conditional mean estimates are reported in Panel A, the conditional variance in Panel B and the diagnostics in Panel C.

⁸ In Panel B, the value of b ₂₁, which represents the sensitivity of ADSE's volatility to the DFM, is equal 0.136. The opposite direction parameter b ₁₂ equals only 0.035. The parameteris also less signifcant.

⁹ The asymmetric impact of negative news on dynamic correlations means that correlations are higher in the case of joint bad news, than in the case of joint positive news of the same magnitude.

¹⁰ The DFM is a significant player in the country news-generating process. The magnitude of the parameters is [μ ₂₁(0.172)>μ ₁₂(0.107), for BEKK and μ ₂₁(0.169)>μ ₂₁(0.116) for AG-DCC].

¹¹ The countries are Egypt, Israel, Jordan, Lebanon, Morocco, Turkey and Oman.