Financial liberalization, structural breaks and stock market volatility: evidence from South Africa

Abstract

This article examines the effect of financial liberalization on South African equity markets using Exponential Generalized Autoregressive Conditional Heteroscedastic (EGARCH) models. It utilises daily data and specifically, it analyses whether volatility persistence increased following financial liberalization. To achieve this objective, the study starts with endogenous structural break tests using Bai and Perron (2003) Ordinary Least Square (OLS)‐type test and the Cumulative Sum (CUSUM)‐type test of Inclan and Tiao (1994) and Sanso et al. (2004) respectively. These breaks are performed both in the stock returns and in the conditional variance over pre‐ and post‐liberalization periods. The significant break points identified through algorithm are incorporated into EGARCH models and to obtain the effect of financial liberalization, the study further adds liberalization dummy using official liberalization dates. The findings show that none of the estimated break dates coincide with the official liberalization dates. The analysis further shows that after taking structural breaks into account volatility declines following financial liberalization. Also using official liberalization dates, the results indicate that the effect of financial liberalization on the stock markets is negative and statistically significant.

Keywords:

• financial liberalization
• structural breaks
• market volatility
• exponential GARCH

JEL Classification::

• C32
• G15

Acknowledgements

I am indebted to Professor Stephen G. Hall for his invaluable comments and incisive suggestions. Many helpful comments and suggestions from the anonymous referees were highly appreciated. I would also like to thank Mikail Karoglou for providing me with the Gauss codes for the Inclan and Tiao (1994) test. Any remaining errors are my own responsibility.

Notes

¹ This can be found in the work of DeSantis and Imrohoroglu (1997), Aggarwal et al. (1999), Edwards et al. (2003), Nguyen (2008), Wang and Theobald (2008), and James and Karoglou (2010) among others.

² MDH states that changes in stock prices and volume are driven by the same information arrival rate to the market. In other words, the MDH simply measures the impact of news arrival on the serially correlated mixture of prices and volume of stock returns.

³ The reforms of 1995 were termed the Big Bang because they brought about price competition, corporate membership and dual‐capacity trading.

⁴ It should however be noted that Bai and Perron (2003) usually gives lower number of breaks than those of the ICSS algorithm. Direct comparison may not be possible because, for example, Bai and Perron's method set some constraint on the number of potential breakpoints. The algorithm assumes that there exists a certain distance ‘q’ between points.

⁵ The Inclan and Tiao (1994) ICSS algorithm gives several breaks as indicated in the work of Aggarwal et al. (1999). This according to Andreou and Ghysels (2002) is due to some size distortions.

⁶ The study after the estimates uses the Bai and Perron (2003) OLS‐type only for the conditional variance breaks as the Inclan and Tiao (1994) and Sanso et al. (2004) CUSUM‐type tests gave so many breaks.