Abstract
Whereas the returns on financial assets can be assumed to be serially uncorrelated, although the issue of asset return predictability is still fiercely debated in the literature, see e.g. Campbell and MacKinlay (1999), it has become widely accepted that return volatility is highly autocorrelated. If the volatility process of financial asset returns were to be perfectly correlated, i.e. if volatility was constant over time, the variance of the underlying series could simply be calculated as
where y t is the continuously compounded return on the relevant asset at time t, T is the total number of observations and the sample mean \(m = \frac{1}{T}\sum\nolimits_{t = 1}^T {yt} \) A practical question that arises is what the length of the sample should be when volatility changes over time. Conventional wisdom would suggest that the sample should consist of a large number of observations as this would improve the statistical accuracy of the volatility estimates. On the other hand, it also implies that extreme shocks to the return process that took place a relatively long time ago, and which contain little information about the current volatility level, will still have a major impact since all observations in the sample are weighted equally. An alternative could be to select a smaller sample size and use a rolling window principle, where the sample size T remains constant and the variance is recalculated with each new observation that becomes available. The problem of extreme values remains however as the variance estimate of equation (2.1) will drop considerably in value when observations relating to this event fall out of the sample. Exponentially weighted methods have been proposed to remedy this problem but the reported high autocorrelation of the volatility process has led to the development of more intricate time-varying volatility models which incorporate this feature.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
We refer to these models collectively as GARCH models. For surveys on the extensive GARCH literature we refer to Bollerslev, Chou and Kroner (1992), Bera and Higgins (1993), Bollerslev, Engle and Nelson (1994), Diebold and Lopez (1995) and to the comprehensive selection of influential (G)ARCH papers in Engle (1995).
In order to ensure that the process is covariance stationary is set to be less than one. The positivity constraints have been questioned by Nelson and Cao (1992) who give necessary and sufficient conditions to ensure a non-negative variance which are less stringent. These weaker conditions are however mainly relevant for the higher order GARCH models.
The development of the GARCH model was primarily motivated by the fact that in empirical applications high orders of q were needed.
Nelson (1991) suggested that εt follow a Generalised Error Distribution (GED) which is a family of distributions including the normal distribution for v = 2 and fatter tailed distributions for v 2.
SV models are reviewed in, for example, Taylor (1994), Ghysels, Harvey and Renault (1996) and Shephard (1996).
See Hull and White (1987), Scott (1987), Wiggins (1987) and Chesney and Scott (1989).
If the correlation coefficient between ε t and ηt+1 is estimated and found to be negative, a leverage effect similar to that discussed with regard to the EGARCH model is observed. Also see the discussion and references in Harvey and Shephard (1996).
See, for example, Jacquier, Polson and Rossi (1994), Kim, Shephard and Chib (1998), Sandmann and Koopman (1998) and Fridman and Harris (1998).
See e.g. French, Schwert and Stambaugh (1987) and Akgiray (1989) who examined the US stock market and Poon and Taylor (1992) for results with regard to the UK stock market.
For the estimation of the SV model we use the exact maximum likelihood methods as described in Appendices A and B. Computing times for estimating SV models is usually higher than for estimating GARCH models which is due to the relatively more complicated methods required for estimating SV models.
See for example Baillie and Bollerslev (1989). For our data set this is confirmed by the insignificant Q(12) and Q(30) statistics for the monthly squared returns in table 2. 1.
For recent research on SV models with fat-tailed error distributions we refer to Liesenfeld and Jung (2000) and Jacquier, Polson and Rossi (2001).
We note that their SV model estimate of the volatility persistence term is equal to 0.91 which explains the rapid exponential decay of the ACF (see figure 1); the high implied first-order autocorrelation coefficient for the SV model is then mainly attributable to the relatively high estimate for the variance of the log volatility process. Also see the discussion in Ghysels et al. (1996).
Also see the discussion in Ghysels et al. (1996) who show that ρs ”is at maximum for a normal distribution”.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2003 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Hol, E.M.J.H. (2003). Asset Return Volatility Models. In: Empirical Studies on Volatility in International Stock Markets. Dynamic Modeling and Econometrics in Economics and Finance, vol 6. Springer, Boston, MA. https://doi.org/10.1007/978-1-4757-5129-1_2
Download citation
DOI: https://doi.org/10.1007/978-1-4757-5129-1_2
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4419-5375-9
Online ISBN: 978-1-4757-5129-1
eBook Packages: Springer Book Archive