26 Pages Posted: 16 Jul 2009
Date Written: July 1, 2009
Data analyses of daily closing VIX in 2008 reveal that: (1) the value of the lambda parameter, in the one-parameter Box and Cox (1964) family, appropriate for VIX, is around -.4 and is very far from one (no transformation) and zero (logarithm), the values typical of most of the literature, (2) the optimally transformed VIX follows an I(1) stochastic process, so that undifferenced VIX, transformed or not, does not have properties propitious for statistical analysis and (3) there are occasional large and unforecastable anomalous events in VIX data that, if ignored, can seriously distort any sort of statistical analysis. We achieve a linear, normal, homoskedastic representation of this data by transforming VIX in the B-C sense, modeling anomalous values explicitly and with a very simple IMA(1,1) model that is almost a random walk.
Keywords: time series analysis, volatility index, forecasting
JEL Classification: C32, G13, G15
Suggested Citation: Suggested Citation
Gonzalez-Perez, Maria T. and Guerrero, David E. and Treadway, Arthur B., The Daily Closing VIX Data for 2008 Reveal Unrecognized Properties (July 1, 2009). Available at SSRN: https://ssrn.com/abstract=1435020 or http://dx.doi.org/10.2139/ssrn.1435020