No abstract

This study highlights the importance of the information contained extreme value ratios (or returns) in the volatility estimation of financial assets. Most popular extreme value estimators like Parkinson (1980), Garman Klass (1980), Rogers Satchell (1991) and Yang Zhang (2000) use a subset of all available extreme value ratios but not the full set. We examine if there are other extreme value ratios which contain more information than the most widely used ratios. This study shows empirically how much information is contained in various extreme value ratios of financial assets, using both real and simulated data. Using information theory, we find out their variability in relation to a uniform distribution in each quarter. We then rank them using the Kullback-Leibler metric (in accordance with a scoring methodology we developed in this study) to ascertain which set of ratios are more variable than others and thus may provide better estimation in computing volatility. We also calculate the rank of the matrix to identify the set of linearly independent ratios, for ascertaining the number of ratios that would be enough to generate a class of volatility estimators. The empirical results demonstrate that the need for incorporating other ratios in volatility estimation. We also observe that each dataset has other more informative ratios which are uniquely attributed to that dataset.