Issues in Hedge Fund Analysis : What a Difference a Day, Week, Month Makes

In this analysis, we simply take a step back and remind investors and researchers alike, that there is no simple answer to the dependency of empirical results on the data, period of analysis, or methods of quantitative analysis used to address issues of academic research. Over a common time frame and data source daily data is then used to create a series of weekly (five day), and twenty day return intervals which form the basis for a series of empirical comparisons. Results indicate similar time adjusted return and risk measures (standard deviation) determined from the use of daily, weekly, and twenty day time frames, however, results also indicate that the use of the various data interval impacts measures of beta estimation and auto correlation. Lastly, in contrast to using a common model approach (e.g., robust estimators) to reduce the impact of outliers on empirical analysis we simply show the results of removing a particular time frame (e.g. October 2008) on the empirical results. Results indicate that adjusting an extreme data point (a day (October 15), a week (October 13-18) or a month (October 2008) to a simple assumption of zero has major impact on beta estimation and autocorrelation results. Researchers may simply wish to use their own knowledge of data dependency to adjust data to reflect expected conditions rather than use models of general applicability which may offer results which are not truly reflective of conditions outside of that unique data period. In short, he results in this analysis indicate that some of the empirical results based on the use of monthly data may not be reflected in the use of weekly or daily data