A NOTE ON SKEWNESS AND DATA ERRORS

The observation that popular data bases have non-trivial errors is applied to the work of McEnally dealing with the return behavior of high risk common stocks. Carefully audited raw data is vital to the usefulness of research. Potentially erroneous data can be identified by examining extremely skewed series. Data series should be stratified according to both raw and relative skewness to identify the potential errors. Extreme observations have greater and greater impacts on higher and higher moments of distributions. Researchers doing "high moment" studies should be concerned about the quality of their data bases. They should decide on a study by study basis whether an analysis is more meaningful with or without legitimate outliers.