Evaluating performance of institutional mutual funds using kernel density estimation

The purpose of this dissertation is to evaluate the performance of U.S. Institutional Mutual Funds using Kernel Density Estimation. Although there is substantial amount of literature on mutual fund performance using unconditional and conditional models, nonparametric methods in general and kernel density estimation in particular have not been used widely. Further, Institutional Mutual Funds have not been considered as a separate category even though they are different from other mutual funds in several respects that could affect their respective performance.Institutional Mutual Funds are evaluated in nine different styles based on Morningstar's style box. The kernel densities of all funds in a style category and the appropriate benchmark were estimated based on 5 years worth of return data calculated from weekly historical prices of funds and the benchmark provided by Morningstar using a specially written Matlab code. Areas under the curve were calculated corresponding to selected returns in the left and right tails which were then used for regression analysis. For each style category, kernel density estimates identified the funds with abnormal performance. This approach can also be used to compare fund performances across style categories and to compare performances of IMFs to those of regular mutual funds.