CONDITIONAL COVARIANCE MODELING AND APPLICATIONS IN MUTUAL FUND PERFORMANCE EVALUATION
MetadataShow full item record
This dissertation contains two essays. The first essay proposes a new model for conditional covariances based on predetermined information instruments. The model is based on a restricted Cholesky-like decomposition and ensures positive definiteness and invariance to variate order by construction. Comparing to existing time-series models, the new model provides a parsimonious and accurate description of second moments. The second essay develops a new conditional alpha model to investigate the risk-adjusted performance of mutual funds of different size as measured by assets under management. The model incorporates the conditional covariance methodology in the first essay and allows both time-varying risks and expected returns to rationally evolve with economic fundamentals. Using fund flow as an information instrument, we find that small mutual funds significantly outperform large funds and that these performance gains are most apparent following fund inflows. In contrast, larger funds display poorer performance after inflows. Our findings support the Berk and Green (2004) notion that small funds are more nimble in deploying optimal strategies than large funds. Results are consistent in both unconditional and conditional performance models.