| Summary: | 碩士 === 東吳大學 === 財務工程與精算數學系 === 104 === This thesis focuses on the relationships of investment proportions of standard ETFs and inverse ETFs among different assets, which are equities, bonds and REITs traded in U.S. securities exchange. The empirical model uses the multivariate BEKK-GARCH model proposed by Engle and Kroner (1995) to estimate the variations of the investment proportions of these assets in our sample period. In addition, to understand the effects of key variables on the weight of each asset in U.S. financial market, the interest rate of ten-year U.S. government bonds, the dollar index, the VIX index, and the MOVE Index are also considered into VAR models. Finally, the results from VAR model will help us to observe the correlations between investment weights and exogenous variables and to further analyze the impulse response and variance decomposition.
The empirical results indicate that the changes of the investment weights among three ETFs are affected by the other assets, and there are four lags correlations. In the exogenous variables, the investment weights of three ETFs are consistently affected by dollar index, the VIX index, and the MOVE Index, but the influences do not present the same directions. Moreover, the impulse responses show that the shock from bond ETFs market has significant and long-run influences to other ETFs markets. As to variance decomposition, when the variations of investment weights of equity ETFs and REITs ETFs can be effectively explained by bond ETFs. Therefore, the bond ETFs market plays an important role and it provides well explanations in allocations of ETFs assets.
|
|---|
