The purpose of this study was to investigate the problem facing institutional investors when forming their dynamic asset allocation strategy. Allowing investors to rebalance their portfolios intertemporally can be seen as an expansion of their opportunity sets, thus making more interesting portfolio strategies available relative to static frameworks. However, to induce optimality of rebalancing it is essential that the investors regard their investment opportunity set as time-varying and thereby potentially predictable. To analyze the implications of such investment dynamics the first objective of the thesis was to develop a framework which allows for a tractable solution and simple empirical implementation. Knowing the familiar issue of extreme portfolio weights resulting from such models, the second objective was to present an extension facilitating applicability for practitioners. This was done by employing the Kalman filter to decrease the investor's confidence in the signals driving her allocations. The study took its theoretical grounding in the dynamic asset allocation model originally developed by Campbell et al. (2003), but applied the specific setup from Sørensen and Trolle (2010). As such, the model is based on an approximation of portfolio returns in order to yield a closed-form solution. To implement this framework, the investment dynamics were assumed to be described by a first-order vector autoregression, where the state variables were the result of a Top-down model selection analysis. The resulting portfolio strategy showed that the investor was aggressively timing the market making her portfolio rebalancing highly dependent on the state variable dynamics. In particular, excess stock returns were found to be predicted primarily by the dividend-price ratio, whereas the yield-spread was the main predictor of excess bond returns. Further, the study was able to confirm and explain the well-known horizon effect where investors tend to increase their stock holdings for longer investment horizons as well as the increasing stock allocation for more risk tolerant investors. Finally, the thesis showed the benefits of generalizing the simple model using the Kalman filter. The optimal portfolio strategy resulting from the extended setup had a greatly reduced allocation variability as the investor became more reluctant to adapt her market view to new information. Based on these findings the study concluded that dynamic asset allocation modeling is an interesting opportunity for institutional investors and that the generalized model presented herein can be seen as one potential way to make such models practically feasible.
|Educations||MSc in Advanced Economics and Finance, (Graduate Programme) Final Thesis|
|Number of pages||151|