Asset Allocation Strategies Based On Black-Litterman Model

As the supply of financial products and the demand for effective asset allocation increase,how to diversify one's assets has become an important issue between individuals and institutions.It is shown that the variation in investment performance can be accounted for 90%by asset allocation and selection.Since the portfolio optimization theory was proposed by Markowitz,the model based on Bayesian technique introduced by Black and Litterman has received much focus.The author utilizes Black-Litterman model proposed by Black and Litterman in 1992,studies the asset allocation strategies using data from 2010 to 2016.There are some extensions in this paper:1,it summarizes the ways of inputting weights in reverse optimization,these inputs are generated by EGP model proposed by Elton(1976)et al.2,the paper points out a mistake made by Shiqiang Zhang(2008)when he was setting variances in his thesis.3,this paper expands the duration of back test,uses extra statistics to judge the stability of strategies.Moreover,views and its variances in this paper are generated by GARCH related model,weights as inputs in reverse optimization are generated by EGP model,this strategy is compared against the one that directly uses market capitalization as inputs,and it turns out that this strategy performs better than the latter in two different stages.So this can act as a reference for investors in their asset allocation.This thesis can be divided into six parts.Chapter 1 shows the background of the research,briefly introduces the routine and highlights of this thesis,and points out the significance of this study.Chapter 2 does a summary of literatures domestic and abroad,discusses portfolio optimization theory,Capital Asset Pricing Model,concentrates on former researches concerned about asset allocation and Black Litterman model,and summarizes their characteristics,points out three gaps between former researches and this thesis.Chapter 3 and 4 are the theoretical foundations,Chapter 3 begins with mean variance model,analyzes Black Litterman model in depth,gives a derivation from a Bayesian perspective,meanwhile we talk about the various inputs of BL model,focusing on weights in the inverse optimization,views and its variances,and then this chapter corrects a mistake in some researcher's work.At the end of Chapter 3,an asset selecting method based on factor model(EGP model)is introduced.Chapter 4 is mainly concerned with econometric model estimating views and its variances.Chapter 5 is empirical,first it talks about the data,analyzes data using basic statistics,then builds up BL model,derives the weights used in the reverse optimization,and determines the items and their lag orders in GARCH related models,and then the chapter predicts the views and variances,implements back tests during year 2015 and 2016 respectively,at last it calculates some statistics of the returns,compares two strategies from several aspects.The main conclusions are:1,for weights in reverse optimization,it would be more appropriate to generate them in the framework of capital market equilibrium,methods that only take variances into account have not been fully explained yet;2,asset selecting methods based on factor model can reflect historical return in the weights it derives,and posterior return in BL model lies between the views and the implied returns;3,the BL model with weights given by asset selecting model based on factor model performs better than the one with market weights in the sense of cumulative returns,stability of returns,and fluctuation of weights.