Bayes-Stein Estimators in Mean-Semivariance Portfolio Optimization
Tcysin, Aleksei; Pivovarov, Dmitrii (2017)
Lataukset:
Tcysin, Aleksei
Pivovarov, Dmitrii
Lapin ammattikorkeakoulu
2017
Creative Commons Attribution 3.0 Unported
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2017111617182
https://urn.fi/URN:NBN:fi:amk-2017111617182
Tiivistelmä
The aim of this thesis was to design and implement a semivariance-based portfolio optimization model with application of shrinkage estimators.
Harry Markowitz’ Modern Portfolio Theory served as a basic theoretical framework; it is further extended by using semivariance computational procedure proposed by Javier Estrada and adjusting the vector of expected returns with Bayes-Stein estimator, suggested by Philippe Jorion.
Back testing was applied in order to check the performance of the suggested scheme. Investment strategy was tested on 30 stocks representing Dow Jones Industrial Average; minimum-risk mean-variance portfolio and S&P 500 index were used as performance benchmarks. Three sets of tests were conducted to check the model in various market conditions.
Results indicate that the proposed framework is somewhat successful, outperforming both benchmarks in bullish and stagnant environments. However, further experiments under various conditions and parameters are necessary before utilizing the suggested approach in practice. Therefore, future work concerns mostly testing and evaluation routines.
Harry Markowitz’ Modern Portfolio Theory served as a basic theoretical framework; it is further extended by using semivariance computational procedure proposed by Javier Estrada and adjusting the vector of expected returns with Bayes-Stein estimator, suggested by Philippe Jorion.
Back testing was applied in order to check the performance of the suggested scheme. Investment strategy was tested on 30 stocks representing Dow Jones Industrial Average; minimum-risk mean-variance portfolio and S&P 500 index were used as performance benchmarks. Three sets of tests were conducted to check the model in various market conditions.
Results indicate that the proposed framework is somewhat successful, outperforming both benchmarks in bullish and stagnant environments. However, further experiments under various conditions and parameters are necessary before utilizing the suggested approach in practice. Therefore, future work concerns mostly testing and evaluation routines.