Portfolio Selection by Using Time Varying Covariance Matrices
JOURNAL OF ECONOMIC & SOCIAL RESEARCH, cilt.9, sa.2, ss.1-22, 2007 (Hakemli Dergi)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 9 Sayı: 2
- Basım Tarihi: 2007
- Doi Numarası: 10.5281/zenodo.23183572
- Dergi Adı: JOURNAL OF ECONOMIC & SOCIAL RESEARCH
- Sayfa Sayıları: ss.1-22
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- İstanbul Üniversitesi Adresli: Evet
Özet
Markowitz mean-variance portfolio theory is one of the most widely used approaches in portfolio selection. Since Markowitz portfolio theory uses equally weighted data, it does not exhibit the current state of the market. It reflects market conditions which are no longer valid by assigning equal weights to the most recent and the most distant observations. To express the dynamic structure of the market, one can use exponentially weighted variances. Exponentially weighted data gives greater weight to the most recent observation. Thus, current market conditions are taken into consideration more accurately. Additionally, to handle the dynamic structure of the volatility in the market, generalised autoregressive conditionally heteroscedastic models can be employed to estimate the covariance matrix.
This paper presents the use of exponentially weighted moving averages and generalised autoregressive conditional heteroscedasticity techniques in portfolio selection. The security variances and the covariance term between each security are calculated using exponentially weighted and GARCH(p,q) schemes. In addition, equally weighted, exponentially weighted and GARCH(p,q) schemes are used for security returns from the XU030 index and portfolio risk parameters at a certain level of expected return are compared. Deviations from the Markowitz mean-variance portfolio theory are investigated.