Naslov
A Test of Portfolio in Factor Space in the Croatian Equity Market

Autori
Zoričić, Davor ; Dolinar, Denis ; Lovretin Golubić, Zrinka

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Book of Abstracts 17th International Conference on Operational Research KOI 2018 / Arnerić, Josip ; Čeh Časni, Anita - Zagreb, 2018, 113-114

Skup
17th International Conference on Operational Research (KOI 2018)

Mjesto i datum
Zadar, Hrvatska, 26.09.2018. - 28.09.2018

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
efficient portfolio, covariance estimation, PCA method

Sažetak
Recent research for the Croatian stock market has shown that it is hard to construct a portfolio which can outperform the existing benchmarks (CROBEX and CROBEX10), regardless of their inefficiency. Research findings point out that failure of such efforts is mostly related to poor out-of-sample estimation of stocks’ expected returns. However, the need to improve covariance matrix estimation has also been reported. This paper focuses on analyzing potential improvements to (co)variance estimation by examining stock portfolio in factor space. The key benefit of such approach to analysis is that it enables the use of PCA (Principal Components Analysis) method which has been shown to improve covariance matrix estimation in the developed markets by decomposing covariance matrix and exposing the unobservable underlying risk factors from data. Observation period in this paper covers the period beginning in March 2005 and ending in September 2017 with a total of 25 regular revisions of CROBEX index. Therefore, the analysis is carried out based on 25 rolling window time samples of monthly excess total returns matching the CROBEX index composition in each sample. Relying on the principal components which capture most of the in-sample variability, out-of-sample covariance matrix and corresponding GMV (Global Minimum Variance) portfolios are estimated. Thus we obtain a time series of GMV portfolio returns based on which its performance is compared to CROBEX index. We find that unlike in the developed markets, the PCA method does not seem to improve covariance estimation. Compared to earlier research based on the statistical shrinkage method the estimated out-of-sample GMV portfolios performed worse (measured by Sharpe ratio) and more importantly exhibited an increase in volatility. The research results corroborate the view that there is too much noise in the data in illiquid and undeveloped markets. Therefore, even simple sample covariance estimation yielded better and more robust results.

Izvorni jezik
Engleski

Znanstvena područja
Ekonomija

POVEZANOST RADA