Ацканов_резюме_ENG (1137943), страница 3
Текст из файла (страница 3)
Investorsand portfolio managers, when making up an investment portfolio in the first place, tend to beguided by the perspectives of these two groups of stocks, preferring one over another(Bourguignon & De Jong (2003), Chan & Lakonishok (2004)). Other styles have also beenstudied quite well. One may find research on momentum, quality or dividends in some of thefollowing papers – Jagadeesh & Titman (1993), Chan et al. (1996), Chan et al. (2000), NovyMarx (2013), Asness et al.
(2013), Bouchaud et al (2016). At the same time, as noted earlier,studies that link the characteristics of the stock’s investment style with its weight in the portfolioare quite a few. Of these, one can check Brandt et al (2009), Hjalmarsson & Manchev (2012),Flieberg et al. (2016), Fletcher (2017).
It is important to emphasize though, that in these worksthe issue of asset risk is either completely omitted or taken into account only partly (Flieberg etal. (2016), Fletcher (2017)), without considering possible interrelationship of asset returns – oneof the problems that has been solved in this paper by applying CVaR, calculated on the basis ofsimulations on Vine copulas.Methodology of the research and dataThe research actively uses methods of financial analysis, econometric analysis and portfoliotheory. The key tool for the study is the modification of copulas - Vine copulas, which allowsbuilding multidimensional joint distribution to identify the relationship of asset returns and toassess their combined risk. For comparison of stylized portfolios, graphical analysis and variouscoefficients of portfolio management efficiency are also used.
Data collection was implementedin the Bloomberg API and Microsoft Excel. Data preprocessing is implemented in the Pythonprogramming language in the Spyder environment, while the main calculations are performed inthe Matlab environment.Scientific innovation of the studyThe novelty of this study is justified by the combination of a number of research fields –investment styles, portfolio theory, implementation of characteristics to portfolio weightsestimation, risk modeling with copulas. This study:• Develops procedures for optimization of the investment portfolio under regulatory constraintsand in application to the specifics of the Russian stock market.• Uses a relatively new method of Vine copula; its application to portfolio optimization with alarge number of assets presents some value, especially when applied to the Russian market.• Puts and solves the problem of investing in style in new light - through the optimization of thestylized investment portfolio by risk, the level of interrelation and the level of compliance withparticular investment style of each asset• Compares the behavior of various stylized portfolios on the Russian stock market anddetermines the most efficient in terms of various coefficients.Theoretical significanceTheoretical significance is built from the following sequence of results.
First, a procedure wasdeveloped and its advantage is proved for optimizing the investment portfolio using copulas withrestrictions on the weight of the asset.Secondly, the problem of stylized optimization of the stock portfolio is posed. Stylizedoptimization allows obtaining risk-balanced portfolios that best match the specific investmentstyle. Special utility functions were developed for that purpose, connecting weights in theinvestment portfolio with characteristics of each stock. This makes the investment process moreflexible with respect to investor preferences and adds transparency to the strategy.Thirdly, procedures have been developed for optimizing stylized portfolios of stocks for differentinvestment styles-cost, growth, profitability, dividends, and momentum.Finally, the behavior of stylized portfolios has been compared under different market conditions.Practical significance.The results of this research can be applied in practice by both professionals (portfolio managers,traders, risk managers) and private investors.
The proposed procedures for optimizing stylizedportfolios will allow market participants to design optimal investment portfolios in accordancewith the preferred investment style, and use the results to better understand the advantages of aparticular style in different market conditions. The last point still requires additional researchthough.Approbation of the resultsThe main results of the research were presented at the conferences: "The Third RussianEconomic Congress (REC-2016)", XVII April International Conference (Session Ea-13). Theresults of the research were also discussed at research seminars on economics at the HigherSchool of Economics.PublicationsThe main results of this study are presented in three articles in the Russian journals "AppliedEconometrics", "Finance and Credit" and "Financial Management".
These journals at the time ofpublication are peer-reviewed scientific journals recommended by the Higher AttestationCommission for publication of the main scientific results of the Ph.D. thesis.The structure of the paperThe paper consists of an introduction, three chapters, a list of used literature and appendix. Thetext of the dissertation 147 pages long, contains 17 figures, 15 tables.
The bibliography includes135 sources.2. The main points of the dissertationClassification of different investment styles occurs not only in the scientific papers, but alsowhen certain professionals in the asset management industry define their strategy.One may find a significant number of investment funds implementing particular investment styletargeted at a specific range of investors. These funds can be both passive (for example, ETFfunds of companies such as Black Rock Inc., The Vanguarg Group), and active (for example,funds of Cornerstone Advisors, MFS Investment, AQR Capital, Wisdom Tree and others). Themost widespread investing styles are value, growth, profitability, dividends, and momentum.
Atthe same time, the overwhelming part of the literature on investment styles concentrates mainlyon the criteria of classification and relative profitability and ignores the issue of constructing anoptimal stylized portfolio, the weights in which are linked to the characteristics that determinethe correspondence of each of the stocks to particular investment style.In this paper, equity portfolios of 5 investment styles are considered:• Growth stocks are generally defined as stocks of companies, which market value may seem tobe overvalued relative to current company’s fundamentals.
At the same time, this apparentdiscrepancy can be explained by the fact that the expectations of company’s profit growth in thefuture are significantly higher than the average growth rates for the market/sector /country. Inthis study, the price-to-book multiple is used to identify growth stocks. Among the papers usingthis multiple for defining growth stocks, some works use this multiple to define value stocks aswell – Graham & Dodd (1934), Fama & French (1998), Bauman et al (1998), Fama & French(2007), Bragg (2007), Bodie et al (2009), Pinto et al, (2010)).• Value stocks are defined as shares of companies which market value does not fully reflect thecompany's financial performance in comparison with the average of the market / sector / country.That is, the company is undervalued compared to others and therefore has a greater growthpotential.
Value stocks are often opposed to growth stocks. Bourguignon & De Jong (2003) notethat the most common way of classifying shares by investment style is using multiples. Price-tobook multiple is used in this study to define values stocks, the lower the multiple the better astock suits value investment style (Graham & Dodd (1934), Fama & French (1998) , Bauman etal (1998), Fama & French (2007), Bragg (2007), Bodie et al (2009), Pinto et al, (2010))).• Momentum stocks in this paper are defined as stocks, with the largest change in the price over acertain period of time. These are market leaders which, through their historically high returns,attract the attention of new investors who expect that their capitalization will continue to grow.• Profitability stocks in this study are defined as stocks of companies with highest financialindicators such as profitability or return on capital.
In this study, the ROE coefficient is used toselect such stocks (preference is given to ROE, since there are more data available on it).• Dividend stocks in this study are defined as stocks with the highest current dividend yield. Thequestions of the predictive power of the dividend yield in relation to the total return on stockswere investigated among others by Blume (1980), Visscher & Filbeck (2003), Connover et al.(2016).
Although investors in dividend stocks in addition to dividend yield usually pay attentionto other characteristics, due to the limited data on the Russian market, a simplified approach hasbeen chosen allowing covering a longer period of time for analysis.As mentioned earlier, copulas are used in this work to optimize weights of the investmentportfolio.
They allow tracking the interrelation of asset returns and assess their joint risk, whichcan be used to build a portfolio with the optimal structure and an appropriate ratio of expectedreturn to risk. This study uses one of the latest copula modifications to track the change in therelationship between assets over time, which is quite important in the case of financial timeseries.In this study modeling of copulas is divided into 2 stages (1) Construction of uniform marginaldistributions of daily logarithmic returns of financial instruments and (2) Construction of jointdistribution of returns of financial assets based on the uniform marginal distributions obtained onthe first stage.For the first stage, one needs logarithmic returns of assets, which requires a number of additionaltransformations of the data.
To model the marginal distributions of returns of each stock, thefollowing model is used:=μ() + σ(ε |Where) × ε ; i = [1, N];~∈(0,1)∀is the daily logarithmic return of asset i at time t;μ() is the mathematical expectation of the return on asset i based on the data availablebefore the time t;σ() is the corresponding standard deviation of the daily logarithmic return on asset i,based on the data available before time t;ε is the corresponding error;is the set of information available before time t.Mean is modeled in ARMA, while GJR-GARCH is used to model the standard deviation. Thesemodels are further used to calculate standardized error:ε =[−μ (; α)]/σ(; α)Where α is the vector of optimal parameters for the models of mean and standard deviation.As the distribution of the residuals is unknown, one needs to either make an assumption abouttheir distribution or use a semiparametric model. In the case of this study the most of the data isstock prices.
Stock prices commonly tend to change their statistical characteristics over time. Inaddition different price shocks tend to happen from time to time which hampers making correctassumptions about their distribution. Taking that into account it is logical to utilizesemiparametric copula model, which uses empirical distribution function for modeling of theresiduals. One may also assume that from the study by Kim & Silvapulle (2007), who prove thatfully parametric models are non-robust against misspecification of marginal distributions andsemiparametric models are better overall.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
