N.Е. Egorova, A.R. Bahtizin, K.A. Torzhevsky Modeling of dynamics of the Russian stock market: macroeconomic parameters and the market's long-term memory length factor
Modeling of dynamics of the Russian stock market: macroeconomic parameters and the market's long-term memory length factor
N.E. Egorova,
Doctor of Economics science, Professor, Chief member of staff of The Central Institute of Economics and Mathematics of Russian Academy of Science (CEMIRAS) (Russia, Moscow, 117418, Nahimovskiy prospekt, 47; e-mail: nyegorova@mail.ru)
A.R. Bahtizin,
Doctor of Economics science, Chief member of staff of The Central Institute of Economics and Mathematics of Russian Academy of Science (CEMI RAS) (Russia, Moscow, 117418, Nahimovskiy prospekt, 47; e-mail: albert@cemi.rssi.ru)
K.A. Torzhevsky
Post-graduate of The Central Institute of Economics and Mathematics of Russian Academy of Science (CEMI RAS) (Russia, Moscow, 117418, Nahimovskiy prospekt, 47; e-mail: me_23@mail.ru)
Аннотация. На основе синтеза фундаментального и технического подходов исследуются динамические характеристики российского фондового рынка, и разрабатывается инструментарий его прогнозирования. Динамика индекса РТС считается зависящей как от макроэкономических индикаторов (рост ВВП, цены на нефть и др.), так и от инерции (памяти) рынка. Трендовые закономерности совместной динамики рассматриваемых индикаторов выявляются с помощью методов технического и эконометрического анализа, а также нейросетевого моделирования. Работа выполнена при поддержке гранта Российского фонда фундаментальных исследований проект № 08-06-00163.
Abstract. Research on Russian stock market's dynamic characteristics and set of tools for its successful forecast is based on both fundamental and technical analysis. RTSI dynamics is considered dependent on mac-roeconomic indicators (GDP growth, oil prices, etc) and on market's inertia (so-called market's memory). Trend patterns of combined dynamic indicators are revealed through methods of technical, econometric analysis and neural networks modeling. Work is executed at support of the grant of the Russian Fund Fundamental Research, the project № 08-06-00163.
Ключевые слова: индекс РТС, индекс ВВП, волны Эллиотта, нейронные сети Keywords: RTS index, GDP index, Elliott's wave, neural networks
Purpose of research is to develop economic-mathematical toolbox for stock markets analysis and forecast (by the example of Russian stock market). Unlike the most current achievements on related problems, this project suggests synthesis of fundamental and technical approach to market analysis. Specific problem is to confirm (deny) the following hypothesis:
1. In the context of fundamental approach -influence on market dynamics: a) basic macroeco-nomic indexes (GDP growth, oil price, us dollar rate, export-import balance, etc); b) international stock markets indexes (Dow Jones, NASDAQ, Nikkei, etc).
2. In the context of technical approach -fractal structures appearance in combined dynamics of RTSI and considered macroeconomic indicators (based on this ground, market's long-term memory length idea is proposed).
3. In the context of both approaches - dependence of RTSI from the set of considered economic indicators and market's memory.
Paper topicality comes on the first place from synergy of different approaches and on the second place - in relatively weak knowledge of Russian stock market.
Practical usage
Developing toolbox can be used on the state level (in regulatory and forecasting processes) and also in stock-trading strategies for medium- and long-term investment (1-2 years). Although, there are numerous research projects on that topic (especially in foreign tech analysis issues [3, 5, 9, 17-20]), however final solutions are far from achievement. Instrumental toolbars are widely used in intraday trading (so-called mechanical or automated trading systems). As for strategic market forecast - they are typically useless, which can be explained by market's complexity as a system itself. Moreover, methodology disadvantage for most forecasts is lack of
macroeconomic indicators consideration over technical approach prevalence. That came evidently clear especial after fall of 2008 year (downside on US stock markets against a background mortgage crisis and huge public debt). Project takes part in developing more complex and valid stock market forecast.
Used materials review
Among foreign authors that studied these topics we can point to famous researches of developing fractal market conception [7, 12]; using Elliott's wave principle [4, 10, 14]; improving technical analysis methods [3, 5, 9, 17-20]; advised to use modeling approach to reflect the emerging markets progresses and others.
Domestic research is presented 1) theoretical issues regarding the stock market as complex nonlinear and no equilibrium system [1, 6, 15]; 2) application study by Eliot's wave's mathematical modeling [2, 8]; fundamental analysis [11, 16]; technical analysis and developing mechanical trading systems [13] and others.
In contrast of studies by [8, 13], this research is not limited with Russian stock market technical analysis and it is not grounded on Elliott's wave concept (in the judgment of the authors, Russian market is too young for full-scale Elliott-waves application). In comparison to [11] research on Russian stock market's fundamental analysis, dedicated to examine the correlation between RTSI-GDP growth we can point four following basic differences: 1) usage of modified approach, composed of studying the correlation between the relative economic indicators - RTSI and GDP growth index and capable of revealing trends in certain economic indicators and eliminate the effect of their different measurement; 2) analysis period was expanded greatly - September 1995 year - May 2008 year, which also helped to receive more reliable results; 3) statistical technology was combined with visual analysis methods,
which are often used in stock markets dynamics research; 4) visual-statistics methods were amplified by neuronal net modeling practice.
Conclusions on this research received so far indicates alternate movement phases in dynamics of figures under consideration, which shows impossibility of precise market prediction based on GDP index and inconsistency with results [11].
Methodology of analysis
A. Data. RTS index (basic and industrial) data from different time frames was downloaded from www.rts.ru and used for neuronal network education and building the econometric model:
1) basic RTSI from 09.1995 till current date, calculated from 50 major Russian companies stock value, among them are: PC (public corp.), AFK Sistema, PC GAZPROM, PC NK Rosneft, PC Sber-bank Of Russia and others;
2) industrial RTS index - "Oil and gas" from 12.1999 till current date, calculated from 12 major Russian oil and gas sector companies share price;
3) industrial RTS index - "Metals and mining" from 12.2003 till current date, calculated from 13 Russian major sector companies share price;
4) industrial RTS index - "Consumer goods" from 12.2004 till current date, calculated from 12 Russian major sector companies' share price;
5) industrial RTS index - "Industry" from 12.2003 till current date, calculated from 10 Russian major sector companies' share price;
6) industrial RTS index - "Telecommunications" from 12.1999 till current date, calculated from 12 Russian major sector companies' share price;
7) industrial RTS index - "Finance" from 01.2005 till current date, calculated from 9 Russian major sector companies' share price;
8) industrial RTS index - "Power industry" from 01.2005 till current date, calculated from 16 Russian major sector companies' share price;
For data comparison to RTSI following indicators were used: GDP, oil price, foreign trade turnover, import and export, measured in us dollars and in total volume of investments. These indices were defined by Russian Federation Ministry of Finance (www.minfin.ru) and Russian Statistics Ministry (www.gks.ru).
To calculate the correlation factor, setting econometric equations and building neural networks, both chained and cumulative (in this case basic period equals 100%) indices were used.
B. Theoretical models. Two types of economic mathematical models were used: 1) econometric; 2) neural networks. Exploratory design shows that despite its acceptable statistical activities, classical econometric models are not suitable to forecast trends in conditions of significant market correction and trend inversion. From that point of view it is essential to upgrade these models, for example, by introducing additional multiplier, reflecting market's "memory".
Neural network contains data on previous market dynamics as input parameters and is capable to reflect the market trend inversion (as during the modeling of autumn 2008 situation). But it is unable to take into account the impact of macroeco-nomic indicators. Meanwhile, market dynamics is determined (in the judgment of authors) by complex superposition of macro factors and system memory, which also requires certain modifying of developed neural networks toolbox. Upgrading and modifying of models designed by authors, its comparison and acceptability appraisal for market forecast - Is forthcoming project's outcome.
A research principle consists in:
I. For econometric models.
1. Based on correlation coefficients significant factors are selected (economic factors) Xi
(i = 1, n), influencing RTSI (Y), where n - is number of picked factors.
2. Empirical dependence charts are built as
y = f (t) and Xi = <Pi(t) (1)
and it's visual analysis performed in order to discover fractal structures and long-term market memory.
3. Inertia link is designed n (Zj ), j = 1, m , which is included as a component part into following econometrics model:
y = F, (x) + n(z) (2)
Supposed to test system memory, RTSI amplitude variation, Hurst index (showing market tendency to follow its
Для дальнейшего прочтения статьи необходимо приобрести полный текст. Статьи высылаются в формате PDF на указанную при оплате почту. Время доставки составляет менее 10 минут. Стоимость одной статьи — 150 рублей.